The LARPs, La and related RNA-binding proteins: Structures, functions and evolving perspectives

Abstract

La was first identified as a polypeptide component of ribonucleic protein (RNP) complexes targeted by antibodies in autoimmune patients and is now known to be a eukaryote cell-ubiquitous protein. Structure and function studies have shown that La binds to a common terminal motif, UUU-3’OH, of nascent RNA polymerase III (RNAP III) transcripts and protects them from exonucleolytic decay. For precursor-tRNAs, the most diverse and abundant of these transcripts, La also functions as a RNA chaperone that helps to prevent their misfolding. Related to this, we review evidence that suggests that La and its link to RNAP III were significant in the great expansions of the tRNAomes that occurred in eukaryotes. Four families of La-related proteins (LARPs) emerged during eukaryotic evolution with specialized functions. We provide an overview of the high resolution structural biology of La and LARPs. LARP7 family members most closely resemble La but function with a single RNAP III nuclear transcript, 7SK or telomerase RNA. A cytoplasmic isoform of La protein as well as LARPs 6, 4 and 1 function in mRNA metabolism and translation in distinct but similar ways, sometimes with the poly(A)-binding protein (PABP), and in some cases by direct binding to poly(A)-RNA. New structures of LARP domains, some complexed with RNA, provide novel insights into the functional versatility of these proteins. We also consider LARPs in relation to ancestral La protein and potential retention of links to specific RNA-related pathways. One such link may be tRNA surveillance and codon usage by LARP associated mRNAs.

Graphical Abstract

Introduction and Background

Brief overview including general properties of La and related proteins

The La motif (LaM) was established as a protein fold coincident with the emergence of the Eucarya and became readily associated with a downstream RNA recognition motif (RRM), together comprising the La module¹ (Fig. 1). The La module in turn became allied to additional motifs as members of the major families of La-related proteins, LARPs 7, 6, 4 and 1 emerged and diversified with specialized functions in the eukaryotes^1–3. Developments in basic and in some cases clinical related research on La and LARPs over the past few years have advanced our understanding to a point where we can begin to appreciate how commonality and individuality contribute to their unique functions.

Figure 1. Schematic representation of architectures of La proteins and selected LARPs.

The La motif (LaM) and RRM comprise the La module. The PDBs used to depict the three dimensional structures for the La module and hLa RRM2α are 2VOD and 1OWX respectively^{19, 106}. Other symbols refer to the following; N: nuclear localization (import) sequence^{28, 29, 33, 45, 298} E: nuclear export sequence^{24, 27, 29}, R: nuclear retention element^{24, 27}, S: Serine-366^{49, 195}, SBM: short basic motif important for recognition of 5' pppG of nascent RNA and nucleolar localization^{25, 32, 49}, DM15: important for direct binding to 7mGpppC-cap-5'TOP motif¹³⁴ The RRM2α, xRRM and DM15 RNA-interaction motifs/domains are reviewed in a separate section. PAM2: poly(A) binding protein interaction motif-2, PAM2*: LARP1-associated PAM2 candidate with atypical features (see text ), LSA: LaM and S1-Associated motif, PBM: poly(A)-binding protein interaction protein motif, RIR: RACK1-interaction region. EI4L in LARP1 refers to eIF4-like region⁶. The different species' La proteins depicted are referred to in the text; for simplicity, only the human versions of the LARPs (hLARP) are shown.

LARPs encompass a range of functions. Some have adopted functions specific to a limited set of RNA target ligands, for example LARP7 members for specific small nuclear (sn)RNAs. Accumulating evidence indicate that others can coregulate subsets of mRNAs that produce functionally related proteins, e.g., LARP1 for the 5'TOP mRNAs that encode ribosomal proteins as well as other mRNAs^4–6, according to the post-transcriptional RNA regulon model^{7, 8}. Human LARP6 (hLARP6) targets a conserved stem-loop (SL) motif found in mRNAs that encode α-collagen I and III to coordinate their translation^{9, 10}. LARPs 4 and 4B appear to coordinate somewhat larger sets to promote stability and translation of their mRNA targets^11–13.

Due to space limitations we cannot review all LARP progress nor offer highly detailed critiques. Our goal is an overview preceded by adequate background to provide what we believe are emerging conceptual advances. Central to the LARPs is the La module RNA binding unit, to which other protein motifs have been added in a LARP-specific manner. These include other RNA recognition motifs and/or motifs that interact with other proteins, some of which are common to multiple LARPs. This has led to a model in which the La module has adapted, structurally and functionally, to befit the LARP-specific function at hand. This overview contains sections focused on each of the LARP families. However, several themes are generally applicable and relevant information in other sections may not be referred to by LARP-specific nomenclature.

Genuine La proteins (Fig. 1) exhibit sequence and length specificity for UUU-3’OH, the terminal motif common to transcripts synthesized by RNA polymerase (RNAP III)¹⁴. Studies on La proteins from protists, yeasts, frogs, mouse and human cells indicate that its function is to protect nascent RNAP III transcripts from untimely 3' exonucleolytic digestion and for some of these it also helps prevent their misfolding by virtue of a separate chaperone activity ^{reviewed in 2, 15}. Although general RNA chaperone functions have been attributed to the La modules of La and the LARPs¹⁶, independent lines of evidence suggest that motifs in the C-terminal regions of La proteins and some LARPs can also contribute to these activities¹⁷ including in mRNA translation¹⁸ (below).

Apparently early in eukaryal evolution a second RRM (termed RRM2) emerged downstream of the La module and this may have been the gene arrangement that gave rise to La and LARP7 members ^{see figure 5 in 1}. LARP7 family members are the closest relatives to genuine La in sequence, architecture and function. The RRM2 of human La (hLa) has an extra structural element that is appended onto the canonical RRM fold, namely the α3 helix that lies over its β-sheet surface¹⁹. Although α-helices located on top of the central β-sheet comprise one of the known variations of the RRM protein fold²⁰, the distinctive elements and functional significance attributed to the RRM2 α3 helix make it a distinguishing feature, as described in a later section; we shall hereafter refer to the RRMs2 of La and metazoan LARP7 as RRM2α. The RRM2α would appear to have been an ancient part of La as it is found in the proteins from extant representatives of phylogenetically deep rooted eukaryotes, Trypanosoma, Leishmania, Giardia and Dictyostelium species²¹ (Fig. 1 and below).

Figure 5. Comparison of the wing 2 motifs and interdomain linkers in LARPs.

A) Sequence alignment focusing on LaM wing 2, interdomain linker and the beginning of RRM1 of LARPs. Human (Hs) La sequence was aligned with 32 LARPs as for figure 4. Black vertical rectangles indicate conserved signatures characterizing wing 2; the dark red horizontal arrows denote the beginning of RRM1 strand β1 determined experimentally for human La, LARP7, LARP6 and LARP4 (^{9, 107, 110}, IC-G & MRC, unpublished). The potential extent of the variable linker regions is indicated with a dotted line above the sequences. B, C) Close-up views of the wing 2 region of hLa superimposed with hLARP7 (B) and hLARP6 (C). Selected residues involved in the interaction of wing 2 with the rest of the domain are highlighted as sticks. hLa (PDB 2VOS) is depicted in yellow, hLARP7 in light orange (PDB 4WKR) and hLARP6 in cyan (PDB 2MTF).

A later section will review structural details and functional attributes of p65, the LARP7 member in the protist ciliate, Tetrahymena thermophila Tth, in which this type of RRM2α, designated as xRRM, contributes to telomerase RNA folding and hierarchical assembly of the RNP complex^{22, 23}. However, La proteins which lack a RRM2α such as yeast Lhp1, contain a C-terminal extension that can assist pre-tRNAs in correct folding¹⁷ (below).

Thus, with two versions of both La and LARP7, each with and without a RRM2α as residents in early eukaryotes ^{see figure 5 in 1}, the evolutionary stage would have been set for emergence of additional LARP architectures. The genuine La proteins of human and yeast contain multiple intracellular trafficking elements that control nuclear, cytoplasmic and nucleolar distribution (Fig. 1)^24–28. Unlike La and LARP7, which are predominantly nuclear and associated with noncoding snRNAs, the LARPs 6, 4 and 1 appear mostly cytoplasmic at steady state as mRNA-associated proteins. It is noteworthy here that majority distribution at steady state does not preclude dynamic nuclear-cytoplasmic shuttling. Indeed, S. pombe and hLa proteins are mostly (~80%) nuclear but shuttle by virtue of conserved nuclear import and export elements, the latter of which map to the α-helical backside surfaces of their RRMs1, and are controlled by a conserved nuclear retention element in the α3 helix of human RRM2α and of a comparable region of RRM1 of S. pombe La (Fig. 1)^{19, 24–29}. This reflects on the versatility of the RRM as a multifunctional folding platform^{30, 31}. These elements constitute a conserved circuit of RNA-binding and subcellular trafficking elements^{32 see 33}.

Nucleo-cytoplasmic shuttling elements were also identified in hLARP6/Acheron³⁴. A classic Crm1-consensus nuclear export sequence that was functionally mapped^{34, 35} to residues that align to the β1 strand of the RRM of the La module¹ was indeed found as part of β1 in the solution structure of hLARP6⁹ ("E" in Fig. 1). Other LARPs also use RNA binding modes that are juxtaposed to subcellular trafficking elements, other interaction motifs and/or regulatory/signaling elements (below).

Another commonality among several LARPs is the interaction with poly(A) binding protein (PABP) which is clear for human LARPs 1, 4A and 4B, and Arabidopsis thaliana (At)LARPs 6b and 6c^{3, 11, 12, 36, 37}. As will be detailed below, existing and emerging data indicate that the La modules of hLARPs 1 and 4 and plant 6c exhibit preferential binding to poly(A) RNA itself^{3, 12} and for hLARP4B, A-rich RNA¹³. Therefore, a theme for the cytoplasmic LARPs is the association with the 3' end poly(A) region of their substrate mRNAs, via PABP and the mRNA 3' UTR or the poly(A) tail, in any case a polar orientation reminiscent of La and LARP7 on the UUU-3' OH ends of their snRNA ligands. From this another common theme follows, mRNA circularization, first proposed for LARP4B¹¹, mediated by simultaneous interactions with the 5' and 3' regions. We will review emerging data on how the different LARPs manage this. mRNA circularization promotes ribosome cycling and ensures that only intact mRNAs, i.e., with a 5’ m7GpppN-cap and 3’ poly(A) tail are efficiently translated, and this is commonly referred to as the closed loop model of translation^38–40.

The La protein La module and RRM2α existed in ancient eukaryotes and persists widely in extant species

Recent advances relevant to the phylogenetics of La came from two sources, a study that sought to uncover the ancestral gene repertoire of the most ancient animal stem cells⁴¹, and secondary structure examination of La protein from the amoebozoa Dictyostelium discoideum^{21, 42}, which represents one of the earliest known branches from the last common ancestor of all free living eukaryotes⁴³. D. discoideum La protein contains a typical La module comprised of a LaM and RRM1 as well as a RRM2α with a short β4 strand followed by an α3 helix^{21, 42}. The La protein of the intracellular parasite Trypanosoma brucei was also known to contain a typical La module^{44, 45}. The report of RRM2α in D. discoideum La prompted further examination of more phylogenetically distant La proteins of some the simplest eukaryotes of the Trypanasoma, Leishmania and Giardia species. We began with a JPred-4 sequence structure prediction analysis of the C-terminal region of T. brucei La. This predicted a characteristic RRM2α with an additional β strand following β4 (β4’) and an extended α3 (Fig. 2A), in agreement with the RRM2α of hLa whose high resolution structure has been determined⁴⁶ (Fig. 2B and see below). The protein sequences of other intracellular parasites, from two Leishmania species as well as two parasite, Giardia species were aligned with other La proteins and compared to the RRM2α primary and tertiary structure from hLa (Fig. 2C). The sequence mapping to hLa RRM2 helix α3 is enclosed in the orange-green box and its salient conserved residues are annotated by asterisks in Fig. 2C. Although not shown in Fig. 2C, the bona-fide La protein in Arabidopsis thaliana, AtLa1 also has a RRM2 (as well as La homologs in other plants)⁴⁷ bearing the key conserved residues in helix α3 (not shown).

Figure 2. RRM2α and SBM (short basic motif) regions of La proteins representing early eukaryotes.

A) Results of secondary structure prediction for sequence of putative RRM2 region of T. brucei La protein (GenBank: AAF34598.1) performed by Jpred-4²⁹⁹. The secondary structure elements are indicated below the sequence; α-helices depicted by red cylinders and β-strands by green arrows. B) Solution structure of hLa RRM2α shown for reference (PDB 1OWX) ¹⁹, with α3 colored to match underlying panel C: red to coincide with boundaries of red rectangle, magenta to match conserved positions demarcated by asterisks, and green to coincide with rectangle downstream sequence boundary in C. C) Alignment of La protein sequences beginning from the approximate middle of RRM2 extending toward their C-termini. The overlapping green and orange rectangles represent the sequences aligning with hLa α3 region of RRM2α as depicted in figure 6, and a conserved sequence block respectively. Asterisks above the hLa sequence demarcate conserved residues, the first three of which reside in the tract connecting β4' to the α3 helix, colored in magenta in the structure in panel B. The blue rectangle encompasses the short basic motif region described for hLa (see text).

The alignment also revealed that in several of the La proteins of the deep rooted eukaryotes, the RRM2α is followed by a conserved G/K/R rich region (Fig. 2C, blue rectangle), also conserved in AtLa1 (not shown). In hLa this region is termed the short basic motif (SBM, Fig. 1) which was proposed to share sequence features with a Walker A-like nucleotide-binding motif⁴⁸ and was later found to recognize the 5'-ppp triphosphate terminus of nascent pre-tRNAs and whose binding is attenuated by phosphoSer-366 to control 5' processing by RNAse P^{49, 50}.

It is notheworthy that several of the La proteins, including AtLa1 (not shown), contain multiple copies of the potential arginine methylation motif, GRR/GR⁵¹ in this region. We also note this part of hLa overlaps with the sequence recently found to be required for CCND1-IRES mRNA chaperone activity¹⁸.

Because Trypanosoma, Leishmania, Giardia and Dictyostelium are more deep rooted on the phylogenetic tree than are the yeasts^{see figure 5 in 43}, which lack a RRM2, these observations suggest that RRM2-GR/K existed widely in eukaryotes and that the yeast/fungal lineage lost their RRM2 (Fig. 1). However, the yeasts S. cerevisiae Lhp1 and S. pombe Sla1, as well as other La proteins have intrinsically unstructured regions in their C-terminal regions¹⁷. For Lhp1 the disordered sequence bears some resemblance to GR/GK region and was shown to function in posttranscriptional biogenesis of tRNA and other cellular RNAs¹⁷. For human La this disordered region predicted by the ISOPRED2 server¹⁷ and confirmed by NMR and Circular dichroism experiments¹⁹ includes the SBM GR/GK region and mRNA chaperone domain, RCD¹⁸.

The second phylogenetic observation is also related to Dictyostelium. D. discoideum can undergo regulated differentiation of homogeneous cells into distinct types, and many of the genes involved in the process were inherited by the Metazoa⁴³. Toward defining a relevant ancestral gene repertoire of animal stem cells, Alie and coworkers found that of the 44 conserved RNA binding proteins (RBPs) involved, La was among the 25 that were previously shown to be expressed in, and in the case of La, essential for the survival of⁵², mouse ES cells⁴¹. In the ES cell-essential group of 25 were a number of RBPs involved siRNA/piRNA and Piwi-related pathways⁴¹. In this regard it is notable that La in animal cell nuclei is important for preventing nascent pre-tRNAs with potential for folding into alternate structures from entering the miRNA processing pathway⁵³. These observations add support to the idea that ancient La was critical during early eukaryotic evolution and was fixed as a fundamental cellular component thereafter. As an ancient RBP with a complexity of RNA-binding and subcellular trafficking motifs, it should not be surprising that La function would extend to multiple aspects of RNA metabolism and that the La module (and RRM2α) would be appended to other motifs in the LARPs.

La is a Eukaryotic Pre-tRNA Chaperone that Supports tRNAome Diversity

La acts on RNAP III transcripts

La is an eukaryote-specific factor whose phylogenetic emergence is consistent with co-evolutionary appearance with RNAP III in an ancient common ancestor prior to eukaryotic radiation^{1, 54, 55}. The term tRNAome used here refers to the tRNA gene content of a genome. Most tRNA genes in bacteria and archaea are included in operons with other RNA types, all of which are under the control of a single RNA polymerase ^{see 56, 57}. By contrast, all eukaryotes divide transcription among three RNAPs, I, II, and III, with the tRNAs further distinguished as individual transcription units under the dedicated control of RNAP III. The large ribosomal RNA is synthesized by RNAP I while RNAP II synthesizes the mRNAs, as well as long and short noncoding regulatory RNAs. RNAP III synthesizes only short RNAs and does so with high efficiency for its most abundant products, the tRNAs^{58, 59}. For example, the S. cerevisiae RNAP III transcriptome is limited to just seven or so types of genes for conserved small RNAs; U6 spliceosomal snRNA, scR1 small cytoplasmic RNA of the signal recognition particle (SRP), the RPR1 RNase P RNA, NME1 RNase MRP RNA, snr52 a small nucleolar RNA, the 5S rRNA and the tRNAs (and a few tRNA-like transcripts, TLTs1–6, of unknown function)⁶⁰. Except for the tRNAs and 5S rRNA, the others are single copy genes that encode RNAs that undergo minimal if any processing as compared to pre-tRNAs and become stably associated with one or more polypeptides in their corresponding RNPs. By contrast, S. cerevisiae tRNAs are synthesized from 275 individual genes that are widely dispersed to all chromosomes and occupy the great majority of RNAP III^60–62.

Nascent pre-tRNAs undergo the most complex processing and maturation process of the RNAP III transcripts, requiring multiple cleavage and modification activities. tRNA biogenesis is further elaborated by their collective sequence diversity and potential to form alternate structures yet they must compete for several shared processing activities and other limiting factors which include the first protein they interact with, La, and certain tRNA modification enzymes^{59, 63–66}.

With seventeen integral subunits, RNAP III is specialized for efficient termination-reinitiation recycling and high production of short RNAs^{67, 68}. In all eukaryotes examined, the critical control element for transcription termination of RNAP III-transcribed genes is the oligo(dT) tract following the coding sequence, which in the nascent RNA becomes the high affinity UUU-3'OH motif to which La binds. Three RNAP III-specific subunits, C53, C37 and C11 contribute to termination and efficient reinitiation^{65, 68–70}. The C11 subunit is a RNAP III-intrinsic RNA 3' cleavage factor that can impose 3'-oligo(U) length differences on terminated transcripts⁶⁵. Thus, termination is a critical feature of RNAP III and La is mechanistically directed to survey the termini of its newly synthesized transcripts¹⁵. La binding is oligo(U) length-dependent^{14, 65} and on this basis, it sorts or channels nascent RNAP III transcripts to different processing and maturation pathways ^{reviewed in 15}. In fission yeast, La is limiting relative to the pool of nascent RNAP III transcripts so the efficacy with which a pre-tRNA competes for it is dependent on the length of its oligo(U) tract and abundance⁶⁵. This was demonstrated in vivo by genetic mutations in C11 that cause increase in the 3' oligo(U) length of the nascent transcript for a suppressor-tRNA⁶⁵. By increasing 3' oligo(U) length, a pre-tRNA can be more efficiently bound by La, more efficiently processed to mature tRNA, and lead to higher levels of tRNA-mediated suppression⁶⁵. Thus although La is very abundant, it can be functionally limiting, because the rate of nascent transcript synthesis by RNAP III is exceedingly high⁷¹.

While La interacts with UUU-3'OH on the spectrum of RNAP III transcripts, its most functional impact is on the pre-tRNAs^{63, 72}. This distinction reflects that La has two activities that confer biological function, 3' end protection from exonucleases^73–75 and RNA chaperone² the latter of which differentially benefits pre-tRNAs. Genome-wide screens in yeast uncovered multiple tRNA genes in which single nucleotide substitutions cause cells to require La for growth^{17, 72, 73, 76 see 63}. To our knowledge no gene for a non-tRNA RNAP III transcript has been uncovered this way although mutations in genes encoding proteins involved in U6 snRNP assembly have been identified^{77, 78}. That La can shield scR1 and other non-tRNA RNAP III transcripts from untimely decay reflects its 3' end protection activity^{75, 79}. Also, while pre-5S rRNA is the next most abundant RNAP III transcript after tRNAs, it undergoes little end processing and is largely independent of La ^{e.g., see 53}. Thus, pre-tRNAs are the RNAP III genes whose transcripts would appear to be the critical beneficiaries of La chaperone activity.

La function extends beyond 3' end protection for pre-tRNAs by preventing misfolding of those with propensity to form alternate structures^{17, 63, 72, 76, 80}. Structurally-challenged pre-tRNAs succumb to nuclear surveillance-mediated decay in the absence of La^{73, 74, 80}. The RNA chaperone activity is important for yeast growing at low temperature, due to insufficient amino acid charging of one or more susceptible tRNAs in the absence of La⁷². As reviewed above, some tRNA gene alleles have been found to be conditionally dependent on La for efficient maturation of their transcripts.

Molecular chaperones can mask mutations in their substrates

Chaperones can buffer mutations in their substrates and act as capacitors by allowing mutation-bearing gene alleles to exist as a means to evolutionary change ^{81, 82}. The highly abundant molecular chaperone, Hsp90 assists nascent polypeptides in avoiding "off-pathway" folding events and in achieving their correctly folded form. Beyond this important function, it has been established that Hsp90 additionally serves as a capacitor of genetic variation that can affect evolutionary change by buffering polymorphisms⁸³. This results from the abundance and power of Hsp90 chaperone activity such that it can mask mutations in its substrate polypeptides that would otherwise disrupt their correct folding. Any number of these polymorphisms can accumulate throughout the genome in a variety of independent genes because they are silent under the protective activity of the chaperone. However, when future conditions change including when the amount of Hsp90 is decreased relative to its substrate load, some of these mutations will be unmasked and if beneficial (or detrimental) may become deterministic toward phenotype⁸³. Thus, the chaperone acts as a “capacitor,” serving to store the genetic variations and release them⁸³. This has indeed been observed in wild populations of genetically distinct Drosophila; as Hsp90 is debilitated, different phenotypes emerge in different populations due to unmasking of a number of polymorphisms in otherwise unrelated genes⁸⁴, linked by the fact that they share the same chaperone.

La is a chaperone that has been reproducibly shown to mask mutations in tRNAs that would otherwise disrupt their structure and lead to their decay^{44, 65, 72–74, 80}. La can also mask mutations in tRNA modification enzymes, without which their tRNA substrates fail to accumulate^85–87. It is therefore reasonable to consider that La may have a similar but unique relationship with tRNA genes as Hsp90 has with its substrates, unique because of the natural high copy number, inherrent variability and propensity for amplification of tRNA genes. As tRNA gene number appears to be dynamic in eukaryotes (below) the potential storage of variation could be large.

tRNAome gene expansions in eukaryotes

Comprehensive analysis found that the only eukaryotic group lacking a genuine La homolog comprised the intracellular parasites, Plasmodia¹. Most Plasmodium species have only 40–60 tRNA genes, and the rest a maximum of 78⁸⁸, similar to many bacteria and archaea which typically contain 45–85⁸⁹. By comparison, free-living eukaryotes such as yeast contain 200–300 tRNA genes while plants and animals typically contain 400–700, although quite a few contain several thousand⁸⁹. Thus, tRNA gene numbers can vary very widely among related species⁹⁰ and even among members of a species^91–94. tRNA genes number from 171 to 322 in four fission yeast Schizosaccharomyces species despite very high similarity in all other gene types, overall architecture and other components of their genomes⁹². tRNA diversity accompanies differential tRNA gene amplification including appearance of new isoacceptors^{92, 95}. Although the mechanisms of tRNA gene amplification are unknown, encounters between DNA replication forks and stable RNAP III transcription complexes, and known links between the molecular machineries involved^96–98 likely underlie this propensity.

Nascent transcripts from individual tRNA genes contain sequence complexity in the 5' leaders, 3' trailers, and for some, the introns⁹⁵. It is reasonable to expect that this diversity would render some pre-tRNAs more prone than others to misfolding. Some eukaryotic tRNAomes include diversity within the mature tRNAs of the same anticodon families, referred to as isodecoders⁹⁵, a source of additional potential for misfolding of the pre-tRNA. It is proposed that under the protective and chaperone activities of La, the transcripts of some tRNA genes might survive to maturity that otherwise might not. Hence, La might support a capacitor function of tRNA gene variation, consistent with a role for chaperones as capacitors for evolution^{81, 82}, as discussed above. It is therefore tempting to speculate that the functional organization of eukaryotic tRNA genes which includes RNAP III-mediated expression that is mechanistically linked to the UUU-3' OH-binding activity of La, contributed to the expansion and variation of eukaryotic tRNAomes by also linking tRNA fate to the chaperone activity of La. Other factors would have contributed, for example, the tRNA gene amplification mechanism(s) itself, as well as certain nuclear tRNA modification enzymes that are known to act redundantly with La to assist structurally-challenged pre-tRNAs avoid decay and promote maturation^{87, 99–102 reviewed in 103}.

The eukaryotic arrangement of tRNA genes as separate transcription units also empowers them as individual genetic units that may be acted on as selectable. Diversity and flexibility in the tRNAome might lead to or coevolve with enhanced potential for biased codon use by the transcriptome and adaptation to translational demands ^{92, 104 see 59}, a perspective consistent with the possibility that the La module was keyed to and rooted in tRNA surveillance.

Conserved and Diverged Features of the La Module RNA Binding Unit

The La module: a novel RNA binding platform with unique features and LARP adaptability

'La module' is the term used to refer to the tandem arrangement of the two RNA binding motifs in which a La motif (LaM) (a unique fold, below) is followed by an RRM (termed RRM1 in La and LARPs)^{3, 105} (Fig. 1). The LaM and RRM1 are comprised of approximately 70 and 80 amino acids respectively, and each is an independent structural 'domain' and will be referred to as such hereafter as appropriate. The La module is an exceptional RNA binding unit discovered from structural studies of hLa^106–108. While several hundred RRMs are found encoded in metazoan genomes^{30, 31}, their pairing as part of a La module is limited to 6–9 unique LARPs per genome. Furthermore, phylogenetic analysis revealed the LaM and RRM1 coevolved in LARPs¹, with only 5 proteins out of the 134 analyzed suggested to contain the LaM alone, underscoring the importance of the La module in its entirety to RNA recognition. Notably, LARP-specific features of the La module have been emerging from phylogenetic and structural studies as discussed below.

In La, the LaM and RRM work synergistically to interact with their RNA targets^{106, 107} and, albeit not demonstrated yet, this is likely true for the LARP superfamily. In hLa protein, the LaM and RRM1 are connected by a largely flexible linker and move about each other independently in the absence of RNA¹⁰⁶, but adopt a fixed ‘V-shaped’ conformational clamp onto the UUU-3’OH target, forming a sequence-specific binding cleft at the interdomain interface that appears to be perfectly tailored for this target (Fig. 3A). While all but one of the specific interactions to the RNA are by side chains of the LaM, recognition of the most sequence-specific determinant of the target, the penultimate uridylate, U-2, is achieved by a cooperative set of protein–RNA and protein–protein interactions involving residues from both domains (Fig. 3B, C)^{106, 108reviewed in109}. This induced fit of the binding pocket around U-2 explains the high synergic nature of this interaction (Fig. 3). This mode of binding was also replicated in LARP7¹¹⁰.

Figure 3. The La module of human La protein in complex with UUU 3’OH RNA.

A) The La module of La (PDB 2VOS) is composed by the LaM (yellow), an interdomain linker (pale green) and the RRM1 (brown). The UUU 3’OH RNA is shown as sticks, color-coded by atom type. For clarity the bases beyond U-3 have been omitted. Cartoon under the structure shows schematic V-shaped model with U-2 in the cleft between the LaM and RRM1 (see text). B) Close-up view of La-RNA interaction showing the three 3’ terminal bases in the same orientation as panel A. Selected side chains are shown as sticks; dashed lines indicate hydrogen bonds. C) Close-up view of the interaction, rotated by 90°.

An intriguing aspect of the LARPs is that despite the multiple conserved features detailed below, their different La modules recognize rather distinct RNA targets. Thus, a challenge is to reconcile differential RNA recognition among LARPs given a high degree of primary structure conservation, especially in their LaMs. In this review we will summarize key findings of how structural studies have started to unveil divergent features that may account for or at least explain in part the different RNA binding properties of LARPs.

The La motif

The LaM was discovered in La as an elaboration of a previously known fold, the 'winged-helix motif,' which in LaM contains three additional α-helices^{44, 107} appended onto the canonical winged-helix scaffold^{75, 77}. These extra elements (α1’, α2 and α4, Fig. 3A) allow the formation of a hydrophobic cavity with RNA binding capabilities, that enables La to adopt a novel mode of RNA recognition that uses entirely different binding surfaces as compared to canonical winged helix domains. Within this hydrophobic pocket, specific and non-specific contacts are made with the UUU-3’-OH ligand, in particular mediated by side chains of 6 key residues that are conserved in LARPs: Q20, Y23, Y24, D33, F35 and F55 (Fig. 3B, C and Fig. 4) (numbering is for hLa, which will also be used to refer to corresponding positions in LARPs throughout this review)^{106, 108}. Thus, the LaM-specific extra helices attached to the winged helix scaffold are integral to RNA recognition and delineate unique function in the context of the La modules of La and LARPs.

Figure 4. Sequence alignment of the LaM and their N-terminal regions in La and LARPs.

For each family, the sequence of the human LaM was aligned with proteins of different species using Clustal Omega in Uniprot portal (http://www.uniprot.org/align/³⁰⁰). The alignments were edited and analyzed with Jalview software³⁰¹. Residues were colored in grey scale according to extent of conservation. Species selected include vertebrates-eutherians (Hs, Homo sapiens, Mm, Mus musculus), vertebrates (Gg, Gallus gallus), invertebrates (Dm, Drosophila melanogaster, Nv, Nematostella vectensis), plants (At, Arabidopsis thaliana) and protists (Dd, Dictyostelium discoideum, Ps, Phytophora sojae). The secondary structure elements for human LaMs appear at the top of the sequence for each family, where α-helices are depicted by cylinders and β-strands by arrows. For La, and LARPs 7, 6, 4, these structured elements have been determined experimentally for the human proteins (^{9, 107, 110}, IC-G & MRC, unpublished). For HsLARP1a, a prediction of secondary structure was performed with Jpred-4 server (http://www.compbio.dundee.ac.uk/jpred/²⁹⁹). The conserved residues of the hydrophobic pocket of the LaM are labeled with asterisks and boxed in orange.

Structural and mutagenic analyses have revealed that the hydrophobic RNA-binding pocket identified in the LaM of La appears functional in LARP7 for its interaction with 7SK RNA UUU-3’-OH¹¹⁰, and surprisingly also for hLARP6 for its internal recognition with collagen 5’UTR stem-loop (SL) RNA⁹. Whereas this was anticipated for LARP7 based on high degree of amino acid homology with La as well as similarities in RNA termini recognition, for hLARP6 this finding is particularly significant given the difference in size and overall stem-loop structure of its ligand, the collagen SL RNA. More intriguingly, the 6 key residues that in La mediate UUU-3’-OH specific recognition appear to retain a central RNA binding role in hLARP6, albeit the 3’OH does not appear to be an element of recognition for LARP6 proteins (see below). Consistent with this is the high conservation of the 6 residues across all LARPs: considering the non-PAM2 (PABP-interacting motif 2) proteins (below), position 20 (hLa numbering) is occupied by Q in 96.6%; positions 23 and 24 by F/Y/W in 97.9% and 99.9%, respectively; position 33 by D in 99.7%; position 35 by F/Y in 96.2%; and position 55 by F in 97.6% of the sequences examined³. This is astounding conservation over the presumed 2 billion year phylogenesis of all of the LARPs. Fig. 4 shows an alignment of the non-PAM2 and PAM2-containing LARPs.

Invariant Asp-33 in the hydrophobic pocket and potential for 3'-OH end binding

A particularly intriguing issue that is potentially relevant to the La modules of all LARPs is the role of invariant D33 which in La proteins resides deep within the LaM binding pocket and is critical for recognition of the 2’-OH and 3’-OH moieties of the terminal ribose, via formation of a bifurcated hydrogen bond with its side chain carboxylate (Fig. 3B, C)¹⁰⁸. There is phylogenetic divergence of 2 to 3 key hydrophobic pocket residues among LARPs that acquired a PAM2 sequence³; of these, D33 was maintained as universally conserved in LARPs4 & 4B. However, for plant LARPs 6B & 6C, which also acquired a PAM2 sequence and diverged in key hydrophobic pocket residues³, this included divergence of D33 but acquisition of D and E respectively at their adjacent 34 positions (Fig. 4). In the LARPs for which the structures and 3'-end recognition have been examined^{9, 44, 106, 108, 110}, i.e. La and LARP7, bearing D at position corresponding to hLa 33, simply altering the RNA terminus from 3’-OH to 3’-PO4, 3'-O-CH3, or 2'-O-CH3, significantly impairs binding to the wild type (WT) proteins.. For La, depending on how the experiments are done, the source of the protein, the specific RNA target, and which 3' or 2' terminal ribose modification is examined, the negative effects range from small to more than 50 fold^{14, 44, 108}. For La-UUU-3’-OH interaction, in agreement with the structural data, mutagenesis experiments have shown that D33 accounts for La's ability to distinguish RNAs with 3’-OH from 3’-O-phosphate and 2’-O-CH3 termini^{14, 44, 108}. The snugness of fit of the 3′-terminal uridylate (U-1) in the LaM binding pocket in close proximity of D33 (Fig. 3A & B) explains the increasingly deleterious effects caused by D33 substitutions with progressively bulkier amino acid side chains and modifications of the 2′-OH or 3′-OH of the RNA ligands, as larger chemical groups can be less well accommodated in the tight binding slot¹⁰⁸. Still, the nonbulky D33A substitution decreases affinity for the 3′-OH ligands, 2–3 fold^{44, 108}, and the D-to-A substitution in T. brucei La protein led to a nearly six-fold decrease in discriminatory affinity for 3'-PO4 vs. 3'-OH⁴⁴. It should be noted that such discriminatory activity would help target La to newly transcribed transcripts, which end with 3’-OH, and disfavor binding to products of RNA cleavage and/or decay enzymes that produce 3’-PO4 and related termini.

Burying the RNA terminus in a deep binding pocket is likely critical to a principal biological activity of genuine La protein, protection of its ligands from 3' exonucleases^73–75. This 3'-end protection activity serves three roles in tRNA biogenesis: i) orders the early phase of pre-tRNA maturation by directing 5' processing to precede 3' processing⁷³, ii) directs 3'-end maturation by the endonuclease RNase Z rather than by the 3' exonuclease, Rex1^{75, 111}, and iii) protects structurally-challenged pre-tRNAs from 3' polyadenylation by the TRAMP complex and subsequent degradation by the exosome-Rrp6, nuclear surveillance system^{74, 85, 101}. Relevant in vivo, a La D33R point mutant is defective in stabilization of 3'-trailer containing pre-tRNALys⁷⁴. By this La module-mediated RNA-binding mode, La acts as a 3' cap on its UUU-3'-OH RNA ligands, protects them from a variety of exonucleases, and can also chaperone them through different processing pathways^{15, 112}. Is a related activity relevant for LARPs?

An alternative role for invariant Asp-33 in the hydrophobic pocket?

As noted, D33 has been universally conserved by LARPs with exception of plant 6b and 6c, which interestingly possess a D and E respectively at the +1 adjacent positions (Fig. 4). In considering the possible significance of this conservation, there have been some curious observations on LARPs that do not entirely fit with 3'-end protection observed in La and worthy of review. Human LARP4 was shown not to significantly distinguish stem loop (SL) hairpin RNAs that differed only in their 3’-OH or -PO4 termini¹². However, more recent experiments using single stranded oligo(A) suggest that hLARP4 is somewhat more discriminatory against 3’-PO4 and 2'-O-CH3 termini than observed for the SL RNA (Gaidamakov & Maraia, unpublished). For hLARP6, while modification of the 3’-OH of collagen 5' SL RNA to 2’–3’ cyclic phosphate had no effect on high affinity binding, interestingly, D33A substitution in the LaM binding pocket was associated with a significant, 3.5-fold reduction in binding affinity⁹. This echoes what was observed with hLa D33A and the D33A equivalent of T. brucei La^{44, 108}, suggesting a comparable contribution of the hLARP6 D33 equivalent to RNA binding affinity as La, despite the expected differences in its contacts to the ribose, the triphosphate and/or base moieties of the respective ligand⁹. We note that recognition by hLARP6 of the SL in the 5' UTR of collagen mRNA is not expected to involve 3'-OH terminal binding per se, and this will be considered in a model below.

Finally, recent data indicate that the LaM of hLa can exhibit context-dependent sensitivity to the nature of ligand RNA 3' termini¹⁰⁵. Binding of hLa to SL IV of the IRES element of the hepatitis C virus (HCV) RNA involves an alternative mode of molecular recognition compared to UUU-3’-OH, requiring cooperative interplay of the La module with the C-terminal RRM2α, which is largely insensitive to distinguishing 3'-OH vs. 3’-O-CH3¹⁰⁵. This was surprising because analogous 2'-O-CH3 alteration of a U(4)-terminated RNA decreased its affinity for hLa by nearly 40-fold¹⁰⁸. Here too, high affinity binding to this SL in its native setting would be connected to the rest of the HCV IRES and downstream coding sequences and therefore is not expected to include recognition of the 3'-OH terminus of the RNA.

The findings that both hLa and hLARP6 when binding their respective HCV and α-collagen SL RNAs respectively, use their LaM binding pocket and invariant D33, but in a way not involving 3'-OH recognition, raises an important question: can an RNA that engages D33, continue on a course out of the hydrophobic pocket? Examination of the structure of hLa bound to its RNA ligand suggests such an exit path might exist, probably through a space between helices α2 and α3 or α2 and α4 of the LaM, although it is noteworthy that the overall conformation of the non-3’-OH bound RNA is likely to differ somewhat from the one that is 3’-OH recognized (Fig. 3A, U-1 is the 3' end of the RNA). For hLa-HCV SL RNA interaction studies, NMR chemical shift perturbation (CSP) analysis may provide some support for this as a larger number of residues in and around the hydrophobic binding pocket experience CSP with the SL RNA compared with the UUU-3'-OH (figure 5 of ref. ¹⁰⁵). An exit path for the RNA could also be speculated for collagen SL binding to hLARP6, whereby the internal bulge of the RNA has been identified as the putative recognition element^{9, 10}. Beyond this initial speculative model, a deeper understanding of the binding of hLa or hLARP6 to structured stem-loops (HCV or collagen SL respectively) requires further structural data of the protein-RNA complexes.

The collective results suggest that while D33 (and equivalent thereof) appears to contribute to RNA binding across the LARPs, its ability to distinguish and/or discriminate 3' termini is dependent on overall architecture and mechanism of the particular molecular interaction. An outstanding issue is whether a La module binding pocket-D33 interaction is used to protect ligand RNAs from 3' exonucleases in the LARPs, two of which have been shown or suspected to have 3'-end binding^{113, 114}. In any case, the observations suggest the La module as an extremely versatile platform capable of more than one mode of RNA recognition.

LaM domain boundaries

Current understanding of the structural basis of differential RNA target specificity by the La modules of LARPs 1, 4 and 6 is limited. Nonetheless, interesting features of the LaM boundaries that emerged from structural work on hLARP6 may provide valuable insights. At the N-terminal region of LARP6 LaM, the structure revealed an α1 helix that is shorter than in La, with the stretch preceding this helix (residues 81–89) forming an integral part of the LaM domain, establishing non-polar interactions with helix α5⁹. Mutations of W85 and K86 here did not alter RNA binding activity of hLARP6⁹, although a deletion mutant of this region failed to bind RNA, suggesting unfolding of the truncated domain¹¹⁵. Interestingly, this N-terminal stretch is highly conserved in eutherian and other vertebrate LARP6 proteins⁹ but not in invertebrates and plants. Notably, LARP6 orthologs from different species have different RNA binding properties³ but whether the N-terminal region contributes to this remains to be established. Interestingly, inspection of sequence preceding the LaM in other LARP families highlights similar patterns of species-dependent conservation for this region (Fig. 4).

Another divergent feature of the LaM that was not anticipated from sequence alignment, but became evident from recent structure determination, applies to wing 2 at its C-terminal boundary (Fig. 5A–C). In La, wing 2 begins with Arginines 90 and 91 and is characterised by a nearby PLP motif (96–98) (Fig. 5A), which represents the C-terminal end of the LaM, with P96 and L97 establishing hydrophobic interactions with α1’¹⁰⁷ (Fig. 5B, 5C). Residues beyond P98 were shown by NMR backbone dynamics analysis to be flexible, thereby demarcating the end of the LaM and the beginning of the interdomain linker¹⁰⁶. In the hLARP6 LaM, despite the two Arginines and a PVP signature (residues 172–174) (Fig. 5A), P172 and V173 were surprisingly not involved in interactions with α1’: instead, a different configuration of wing 2 positions these residues somewhat away from the rest of the molecule, leaving the downstream residues, L175 and F176, to stabilise it via contacts with α1’ (Fig. 5C). The divergence between La and LARP6 here is two-fold: (i) different structure at the C-terminal of LaM and (ii) shorter interdomain linker for hLARP6, and both parameters are very likely to play a dominant role in determining the relative domain orientation of the tethered LaM and RRM1 within the La module⁹.

Sequence alignment of LARPs (Fig. 5A) indicates that as compared to hLa, hLARP7 contains a shorter wing 2, and the signature PLP is replaced by PLG (residues 114–116). Nonetheless, L115 makes stabilizing contacts with α1’ and the overall configuration of wing 2 is not very dissimilar to that of La¹¹⁰ (Fig. 5B). The wing 2 sequence is somewhat longer in LARP1 members and is interestingly absent from LARP4 members (Fig. 5A). Overall, there are significant differences in the wing 2 sequences across LARPs that could potentially have considerable repercussions in the mode of RNA binding, influencing the relative positioning of the LaM and RRM domains of the respective La modules.

The RRM1

Although the structure of RRM1 in hLa closely resembles the canonical RRM fold (Fig. 6A), it was surprising to find that it uses a noncanonical RNA-binding surface to bind short oligoU sequences^{106, 108, 116, 117}. Phylogenetic classification anticipated LARP family-specific characteristics in the predicted RRMs¹, and the newly solved structures of human LARP7 and LARP6 uncovered significant RRM1 variability: only 3 β strands instead of the canonical 4 were observed in LARP7 RRM1¹¹⁰ (Fig. 6B), whereas LARP6 RRM1 contains additional helices⁹ (Fig. 6C). In addition, work on hLARP7 and hLARP6 suggested that the canonical RNA-binding surface of their RRMs1 are also not positive determinants of sequence-specific RNA interaction, similar to La protein^{9, 110}.

Figure 6. Comparison of RRM1 and RRM2 domains in LARPs.

The RRM1 domains of LARPs are structurally diverse: A) hLa RRM1 (PDB 1S79); B) hLARP7 RRM1 (PDB 4WKR); C) hLARP6 RRM1 (PDB 2MTG). hLARP7 RRM1 lacks strand β4; hLARP6 RRM1 contains additional helices α0’ and α1’. The RRMs2 of D) hLa (PDB 1OWX), E) hLARP7 (PDB 5KNW) and F) p65 (PDB 4EYT) all contain a long C-terminal helix (α3) that obscures the β-sheet platform. G) In p65 the unstructured C-terminal of the helix (α3×) refolds upon RNA binding. H) Close-up view of the interaction between p65 and the S4 RNA. Selected residues are highlighted in stick representation. I) Sequence alignment of the α3 region for hLa, hLARP7 and p65 performed with Clustal Omega in Uniprot portal (http://www.uniprot.org/align/³⁰⁰) and edited and analyzed with Jalview³⁰¹. Residues colored in dark red indicate conservation. Boxed residues denote amino acids that interact with RNA in p65.

The distinctive LARP family-associated traits of RRM1 suggest that this domain plays a key role in their RNA substrate selection specificity. This was demonstrated by loss of binding specificity or affinity when RRMs1 of hLARP7 and hLARP6 respectively were replaced by RRM1 of hLa^{9, 118}. Although the features in hLARP6 RRM1 that mediate specific interaction with the SL in collagen mRNA are yet to be defined, preliminary data indicate more extensive contacts as compared to hLa and hLARP7, and this may also be the case for hLARP4 (Cruz-Gallardo & Conte, unpublished). Detailed structural information on the La modules of additional LARPs in complex with their targets will be necessary to further address this issue.

Interdomain linker and domain-domain orientation in the La module

Robust synergism between LaM and RRM1 for RNA binding that was seen in La is recapitulated in both LARP7 and LARP6 La modules^{9, 110}. In hLa, the linker is flexible in the apo protein but adopts helical conformation in the RNA-bound form, assisting the correct positioning of both domains in an orientation optimized for RNA recognition¹⁰⁶.

Demonstration of the important topological role of the linker in RNA recognition was provided by the observation that replacing the short linker of hLARP6 with the longer one from hLa, which would perturb interdomain distance and degree of conformational sampling, decreased binding affinity for collagen SL RNA by ~10 fold. Importantly, the cumulative analyses from the LARP6 study were not readily compatible with a V-shaped architecture for LaM and RRM1 as occurs in La and LARP7 bound to UUU-3'-OH⁹. Instead, they envisioned a more elongated tandem domain orientation in the RNA bound form, in which the interdomain linker played an integral role in domain architecture and RNA recognition⁹ (Fig. 7).

Figure 7. Proposed arrangements of the La module for RNA recognition by LARPs.

A) Domain arrangement of hLa and hLARP7 in complex with RNA (RNA removed for clarity). The LaM and RRM1 adopt a V-shaped conformation to create the binding pocket to accommodate the RNA ligand. B) Cartoon representation of possible domain-domain orientation of the La modules of hLARP6 and hLARP4. Current knowledge on wing 2 conformation and interdomain linker for these proteins suggests that LaM and RRM1 will adopt a more elongated arrangement to interact with RNA (see text).

Perhaps, of more general significance, the biophysical studies of LARP6 provided initial evidence to suggest that the two domains of the La module may be capable of synergistic RNA recognition via different topological arrangements in different LARPs. An elongated rather than a deep V-shaped (Fig. 7) configuration may also be expected for the hLARP4 La module from preliminary structural data (Cruz-Gallardo & Conte, unpublished), and this would be consistent with a binding requirement for a relatively long tract, 15 nt, of single stranded RNA¹². Whether the La module tandem domain arrangement in hLARP6 and hLARP4 differs from hLa and hLARP7 awaits confirmation from structural data of their complexes with target RNAs.

In summary, the length and amino acid composition of the LaM-RRM1 interdomain linker should be expected to be key in determining the functional structure of La modules across the LARP superfamily, with consequent effects on RNA binding. Interestingly, sequence and length of the linker indeed vary among LARP families, similar to that noted for wing 2 (Fig. 5A).

Use of canonical and noncanonical RNA-binding surfaces

Formation of the unique binding pocket for UUU-3'OH recognition in La protein by the LaM and RRM1 does not involve the RNA binding surfaces expected for canonical winged helix and RRM domains respectively, and therefore leaves open the possibility for additional, perhaps simultaneous, modes of RNA binding^{108, 116, 117}. In support of this possibility it should be underscored that La RRM1 bears a number of hallmark features of the canonical β-sheet RNA-binding surface, including RNP-1 and RNP-2 motifs on β-strands 3 and 1, respectively, whose key aromatic residues have been conserved e.g., ^{see Fig 3 in 74}. Functional involvement of the canonical RRM surface in pre-tRNA binding was demonstrated for hLa^{74, 80}. Moreover, the UUU-3'-OH and the additional RRM1 canonical surface binding mode could be functionally linked to the two different activities of La, 3' end protection and pre-tRNA chaperone activity, respectively⁷⁴. Mutation of hydrophobic pocket residues in the LaM impaired UUU-3'-OH binding while substitution of loop-3 residues weakened binding to other parts of the pre-tRNA⁸⁰. Mutagenesis coupled with in vivo functional assays indicate that the conserved aromatic residues on RNPs 1 and 2 and key basic residues in loop 3, that connect β2 and β3 of RRM1, are required for pre-tRNA chaperone activity⁸⁰. Analogously, the β-sheet of LARP7 RRM1 has been suggested to contact 7SK RNA beyond the terminal U triplet¹¹⁰.

It is noteworthy that a relatively short stretch of RNA, (U)UUU-3'OH is specifically recognized by the La module^{106, 108} of La and LARP7 to comprise most of its high affinity binding despite the potential extra RNA binding surface. In contrast, the La modules of hLARPs 4 and 6 do not recognize equally short RNAs: e.g., hLARP4 requires at least 15 nucleotides of its highest known affinity ligand, oligo(A)¹².

In summary, evidence exists for canonical and noncanonical type RNA binding by the RRM1 in La protein, the latter of which provides the clearest and best established structural models by which it recognizes its most prevalent and biologically relevant ligands, the UUU-3'OH termini of nascent RNAP III transcripts. However, as reviewed above, biochemical and functional biological data also indicate use of the canonical RRM surface of hLa in tRNA biogenesis^{74, 80}.

While our collective understanding of RNA recognition by La modules has advanced enormously due to detailed structural studies coupled to mutagenesis and affinity measurements, many interesting and important questions remain unanswered. Genuine La protein itself bears multiple RNA-binding surfaces and can exhibit more than one mode of RNA recognition. Therefore, models based on UUU-3'OH binding by La protein should not necessarily be expected to explain RNA binding by a LARP. Evidence of noncanonical arrangement and combinatorial potential of the La module should also caution predictions based on sequence and structure alone. On the other hand, the possibility that the La module can adopt alternate architectures (Fig. 7) enlighten another perspective. The extent to which a La module may be able to switch between topological architectures upon interactions with different RNAs is an intriguing possibility.

Proteins with a LaM and no apparent RRM

It is also noteworthy that a few proteins have been identified that contain a LaM without an adjacent RRM, including S. pombe Slr1p, A. niger AnLARP, P. sojae PsLARP, as well as the characterized S. cerevisiae Sro9p and Slf1p¹. Sro9p and Slf1p appear to function in mRNA translation¹¹⁹ and were shown to selectively bind homopolymer RNA¹¹⁹. Slf1 was known to be involved in copper metabolism in yeast before it was known to harbor a LaM¹²⁰. RNA co-IP experiments revealed association of Slf1 with mRNAs related to copper resistance and oxidative stress, apparently dependent on the LaM, as the triple mutant Y24A-F35A-F55A (hLa protein numbering) resulted in decreased association¹²¹. An oxidative link to mitochondria through Slf1 was uncovered^{122, 123}. Slf1 and Sro1 appear functionally related to LARPs 1 and 4 families and similarly to these, associate with translating ribosomes and PABP and interact with overlapping sets of abundant mRNAs including those that encode ribosomal proteins¹²³. Indeed, Slf1p was found as one of a few nonribosomal proteins in the yeast 77S complex that contains closed-loop-specific translation components¹²⁴, reminiscent of other cytoplasmic LARPs (below). Although much remains to be uncovered about the mode of RNA binding of these unusual LARPs, it may be possible to hypothesize the existence of an additional non-canonical RNA recognition mechanism by LARPs that involves the LaM but does not require a RRM1 at all. Alternatively, a cryptic or nonrecognizable RNA-binding motif may reside C-terminal to these LaMs. Further analysis of their RNA binding requirements will be required to test this.

LARP Regions Beyond the La Module Involved in RNA Recognition

Although the La module is a unique RNA binding unit and a main locus of RNA recognition, it is allied to other motifs and domains that are distinctive to each LARP family, and these can contribute to discrete RNA binding by these proteins. Here, we will use three recent examples of RNA binding by downstream motifs of different LARPs, the xRRM, RRM2α and DM15.

The RRM2α and xRRM

La and LARP7 families contain an additional RRM in their C-terminal region, of a subtype with an α3 helix that lies across the β-sheet²⁰, referred to here as RRM2α (Fig. 1, Fig. 6D, E & F). In p65, the Tetrahymena homolog of LARP7, the RRM2α was dubbed xRRM based on specific novel features not shared by La RRM2α. The p65 xRRM interacts specifically with the terminal stem-loop 4 (S4) of telomerase (TER) RNA independently of the La module using an extension of helix α3 (termed α3×) that transitions from an unstructured state in the apo protein to helical conformation upon binding to S4 RNA²³ (Fig. 6G, H & I). Structural studies revealed that α3× binds S4 across the major groove adjacent to a GA bulge with aromatic residues F521, F524 and F525 engaging in hydrophobic interactions (Fig. 6G, 6H). Residues Y407 and D409 on β2 and R465 on β3 provide further hydrophobic contacts and hydrogen bonds between the edge of the β sheet and the RNA²³ (Fig. 6G, H). Additional interactions, largely with the RNA backbone, involve basic residues on the α3× helix (K517, K518, R522, K528, K529).

The unusual features and novel mechanism of RNA recognition that led to the definition of xRRM, are based on: (i) absence of RNP2 and RNP1 consensus sequences in β1 and β3 respectively; (ii) conservation of arginine (R465 in p65) within the non-consensus RNP1 which recognized both nucleotides in the GA bulge; (iii) presence of a new Aromatic-X-D/Q/E/N motif on β2, involved in nucleotide recognition, that was termed RNP3; (iv) an α3 helix that lies across the β-sheet surface with a C-terminal tail required for RNA binding that is disordered in the free protein but forms the α3× extension upon RNA binding; and (v) high affinity binding to RNA that depends on contacts with α3× and is independent of the La module²³ (Fig. 6G, H &I).

Recent studies on hLARP7 RRM2α reveal similarities and differences with p65 xRRM. First, contrary to p65, LARP7 RRM2α does not bind a two-nucleotide bulge, but interacts with both unpaired and base-paired nucleotides in the stem and apical loop of hairpin 4 (HP4) of 7SK RNA^{110, 125}. LARP7 RRM2α shares with p65 xRRM the hallmark RNP3 sequence in β2 (YVD) and the conserved Arg on β3, which are involved in the 7SK HP4 RNA binding according to NMR chemical shift perturbations analysis¹²⁵. Despite the presence of helix α3 in LARP7 RRM2α that appears to participate in RNA binding, the extent of involvement of a C-terminal tail that would form an α3× helical extension when bound to RNA is still somewhat unclear. The very weak chemical shift perturbations (CSP) detected upon RNA binding beyond residue 546¹²⁵ are inconsistent with a newly folded α-helix which would typically be accompanied by extremely large CSP^{106, 126}. Furthermore, C-terminal truncation of 5 residues (551 to 546) decreases LARP7 binding affinity for 7SK HP4 by ~10 fold; however the equivalent truncation in p65 xRRM (to residue 519) completely abolishes binding to TER S4²³, indicating dissimilarities in RNA recognition mode between LARP7 and p65 RRM2. Supporting this, α3 sequence alignment shows that hLARP7 lacks the 3 Phe aromatic residues and some of the basic residues that bind to TER S4 in p65 (Fig. 6I).

The structure of the RRM2α of hLa was the first that showed presence of a long C-terminal helix α3¹⁹ adopting a configuration that was largely recapitulated in the RRMs2 of LARP7 and p65. RRM2α of La behaves quite differently from LARP7 and p65, in that it has very little, if any, RNA binding capability in isolation, and appears to contribute to RNA recognition only when working in synergism with the La module to recognize internal SL structures, such as in the HCV IRES and pre-miRNAs^{127, 128}. RRM2α also appears important for binding to HBV SL and other mRNAs^{129, 130}.

Closer inspection of sequence alignment shows that in La the RNP-3 sequence is somewhat conserved on β2 (WID)¹²⁵, whereas the key Arg on β3 and the basic and aromatic residues on α3 and α3× are not (Fig. 6I), perhaps providing an explanation for the difference in RNA binding capability of the RRM2α in La as compared to the LARP7 family.

While the isolated RRM2α of hLa (225–334) exhibits very little if any RNA binding on its own, addition of sequences about 20 residues downstream of α3 which includes a tract of highly basic residues (a.k.a., short basic motif, SBM, in Fig. 1) increases RNA binding^{19, 49, 131} and probably contributes to yet other modes of RNA recognition by La, including to the 5’ triphosphate of nascent RNAP III transcripts^{49, 132}. A new study employing time-resolved electrospray ionization hydrogen–deuterium exchange (TRESI-HDX) identified potential changes that accompany RNA binding in the entire hLa C-terminal domain, encompassing RRM2α and the SBM¹³¹. Interestingly, a region just following helix α3 (residues 321–326) was proposed to become less structured in the presence of ssRNA and more structured when mixed with SL IV of HCV RNA¹³¹. However, exact molecular details await further studies, including clarification as to whether any cooperation exists between the basic region and the RRM2α for RNA recognition.

The DM15 domain

LARPs 1A and 1B possess a highly conserved C-terminal region, namely DM15, composed of one to four conserved DM15 repeats/boxes¹. The DM15 domain of human LARP1A was found to adopt an atypical HEAT-like fold containing three helix-turn-helix repeats, that map to the three conserved DM15 boxes (A, B and C)¹³³. Importantly, this domain is capable of mediating RNA binding, independent of the La module¹³³. Very recently the DM15 domain was unveiled as a new cap binding protein, recognizing specifically the 5′ cap motif m⁷GpppC – with a C as the first base – which is characteristic of 5'TOP mRNAs¹³⁴ (Fig. 8). In agreement with previous observations, in a classical case of convergent evolution the specific mode of recognition of the m⁷G cap by DM15 follows a common theme shared with other cap binding proteins, whilst the overall domain fold diverges¹⁰⁹. Specifically in DM15, the charged methylated m⁷G base is stacked between two tyrosines, Y883 and Y992, in a classical ‘cation-π-sandwich’, whilst its nitrogen atoms N1 and N2 establish hydrogen bonds with an acidic residue (E886) in the binding pocket, all of which are common features of cap recognition. Additional H-bonds are formed between N1/N2 and Ser882 in the cap-binding crevice (Fig. 8). Interestingly, the first C base engages in a specific network of H-bonds with arginines 847 and 879. The functional repercussions of this specific recognition are considerable, as most cellular mRNAs have a Guanine in the +1 position, contrary to 5'TOP mRNAs. The net preference of DM15 for a C in the +1 position was unambiguously demonstrated in EMSA experiments comparing different capped oligo-RNAs¹³⁴. Strikingly, the LARP1 DM15 domain both outcompetes cap-binding protein eIF4E for binding to m⁷GpppC-TOP RNA and can displace the cap-C TOP RNA from eIF4E¹³⁴.

Figure 8. Structure of the DM15 domain of hLARP1 in complex with m⁷GTP and 5’TOP RNA.

The DM15 domain is composed by three helix-turn-helix repeats denoted by DM15 boxes A, B and C. The figure shows a superimposition of 2 crystallographic structures: (i) DM15 bound to the RNA sequence 5’-CUUUUCCG-3’(PDB 5V7C) and (ii) the complex DM15-m⁷GTP (PDB 5V4R). For clarity the bases beyond U4 in the 5'TOP mRNA have been omitted. Selected side chains involved in protein-RNA contacts are shown as sticks and dashed lines indicate hydrogen bonds.

Beyond the cap and the first base, only the first 4 nucleotides of the bound RPS6 5'TOP mRNA sequence contact the positively charged concave surface of the DM15 domain, with interactions between the phosphate backbone of the RNA and side chains of R879, R840 and H800¹³⁴ (Fig. 8). No further contacts were seen with ribose or base moieties beyond C+1, suggesting that the 5'TOP mRNA sequence would not be a determinant for substrate discrimination by DM15. Thus the basis of the preference reported for the DM15 domain for some 5'TOP sequences over others appears to remain unexplained by the recent structure. Nonetheless, direct interaction of DM15 with the 5′TOP motif¹³⁴ is particularly interesting given that a putative mTORC1-recognition sequence is located in a flexible loop C-terminal to DM15 repeats, thereby ascribing a role for LARP1 in directly linking TOR signaling to ribosome biogenesis (see below).

The Nuclear La Protein

La protein has two recognized functions, nuclear, to assist nascent RNAP III transcripts in their processing and maturation, and cytoplasmic, in mRNA translation. Furthermore, studies from multiple laboratories have shown that a fraction of La concentrates in nucleoli in various cell types although its function remains unknown. In addition to binding cellular transcripts, the intimate association of La with specific transcripts synthesized by viral RNA polymerases, first elucidated by work on RNA viruses, continues to deepen our understanding of its impact on an expanding field of virology. Thus, it is appreciated that La is a multifunctional protein. Below we will review recent findings on nuclear functions of La and in a later section its cytoplasmic functions.

Nuclear La function in nascent RNA sorting

As noted above, a recent advance in La function was in the channeling of nascent RNAP III transcripts⁵³. Related to this, it had been reported that the activity of Ago2, a Piwi-homologous, active component of the siRNA-mediated silencing complex (RISC) that uses siRNA to guide cleavage of cytoplasmic mRNA, could be enhanced by La¹³⁵. Evidence was later presented that La was additionally involved in another part of the miRNA-mediated silencing pathway. It was shown to associate with, stabilize and regulate nuclear pre-miRNAs by SL binding in an UUU-independent manner requiring its La module and RRM2α¹²⁸, similar to the domain requirement used for binding to the hepatitis C virus (HCV) mRNA¹²⁷. The recent results indicate that La prevents nascent pre-tRNAs with propensity for misfolding from routing to the pre-miRNA pathway⁵³. In the absence of La some pre-tRNAs undergo alternate folding to structures recognized by the nuclear export factor Exportin-5 and the processing factor Dicer and get loaded into Ago-containing RISC; however when La is present its early access to nascent pre-tRNAs prevents their mischanneling⁵³.

La binding to spliced pre-tRNA intermediates in animal cell nuclei

As noted in preceding sections, the La proteins in all species examined representing ancient eukaryotes that predate the evolutionary emergence of yeast, as well as other species examined including plants, contain a RRM2α, whereas the yeast lineage appears to have lost its RRM2α¹. A feature here that is of potential relevance to the nuclear function of La is that the miRNA pathway was lost from some yeast¹³⁶. However, a more striking difference is that pre-tRNA splicing occurs in the nuclei of animal cells but in the cytoplasm of yeast¹³⁷. This is relevant because La is found associated with pre-tRNA splicing intermediates in animal cells whereas it is physically separated from the splicing pathway in yeast. Indeed, mouse brain cells genetically deleted of La protein become deficient in a 3' trailer-containing spliced intermediate pre-tRNA that is normally bound to La in wild type brain cells¹¹².

LARP7 Family Members

These are the LARPs most similar to La, not only by homology and phylogeny¹, but their La modules exhibit the same UUU-3’OH sequence-specific RNA recognition mode and they also share similar RRM2α architecture (Figs. 1, 6). However their distinctiveness is that while La binds UUU-3’OH terminal motifs on all RNAP III transcripts, LARP7 members recognize this motif on a specific subset of RNAP III transcripts, the identity and function of which can differ in different species. Another functional distinction is that while La binding is almost always transient for the spectrum of nascent RNAP III transcripts, LARP7 members remain associated with their RNA as an integral part of a stable RNP, e.g., metazoan 7SK snRNP or ciliate telomerase RNP.

LARPs7 provide a clear important example of how sequence homologous members of a single LARP family can adopt different ligands and functions in different species. Vertebrate and other metazoan LARP7 members stably bind to the ~331 nt 7SK RNA as a component of an abundant snRNP. Physiologic roles of the 7SK snRNP include balancing the growth and other regulatory functions of P-TEFb (positive transcription elongation factor-b) activity which releases RNAP II from its promoter proximal paused state to a productive elongation state^138–140. LARP7 association with 7SK RNA is highly stable, making it an integral component of the snRNP^141–143. The LARP7 members of the two ciliate species that have been studied, Tetrahymena p65 and Euplotes p43 (Fig. 1), are stably associated with telomerase RNA, which in ciliates is a ~160 nt product of RNAP III (whereas RNAP II synthesizes telomerase RNA in metazoa). Because polyploid ciliate macronuclei contain a large number (200–300) of linear chromosomes, they need very high levels of telomerase RNA, a function well suited for RNAP III^{58, 59}. p65 and p43 are stable components of their respective multisubunit RNP telomerase enzyme complex and integral to function^{144, 145}. Both of these types of RNPs are responsible for critical functions that are central to proper growth and development in their respective biological systems, and much work has gone into their structure-function analyses.

Data suggest that hLARP7 and p65 exhibit RNA chaperone-like activity. This involves RNA folding and/or other facets of dynamic RNP organization upon binding to 7SK snRNA or telomerase snRNAs^{23, 141, 146}.

Transfer of pre-7SK RNA from La to LARP7 preceding assembly of 7SK snRNP

It is known that as a nascent transcript, pre-7SK RNA is associated with La via its UUU-3’OH before transfer to LARP7 for assembly of the 7SK snRNP ^{141 reviewed in 2}, but an open question remains on how transfer from La to LARP7 occurs². A first observation is that LARP7 appears tailored for 7SK because in addition to La module recognition of UUU-3’OH, its RRM2α makes tight interactions with the apical loop of HP4. Both docking sites, UUU-3’OH and the HP4 apical loop are required for LARP7 interaction: the U tail alone is unable to support in vivo binding¹⁴¹ and LARP7 mutants lacking the C-terminal region that contains RRM2α are unable to bind 7SK in vivo¹¹⁸. By contrast, the La RRM2α is unable to interact with HP4 and appears to have different ability for RNA binding¹²⁷ (and see above). It is not clear however that increased avidity of LARP7 for 7SK RNA is sufficient to displace La.

A nucleotide triphosphate is a universal motif incorporated at the 5' end (5'-pppG/A) of all newly synthesized RNAP III transcripts, which has been shown to be a target of the short basic region (SBM) in the C-terminal region of hLa^{49, 132}. The 7SK RNA sequence contains a 5' proximal motif that promotes the modification of its 5'-pppG to a monomethyl γ-phosphate cap (5'-me-ppG) by the methyltransferase, MeCPE¹⁴⁷, which decreases the affinity for La protein¹³². Moreover, MeCPE remains stably associated with 7SK RNA in the complex, prompting the suggestion that it may participate in the transfer process. Intriguingly, immunoprecipitation of La from HeLa cells detected 7SK RNA and MePCE but not other components of the regulatory 7SK snRNP (LARP7, HEXIM, CycT1 and hnRNP A1)¹⁴¹, suggesting that MeCPE may associate with La-bound 7SK and methylate the 5'-pppG-RNA, which in turn reduces the affinity for La^{2, 132}.

LARP7-RNA interactions are necessary for LARP7 functions

In the first characterization of the biochemical and biological activity of hLARP7 it was named PIP7S¹¹⁸. LARP7 specifically recognizes the 3’ hairpin of 7SK (termed HP4), in that the UUU-3'OH element is recognized by the La module and the apical loop is specifically contacted by the RRM2α in the C-terminal part of the protein¹¹⁰ (Fig. 1). Both elements have been shown in vivo to be essential docking sites of LARP7¹⁴¹. The atomic details of these interactions have been discussed previously. Intriguingly, recent EMSA experiments revealed that LARP7 can also bind to HP1, the conserved N-terminal hairpin of 7SK that is also the locus for HEXIM interactions¹⁴⁸. It was proposed that the long, mostly unfolded linker between RRM1 and RRM2α (encompassing residues 210–450 in hLARP7), which contains several stretches of basic residues, may mediate this binding¹⁴⁸. These findings also suggest intriguing functional interplay between HEXIM and LARP7 that merits further investigations.

Interactions between LARP7 and 7SK RNA is prerequisite to recruitment of P-TEFb to the 7SK RNP¹⁴¹, with inhibition of P-TEFb kinase activity controlled by the interaction of CycT1 with the C-terminal domain of HEXIM1/2^{138, 149}. The molecular association of HEXIM1/2 and P-TEFb with 7SK RNA/RNP is a dynamic process involving conformational rearrangements and intricate temporal and spatial multi-partner interplay tht is not yet fully understood^{141, 148, 150}. Similar dynamic assembly of telomerase RNP has been attributed to LARP7 p43¹⁴⁶.

A recent biochemical study that reconstituted the 7SK RNP from multiple purified components, showed that it was functional for release of active P-TEFb and that its MePCE-mediated 5'-me-ppG capping activity was inhibited by LARP7¹⁵¹. This further showed that the region C-terminal to LARP7 xRRM is required for interaction with MeCPE in the context of 7SK RNA, for inhibiting its capping activity. This also provided evidence that xRRM in full length LARP7 recognizes the 7SK RNA bulge at nt 320–321 as important for the inhibition¹⁵¹, and that this recognition was proposed to occur in a manner similar to how p65 xRRM interacts with telomerase RNA^{22, 23}. The new data also supported two conformational states of the 7SK RNA¹⁵¹, including a 'closed state' consistent with juxtaposition of the 3’ and 5’ ends of 7SK RNA¹¹⁰.

Involvement of LARP7 in other activities

In agreement with a proposed role of LARP7 in p-TEFb recruitment to 7SK RNP, LARP7 mutations or its down regulation are associated with gastric, breast and cervical tumorigenesis^152–154, presumably by increasing the growth-promoting activity of P-TEFb^{2, 118, 155}. However, data have shown that in embryonic stem cells (ESCs) knockdown of LARP7 is paradoxically not associated with increased cell growth and proliferation as expected via activation of P-TEFb, but instead leads to growth failure¹⁵⁶. This apparent embryonic-specific effect may help explain cases of loss-of-function mutations of LARP7 associated with primordial dwarfism (PD), a condition of severe growth restriction and associated symptoms^157–160. The exact mechanism of the unexpected activity of LARP7 in early development is yet unclear although it was proposed to be mediated by interaction with Lin28 mRNA and the poly(A) polymerase Star-PAP in a pTEFb-independent manner¹⁵⁶.

Another study proposed that some symptoms of Alazami syndrome (specific PD with LARP7 mutations) may result from translational and/or nucleolar stress in neural cells¹⁶¹. Although LARP7 is generally nucleoplasmic, it was convincingly found among proteins specifically enriched in nucleoli of rat neurons¹⁶¹. Interestingly, in hippocampal neurons, LARP7 knockdown reduced perikaryal ribosome content and protein synthesis¹⁶¹. How LARP7 executes this function had been unclear but is explained by recent findings that show that the 7SK snRNP plays a positive role in promoting the transcription of small nuclear (sn) RNAs and the small nucleolar (sno)RNAs, U3, U8 and U13 which are involved in processing the precursors of large ribosomal RNA during ribosome biogenesis¹⁶². This function of 7SK snRNP is mediated by interacting with a RNAP II elongation complex that is dedicated to the transcription of small nuclear RNAs, known as the little elongation complex (LEC), and requires LARP7¹⁶².

Recent Advances on the Cytoplasmic LARPs

Ample evidence indicates that members of LARPs 6, 4 and 1 families are mostly cytoplasmic (although nuclear-cytoplasmic shuttling suggests nuclear functions for some) and share a common function in mRNA binding, translation and/or stability. Therefore, we will first provide a brief overview of relevant regulatory factors and other issues related to mRNA translation and metabolism, after which we review advances on cytoplasmic La followed by LARPs 6, 1 and 4. Several excellent reviews on mechanisms of translation initiation are available^{163, 164also see 165}.

Poly(A)-binding protein (PABP): a central factor in mRNA metabolism

For the purposes of this review we will highlight a few relevant points that are schematized in Fig. 9A. mRNAs contain poly(A) 3' tails of variable length that are bound by one or more molecules of PABP, a multifunctional protein that interacts with a variety of other proteins and heteromeric complexes that carry various enzymatic activities, and integrates translation initiation, termination and ribosome reinitiation or cycling, as well as mRNA decay/stability functions. Eukaryotic mRNAs contain a 7-methyl-G(5')ppp(5')N "cap" on their 5' ends, that is recognized by the cytoplasmic cap-binding protein known as eukaryotic initiation factor-4E, (eIF4E) which associates with eIF4G, a large protein that serves as a central hub for several other translation factors¹⁶⁵. PABP contains four RRMs and a C-terminal region that includes a MLLE domain. The eIF4G interacts with the backside of RRM2 of PABP when bound to poly(A)¹⁶⁶. Interaction of PABP with eIF4G somehow increases the functional affinity of eIF4E for the 7mG-cap^{167, 168 see 169}. Bimodal interaction of PABP with poly(A) at the 3' end of the mRNA and with eIF4G at the 5' end accounts for the 'closed loop' model of translation^170–173. Accordingly, mRNAs with 5' capped ends and 3' polyA) tails are efficiently engaged in translation (Fig. 9A). In addition, by binding to the RNA termini, proteins can protect against exonucleases, a major source of mRNA decay¹⁷⁴. In this model, which can be conditional on cell type¹⁶⁴, PABP activities promote initiation, termination and ribosome recycling, as well as stabilization of the mRNA^{38, 175} (Fig. 9).

Figure 9. Working models of translation and LARP-PABP factors.

A) Schematic showing factors involved in an example network of protein-protein and RNA-protein interactions that serve to bridge the 5' and 3' end regions of the mRNA according to a closed loop model of translation. A translation initiation (start) codon AUG, and a termination (stop) codon, UAA are to orient polarity. Individual factors are referred to in the text. Multiple copies of PABP can bind to the poly(A) tail via its four RRM domains (depicted in highly schematized form), some of which also serve as separate docking sites for other factors including eIF4G and Paip1^{166, 178}. The MLLE domain of PABP, depicted as red circles, are used to interact with different PAM2-consensus sequence-containing proteins involved in translation initiation (Paip1), termination (eRF3) or recruitment of deadenylase-containing complexes (CAF1-CCR4 and PAN3-PAN2)^{178, 181}. The eIF4E cap-binding positive initiation factor and its association with eIF4G are depicted ^{for more comprehensive reviews see 168, 169, also see 170}. An inset reveals the eukaryotic mRNA two different cap structures; 5’-TOP mRNAs contain a cap-C (m7GpppC), and other mRNAs contain a cap-G/A (m7Gppp-G/A). Under repressive conditions such as nutritional or other stress, 4E-BPs (eIF4E-binding proteins), sequester the cap binding protein eIF4E. B) A potential working model for LARP1 involvement in mRNA metabolism in relation to translation initiation and stabilization. It can bind via the DM15 domain to the 7mGpppC-TOP motif of TOP mRNAs and to PABP and to poly(A) (or other regions) of the mRNA via its La module (see text), although whether it would do so simultaneously as depicted here is unknown. LARP1 may either stimulate or inhibit translation and stabilize poly(A) under some conditions (see text). C) Proposed working model for LARP4/4B involvement in translation and mRNA stability. LARP4/4B interact with RACK1 and with PABP, although whether either of them would do so simultaneously is unknown, three of the combinatorial possibilities are depicted. LARP4B is proposed to bind to mRNA 3' UTR sequences (¹³ see text). LARP4 has been shown to bind PABP via the PAM2 motif, and through a second PABP-interaction motif (PBM). It is unknown to which part of PABP the PBM binds. LARP4 can also bind poly(A) RNA. Through competition with the PAM2 motifs of the deadenylases for the MLLE domain of PABP, LARP4/4B may protect mRNA 3'-ends from deadenylation, leading to poly(A) length modulation¹¹⁴. 4E-BPs: 4E binding proteins, repressors of translation; Paip1: poly(A) binding protein interacting protein-1, a stimulator of translation; eIF4E: eukaryotic initiation factor 4E, a.k.a., cytoplasmic cap binding protein; eIF4G; eukaryotic initiation factor 4G; eIF3: eukaryotic initiation factor 3; eIF4A: eukaryotic initiation factor 4A; RACK1: receptor for activated kinase C, a 40S ribosome subunit; eRF1 and eRF3 are eukaryotic translation termination/release factors. Tob2: transducer of ERBB2; CAF1-CCR4: chromatin assembly factor 1 and CCR4-NOT transcription complex subunit 6, two proteins with poly(A) deadenylase activity; PAN2: PAB-dependent poly(A)-specific ribonuclease subunit PAN2, catalytic subunit of the poly(A)-nuclease (PAN) deadenylation complex; PAN3: PAB-dependent poly(A)-specific ribonuclease subunit PAN3, regulatory subunit of the poly(A)-nuclease (PAN) deadenylation complex.

Members of the cytoplasmic LARPs 1, 4 and 6, have been shown to interact with PABP, by use of a short peptide sequence that conforms to various degrees to a consensus motif termed PAM2^{3, 6, 12}, and for LARPs 1, 4 and 4B evidence of a second interacting region also exists^{4, 11, 12, 37, 176}. Poly(A) is bound by PABP via its 4 tandem RRMs with collective binding affinity of ~20 nM¹⁷⁷. Poly(A) tail length can vary from ~25 to 250 nucleotides^{175, 178}, and multiple PABP molecules can bind with a periodicity of ~30 nucleotides¹⁷⁹.

In addition to interacting with poly(A), PABP is a hub for a multitude of proteins that interact with one or another of its RRMs and/or its C-terminal MLLE domain¹⁷⁸. Paip1 is a translation initiation regulatory factor that interacts with two separate regions of PABP via two Paip1 motifs, one of which is a 12 amino acid PAM2 sequence¹⁸⁰. Several other proteins involved in translation and/or mRNA poly(A) metabolism also contain PAM2 sequences that bind the MLLE of PABP^{181, 182}. These 'PAM2 proteins' include stimulators and inhibitors of translational initiation (Paips 1 and 2 respectively), the translation termination/release factor eRF3, the poly(A) deadenylases (PAN3, and CCR4 complexes), and several others, and their regulated competition for PABP likely underlies important physiologic homeostasis mechanisms¹⁸¹. The PAM2 consensus sequence contains conserved aromatic and hydrophobic residues at key positions that when mutated diminish interaction with PABP MLLE¹⁸¹. LARPs 4 and 4B have a PAM2 just upstream of their La modules that was shown for LARP4 to be functional^{2, 12}.

Similar to Paip1, other PAM2 proteins contain a second region that interacts with one of the four RRM domains of PABP¹⁷⁸. LARP4 contains a second region, downstream of its La module (Fig. 1), that interacts with PABP although the interacting region of PABP is unknown¹². Mapping of interaction between LARP4B and PABP is consistent with this¹¹. Thus, the La module of LARP4 binds poly(A) RNA and is flanked by two PABP-interacting motifs, PAM2 and PBM¹² (Fig. 1).

Phylogenetic analyses suggests that acquisition of a PAM2 sequence proximal to the La module occurred recurrently during evolution of LARPs³. In addition to LARP4 and 4B, a PAM2 was also found just upstream of the La modules of plant LARP6b and 6c and shown to be important for PABP interaction³. We are unaware of any reports of human LARP6/Acheron interactions with PABP, consistent with lack of the same for plant LARP6a³.

LARP1 was shown to bind PABP in an RNA-independent manner^{37, 176}. Co-IP experiments showed that partial or full truncation of the DM15 region of LARP1 substantially reduced but did not ablate interaction with PABP⁴. A PAM2 sequence candidate, that resides in the predicted α1 helix of human LARP1 RRM¹ was identified⁶. Partial reduction of PABP co-IP with DM15-truncated LARP1 is consistent with two interacting regions⁴. It should be noted that this candidate lacks sequence spacing features important in other PAM2 motifs¹⁸¹ and it maps to a predicted α-helix in the RRM whereas other PAM2 peptides are unstructured upon interacting with PABP MLLE¹⁸¹. These features together with significantly decreased binding to PABP after mutation of a key aromatic residue in this putative PAM2⁶ suggests the possibility that this is an atypical interaction. Curiously, acquisition of a PAM2 or variant thereof within the La module of LARP1 would strengthen the prior noted association³. It is tempting to speculate that because its N-terminal region contains an eIF4G-like domain⁶, localization of the PAM2 within RRM would better ensure tight linkage of PABP and La module activity. In any case, we note that this PABP-interaction region maps to the α1 backside of the canonical RNA-binding surface, a region of human and S. pombe La proteins to which residues required for nuclear export reside²⁴. In any case, it will be important to address if this LARP1 candidate PAM2 sequence motif is mediating poly(A) RNA-independent interaction with PABP and if so via which domain.

The m⁷GpppN cap, translation initiation factors, and 5'TOP mRNAs

Genetic and biochemical studies of yeast have been invaluable toward understanding basic mechanisms of translation and regulation thereof, including central importance of the m⁷G cap-binding factor, eIF4E and associated proteins, PABP and mRNA poly(A) metabolism¹⁸³. However, certain differences between yeast and mammalian cells in their mechanistic and regulatory control of translation are pertinent here ^{e.g., see \165}. The mRNAs that encode the ~90 ribosome protein subunits and other translation factors, including PABP are each present at high copy number and collectively comprise ~20% of total cellular poly(A) mRNA¹⁸⁴. In mammals these share a consensus known as the 5' terminal oligopyrimidine (5'TOP) motif¹⁸⁴ (not found on the analogous yeast mRNAs). In addition to a 5' tract of 5–15 pyrimidines followed by G+C-rich sequence, the 5'TOP motif is distinguished by a C at the 5' position, whereas the ~80% of other mRNAs have A or G. Thus, 5'TOP mRNAs bear a m⁷GpppC-cap (5'cap-C) while the others have a m⁷GpppG/A-cap (5'cap-G)¹⁸⁴ (Fig. 9A, right lower).

The mammalian target of rapamycin (mTOR) is a protein kinase with multiple substrates that promotes cell growth and proliferation in response to nutrition-associated factors such as insulin and other cues¹⁸⁵. Effects of mTOR are to increase the protein synthetic capacity of cells in part by regulating the 5' cap-binding protein, eIF4E, and promoting translation of 5'TOP mRNAs that encode ribosome subunits and translation factors. Positive regulation of eIF4E activity, as with stimulation of mTOR activity after treatment of cells with insulin involves its release from the eIF4E-binding proteins (4E-BPs), which are repressors of translation initiation^{186, 187}.

Poly(A) binding by some La modules

As alluded to above, C. elegans LARP1 preferentially binds poly(G) over the other homopolymers whereas hLARP1 was found to associate with poly(A) but not G, C or U in an extract-based system¹¹³. The basis of this species-specific difference is unknown and emphasizes the need for more well defined RNA binding analyses of the La module. Comparison of the four homopolymers as well as RNAs with various combinations of mixed nucleotides, identified poly(A) as a preferred ligand of the hLARP4 La module¹² (S. Gaidamakov & RJM, unpublished). Photoactivatable ribonucleoside–enhanced crosslinking and immunoprecipitation (PAR-CLIP)¹⁸⁸ was used to evaluate hLARP4B for in vivo RNA binding sites¹³. Because this crosslinking method requires a UV-activatable Uridine analog, 4-thioU, it introduces strong bias for U in the target sequence. Nonetheless, the top eight LARP4B binding sites isolated, comprising 76% of all of the target sites quantified, are composed of 66% A residues and 27% U residues¹³. Thus, despite an intrinsic bias for U imposed by PAR-CLIP, LARP4B revealed a preference for A-rich RNA¹³. To the best of our knowledge LARP4B has not been examined for relative binding to the homopolymers, nor has LARP4 been subjected to PAR-CLIP.

The well-recognized binding specificity of hLARP6 is a conserved stem-loop in the 5' UTR of the collagen mRNA^{9, 10} and although it is capable of interacting with oligo(U), oligo(A) and oligo(G) these interactions are between 30 and 100 fold weaker than with collagen SL⁹ (U. Abongwa & MRC, unpublished). Of the plant LARP6 members tested, AtLarp6a, exhibited preference for poly(A), which is curious because this protein has conserved all of the six key LaM residues (see Fig. 4 and below), while AtLarp6c preferred poly(U)³.

The La module of hLa exhibits binding to poly(A) with apparent affinity similar to LARP4 (S. Gaidamakov & RJM, unpublished). Also, hLa exhibits poly(A) binding in the cytoplasm (M. A. Bayfield, manuscript in preparation). This suggests that the La module may have some propensity for poly(A) recognition that further evolved for some of the cytoplasmic LARPs.

Cytoplasmic La Protein

The great majority of La protein is obviously nuclear by cell staining. La is soluble and unlike many nuclear proteins it readily leaks out of nuclei upon cell lysis. This feature as well as the transient nature of its binding (i.e., dissociation) to nascent transcripts noted above, led some researchers to question if La had any function in the cytoplasm of living cells. However, several studies helped characterize cytoplasmic La as a distinct entity of functional significance. As noted above La contains multiple subcellular trafficking elements and was demonstrated to shuttle from nuclei to cytoplasm^{24, 27–29, 33}. It was known early that hLa was targeted to small nuclear (sn) transcript precursors of cytoplasmic tRNA and 5S rRNA and U6 snRNA synthesized by RNAP III¹⁸⁹, as well as the small VAI/II and EBER RNAs encoded by the adenovirus and Epstein-Barr DNA viruses respectively, also transcribed by RNAP III^{190, 191}. However, work on La and RNA viruses demonstrated that only the cytoplasmic fraction of La associated with short leader transcripts synthesized by the viral RNA-dependent RNA polymerase¹⁹². Nonetheless La exhibited preference for the leader transcript with the longer 3' U tail whereas the viral N protein showed no preference¹⁹². Another development that distinguished cytoplasmic La were antibodies specific for phosphoSer-366-specific and nonphosphoSer-366-specific isoforms of hLa which demonstrated that these discrete isoforms were separately localized in the nucleoplasm and cytoplasm respectively²⁶. Parallel biochemical analyses suggested that ~80% of hLa is phosphorylated on Ser 366²⁶. Since ours and other data indicate about 2 × 10⁷ molecules of La per HeLa cell¹⁹³, the amount of cytoplasmic nonphosphoSer-366 was estimated at ~10⁶ molecules/cell.

Multiple phosphorylation sites were mapped to hLa, Thr-302, Ser-325, Thr-362, Ser-366 and Thr-389^194–196 and some were shown to control an activity in vitro and/or in vivo and/or to be of regulatory significance^{18, 25, 26, 50, 197–200}. Of the multiple phosphorylated sites characterized by Pruijn and coworkers, it was noted that Ser-366 was by far the most abundant¹⁹⁴ consistent with the major nuclear isoform associated with pre-tRNAs²⁶. It is important to note that phosphoSer-366 does not direct the nuclear residence or subcellular localization of La protein^{50, 194}. Rather, Ser-366 phosphorylation status more likely reflects colocalization and/or accessibility of the responsible kinase, CKII¹⁹⁹ and relative paucity of phosphatase. This further suggests that upon relocalization from nucleus to cytoplasm, La phosphoSer-366 may be dephosphorylated although this has not been examined. In any case, it has been shown that the phosphorylation status of Ser-366 does not influence the nuclear import or nuclear retention of La, and nor does a nonphosphorylatable hLa-A366 mutant distribute differently from wild-type hLa¹⁹⁴. In addition, although the other phosphorylation sites may appear minor relative to total La levels they could be a substantial fraction of cytoplasmic La. Human La also undergoes sumoylation of the LaM on K41 which promotes its retrograde transport from neuronal axon processes to the nucleus²⁰¹, and on two sites between RRMs 1 and 2 which decrease its binding to 5' TOP mRNAs and internal RNA elements²⁰².

Phosphorylation status of hLa Ser-366 controls SBM activity

Contrary to the lack of effect of Ser-366 phosphorylation on subcellular distribution is its striking effect in vitro and in vivo on RNA binding and related activities mediated by the adjacent SBM region C-terminal of RRM2α (Figures 1, 2). As noted above, nucleoplasmic La is phosphorylated on Ser-366, but when it is forced to be nonphosphorylated, biochemical and genetic evidence indicate that the SBM confers high affinity recognition of the 5' end of pre-tRNAs, and leads to inhibition of their processing by RNase P (in vitro and in vivo) that is alleviated by Ser-366 phosphorylation^{49, 50}. Analysis of tRNA alleles epistatic for RNase P processing support these findings⁵⁰ and La-A366 mutants that can not be phosphorylated in vivo are dominant negative for pre-tRNA maturation^{50, 203}. Direct binding showed that hLa exhibited higher affinity for RNA with 5' triphosphate vs. 5' dephosphorylated RNA, that the 5' triphosphate moiety increased La's inhibitory effect on processing by RNase P, and that inhibition required an intact SBM region in hLa⁴⁹.

This 5' triphosphate binding that was localized to the SBM of hLa is sensitive to methylation of the γ-phosphate of the 5' triphosphate as is found on the 7SK, U6 and a few other RNAP III transcripts, and is also weakened by addition of the m⁷G cap found on mRNAs¹³². The SBM and sequences up to position 366 are part of an unstructured region that is basic whereas amino acid side chains following Ser-366 are acidic. It is plausible that phosphorylation of Ser-366 would alter the local conformation of this region and inhibit the RNA-binding activity of the SBM^{see figure 3 in25 and figure 5 in 18}. Again, nucleoplasmic La is phosphorylated on Ser-366 and does not interfere with pre-tRNA processing whereas the cytoplasmic form is nonphosphorylated on Ser-366²⁶.

Cytoplasmic La and the short noncoding leader transcripts of RNA viruses

The first identified specific association of La with an RNA other than from RNAP III was the short leader noncoding transcript of the negative strand RNA virus, vesicular stomatitis virus (VSV)¹⁹² Similar results were soon confirmed for La specific association with short leader transcripts of additional RNA viruses including the deadly rabies virus^{204, 205} and extended by others to the short noncoding leaders of Parainfluenza and Rinderpest viruses^{206, 207}. These short 5' triphosphate-containing leader RNAs of negative strand RNA viruses are the first transcripts synthesized by the RNA-dependent RNA polymerases and are not modified with a m⁷G -cap, although the subsequent longer protein-coding transcripts are²⁰⁸. General commonalities of these leader RNAs are short length, ~50 nt, tendency toward A+RU-rich 3' tails, and 5' ppp termini. Also notable is that the leader RNAs originate from the 3' end of the viral template RNA that VSV and others cause redistribution of La to the cytoplasm, and that the VSV leader was shown to be bound by cytoplasmic La¹⁹².

The negative strand RNA viruses cause diverse diseases that range in severity from mild to lethal. They typically produce 10⁵ copies of RNA-dependent RNA polymerase in the cytoplasm of an infected host cell within hours of infection²⁰⁸. As with other infectious bodies, foreign material, especially cytoplasmic non-self RNAs, can be strong activators of host cell antiviral innate immune response. Among other host cell RNA sensors RIG-I contributes to the early induction of antiviral type I interferons in negative strand viruses while another RNA sensor, MDA5 is important for a detecting a distinct overlapping set of viral RNAs^{208, 209}. RIG-I requires a 5' triphosphate on the RNA for activation of its interferon response activity^{210 reviewed in208}. The 55 nt noncoding leader of the measles virus ends with a 3' CUU-OH terminus and contains a 5' triphosphate that is required for RIG-I activation²¹⁰. Viruses have been in an arms race with hosts and have collectively developed numerous ways to avoid or delay detection by host cell surveillance, in some cases by hijacking host factors^{208, 211}.

The negative strand RNA, respiratory syncytial virus (RSV) is a major cause of mortality in infants on all continents, second only to malaria as a lethal pathogen worldwide ^see212. The RSV 44 nt noncoding leader RNA contains a 5' triphosphate, ends in an A-rich sequence with a U-OH 3' terminus and is efficiently bound by La early in infection²¹³. RSV infection causes redistribution of La to the cytoplasm; and binding of leader RNA by La facilitates RSV proliferation²¹³. Decreasing La levels by knockdown increases leader RNA association with RIG-I and interferon induction²¹³. Thus, it would appear as if the leader RNAs of negative strand viruses engage cytoplasmic La to mask their 5' triphosphate danger signal which would otherwise activate the surveillance factor, interferon activator, RIG-I. These viruses hijack La to shield their leader transcripts from detection and attenuate early activation of innate immune signaling²¹³.

Cytoplasmic La and virus mRNA IRES-dependent translation

That La associated with viral leader transcripts prompted studies of additional RNA viruses that were found to use La for translation of their mRNAs, including poliovirus, HIV and HCV ^214–220. Some of these, including poliovirus do not m⁷G-cap their RNAs nor use the cap-dependent mechanism of translation initiation and instead use an internal ribosome entry site (IRES) structural element that encompasses the start site AUG to direct initiation²²¹. Indeed, viruses that use IRES-mediated initiation employ mechanisms to deactivate cap-dependent initiation and other means to direct the translation machinery to viral protein production²²². Poliovirus, HCV and other viruses rely on a number of host factors to promote IRES-mediated translation of their mRNAs²¹¹. La was the first such host factor found that promoted IRES-mediated translation initiation^{214, 215}. La function in IRES-mediated mRNA translation was demonstrated in vivo for HCV mRNA²²³ and was shown to promote IRES-mediated translation of several viral and cellular mRNAs^{224 see 211, 225}. Of the ~10 proteins produced by poliovirus, two proteases target host cell cap-dependent initiation factors for cleavage. Protease 3C substrates include PABP and La, the latter of which is cleaved at position 358, which removes Ser-366 and the downstream NLS resulting in cytoplasmic accumulation of a high affinity RNA binding form of La presumably to facilitate virus mRNA translation^{226, 227}.

Biochemical experiments using mammalian translation systems reconstituted with purified factors indicate that La can enhance eIF4E-, PABP- and m⁷G cap-dependent translation by inhibiting spurious initiation at aberrant internal start sites²²⁸. Moreover, inhibition of cap-dependent translation is exacerbated by increasing La as PABP becomes limiting²²⁹. These studies provide insight into La function as a positive factor in directing IRES-mediated translation initiation when cap-dependent mechanisms are inhibited.

Of note are recent advances toward understanding the roles of La in the translation and replication of HCV RNA virus^230–235. These include mechanisms of HCV RNA recognition by La, granzyme H-mediated redistribution of La to the cytoplasm, and effects on telomerase activity in HCV infected cells^{127, 236, 237}. Some studies have led to development of compounds to target La by pharmaceutical inhibition of HCV²³¹. Another line of work has established that La binds and stabilizes Hepatitis B virus RNA^{129, 238}. Phosphorylation of hLa Ser-366 by protein kinase CKII is required for the positive effect on HBV expression¹⁹⁸.

La and TOP mRNAs

As an oligo(U) binding protein it was reasonable that La may function in translation of 5'TOP mRNAs^{239, 240}. Subsequent investigations showed that cytoplasmic nonphosphoSer-366 La is specifically associated with 5' TOP mRNAs²⁶ and that increasing the fraction of nonphoshoSer-366 by inhibiting protein kinase CKII increases the amount of 5'TOP association¹⁹⁹. As noted above, while 5'TOP mRNAs are efficiently translated in proliferating cells their physiologic regulation involves translational repression in times of metabolic and/or growth quiescence. Consistent with this, in a Xenopus cell line, several 5'TOP mRNAs were 80% and 40% loaded on translating polysomes in growing and resting cells respectively²⁴¹. In the same study, over-expression of wild type (wt)La and a dominant-negative (dn)La had no effect on 5'TOP mRNA translation efficiency in growing cells whereas in resting cells wtLa increased translation efficiency from 40% to 50% and dnLa decreased it from 40% to ~30%²⁴¹. Thus, these data indicate that La promoted 5'TOP mRNA translation but only in the resting cells, in what would appear to be under repressing conditions.

Other studies have indicated a negative impact of La on 5'TOP mRNA translation^{199, 242}. Later experiments revealed no specific effects of La on the translation efficiency of 5'TOP mRNAs ^see243.

It is conceivable that with over-expression of La or under conditions in which La redistributes to the cytoplasm, and when m⁷G cap-dependent translation is limited or weak, the 5'-ppp cap like binding activity of the SBM of nonphosphoSer-366 La could negatively modulate m⁷G cap-dependent translation initiation^{199, 228, 229, 242} and promote IRES translation.

La chaperone activity in mRNA translation

Some cellular mRNAs use IRES elements of one type or another to direct cap-independent translation initiation²²⁴ and several of these have been shown to engage La for optimal function^{18, 130, 244–247}. La can therefore function as an mRNA translation factor, possibly employing both its N-terminal and C-terminal fragments both of which harbored comparable RNA chaperone activities²⁴⁸.

It was known from early studies that the ability of La to promote IRES-mediated translation was dependent on amino acids 293–380^{215, 217, 249}. La was shown to promote IRES-mediated translation of cyclin D1 (CCND1) mRNA in tumors¹³⁰. A significant advance toward linking the RNA chaperone and translation activities of La came from studies of the CCND1-IRES mRNA¹⁸. This IRES mRNA chaperone activity was mapped to amino acids 333–390 of hLa and regulated by AKT-mediated phosphorylation of Thr-389¹⁸. It was noted that the disordered C-terminal region of yeast La that is important for correct folding of tRNA may bear some similarity to this region^{17, 18}. As noted in an earlier section, this is in a very intricate region of hLa, comprised in part of the short basic motif (SBM, Fig. 1, 2) proposed to share features with a Walker A-like nucleotide-binding motif⁴⁸ and found to interact with the 5' pppG end of nascent pre-tRNAs as part of a phosphoSer-366-mediated control of their 5' processing by RNase P^{49, 50}, and to overlap with the nucleolar localization signal^{25, 32}.

Recent Advances on LARP6 Family Members

Brief overview of LARP6 phylogeny

LARP6 (a.k.a. Acheron in metazoa) members are characterized by a La module followed by a LARP6-specific domain, denoted LaM and S1-Associated (LSA) due to homology to part of a repeat found in S1 ribosomal protein and structurally related nucleic acid binding folds¹. LARP6 is encoded by one gene in protists, invertebrates and eutherians, two in other vertebrates and 3–6 in vascular plants³. The phylogeny of LARP6 appears more complex than for other LARPs. In higher plants there are three LARP6 subclusters, a, b and c, of which b and c contain N-terminal PAM2 peptides for PABP binding, and these have been characterized for the Arabidopsis thaliana (At)LARP6 proteins³. As noted, phylogenetic emergence of the PAM2 was associated with divergence of some of the key hydrophobic pocket residues in AtLARPs 6b, 6c and LARPs 4, 4B³. Thus, plant LARP6a and hLARP6 are most comparable, whereas plant LARPs 6b and 6c are more distinct from these³.

We should clarify that LARPs 6b and 6c exist in plants but not in vertebrates, LARPs 4 and 4B exist in vertebrates but do not exist in plants, whereas LARP6/a is common to vertebrates and plants. Thus, vertebrates and plants each contain two LARPs that share N-terminal PAM2 sequences and whose LaM hydrophobic RNA binding pockets underwent partial remodeling but are of different LARP subfamily memberships³. It seems plausible that the two PAM2-containing LARPs in vertebrates (LARPs4/4B) and plants (LARPs6b/c) participate in similar mRNA-associated functions.

Arabidopsis thaliana LARPs6

The three AtLARPs6 likely fulfill distinct functions as mRNA-associated proteins with different subcellular localizations and RNA- and PABP-interaction networks³. AtLARP6a mainly localizes to nucleolus and cytoplasm, recognizes poly(A), does not bind or colocalize with PABP nor relocate to stress granules upon hypoxia³. AtLARP6b has similar localization as 6a but associates with PABP and colocalizes to stress granules³. AtLARP6c binds PABP and goes to stress granules, but was also observed in nucleolus, nucleoplasm and cytoplasm³. RNA interaction profiles of two AtLARPs6 show clear differences: 6a prefers oligo(A) while 6c prefers oligo(U) (any preference by 6b is unknown due to limitations of protein stability)³. Also as noted, while AtLARPs 6b and 6c bind PABP, AtLARP6a does not³.

Functional studies on metazoan LARPs6

Studies on the tobacco hawkmoth, Manduca sexta, first uncovered LARP6, referred to as Archeron, as critical for myogenesis in a manner that involves programmed apoptosis; a role in myogenesis for LARP6 was also found in mice and fish^{35, 250, 251}. In vertebrates, LARP6 has been reported to control muscle differentiation and development³⁵ by acting upstream of the key myogenic transcription factor MyoD²⁵⁰. LARP6 may regulate myogenesis via interaction with the CaMKII-like domain of the developmental transcription factor CASK-C and/or association with the Id (Inhibitor of differentiation) transcription repressors²⁵². Since Ids modulate regulatory factors such as MyoD²⁵³, LARP6 may contribute to myogenic differentiation through interactions with Id proteins. Indeed, ectopic MyoD can rescue differentiation in C2C12 myoblasts expressing a dominant-negative form of LARP6²⁵². The complex LARP6/CASC-K/Id may therefore regulate development at the transcriptional level, thereby assigning this function to the nuclear fraction of LARP6. While it is unclear if the above function of LARP6 entails interaction with yet unknown RNA targets, specific RNA binding is unequivocally required for its function in regulated synthesis of collagen α1(I), α2(I) and α1(III) chains, which as reviewed in an earlier section, is mediated by interaction with the SL in the 5'-UTRs of these mRNAs¹¹⁵. These 5’ SL sequences encompass the translation initiation codon and are highly conserved in all vertebrates²⁵⁴.

The role of LARP6 in orchestrating collagen production was recently reviewed²⁵⁵. While synthesis of type I collagen is critical for normal growth and regulated response to wound healing, excessive production can cause fibrosis and be pathologic. Disruption of the LARP6-SL colα1-mRNA interaction decreases collagen production, making LARP6 appealing as a target for treatment of fibroproliferative disorders²⁵⁶. On the other hand, the LARP6-SL interaction is a critical parameter in mediating protective collagen synthesis in atherosclerosis by reducing vulnerability of plaque rupture. In this case, insulin-like growth factor-1 (IGF-1) regulates LARP6 and resultant collagen expression in vascular smooth muscle cells²⁵⁷. It is notable that this study showed that a Colα1 5' SL RNA decoy could block collagen synthesis in response to IGF-1²⁵⁷.

At a molecular level, the La module of LARP6 is proposed to facilitate coordinated translation of collagen α1(I) and α2(I) mRNAs, which occurs with 2:1 stoichiometric ratio, by binding to a bulge in the SL and modulating accessibility to the translation start codon located in the adjacent double helix stem¹¹⁵. LARP6 also participates in trafficking of these mRNAs in the nucleus and cytoplasm¹¹⁵. It was postulated that LARP6 interacts with a number of molecular chaperones to choreograph collagen production, including non-muscle myosin²⁵⁸ SEC61 translocon²⁵⁹, vimentin²⁶⁰, and RNA Helicase A (RHA) ^{261 reviewed in 255}. LARP6 has also been reported to recruit the serine-threonine kinase receptor-associated protein (STRAP) to collagen mRNA, thereby preventing uncontrolled translation mainly of collagen α2(I) mRNA by a mechanism thought to involve eIF4A²⁶². STRAP (a.k.a. unrip) belongs to a family of WD-repeat proteins and was found to interact with the last 27 amino acids of hLARP6²⁶², which map to the LSA domains¹ (Fig. 1). Importantly, in addition to its significance with regard to STRAP, this work also attributes a specific protein-protein interaction function to the LSA motif. On the other hand, vertebrate LARP6 members share with numerous other proteins a slightly larger region encompassing the LSA, termed SUZ-C ^{supp figure 2 in 263}, that was proposed to participate in RNA binding by the SYZ-20 protein. In hLARP6, SUZ-C maps to residues 453–488. Clarification on the functional role of the LSA in the LARP6 family of proteins awaits further investigations.

LARP6 appears to play a role in cancer as it enhances proliferation, lamellipodia formation and invasion activity of breast cancer cells. These functions appear linked to nuclear activity of LARP6, as loss of its NLS abrogates these effects³⁴. Interestingly, the effects on cell morphology and motility appear to be the reverse of what is observed with LARP4, where cell motility was increased by depleting LARP4²⁶⁴ (below).

LARP6 function in collagen synthesis can be modulated by phosphorylation^{265, 266}. Of note, hLARP6 appears to be a downstream target of Akt and phosphorylated on S451, although a known consensus sequence for Akt is not recognizable surrounding S451²⁶⁵, suggesting that it may be phosphorylated by a kinase downstream of Akt. In any case, S451A mutation or inhibition of Akt, reduces the type I collagen secretion²⁶⁵. Furthermore, hLARP6 was reported to be phosphorylated on S348 and S409 by mTORC1, and in cells where the phosphorylation is prevented by mutating these serines to alanines, or by mTORC1 inhibition, collagen secretion decreases and hyper-modified collagen polypeptides are produced²⁶⁶. Also, lack of S348/S409 phosphorylation appears to weaken the interaction of hLARP6 with STRAP, although these serines are far from the putative STRAP interaction site on hLARP6 (464–491). Interestingly, LARP1 and LARP4 as well as many other proteins are also phosphorylated by mTOR²⁶⁷.

Recent Advances on LARPs1

LARP1 was discovered prior to other LARPs and has been studied in a variety of organisms including C. elegans, drosophila, plants, humans and mice. Several early studies, although not limited to those²⁶⁸, uncovered LARP1 through genetic screens to better understand developmental programs or other processes^{37, 269–271}. A recent review covers phylogeny and function of LARP1 in reproduction, proliferation and oncogenic potential, as well as emerging links to translation, mRNA metabolism and stress response pathways²⁷². LARP1 has also received directed focus as a tumor progressor in human cancers, consistent with its mechanistic links to the mTOR pathway that controls cellular growth and proliferation in response to nutritional status, hypoxia and other states relevant to cancer¹⁸⁵, and has also been reviewed along with the other LARPs from this perspective³⁶. In keeping with the aims of this review we will focus on recent advances on the structure and function aspects of LARP1 and its mechanism of action as an RNA binding protein.

Advances came from phylogenetic comparisons and recognition of the conserved region initially referred to as the LARP1 domain^{269 and refs therein} and more recently, DM15 as will be used hereafter¹. A conserved region corresponding to DM15 was previously noted in La-related homologs of human, C. elegans, and two yeast proteins, Sro9p and Slf1p although neither of the latter contain DM15¹¹⁹. The DM15 box is a ~40 amino acid motif found in three consecutive repeats in the LARPs1 of protists, fungi, plants and animals that was shown to bind 5'TOP RNA¹³³.

C. elegans LARP1 fragments containing its La module or DM15 region were examined for RNA binding²⁶⁹; the latter exhibited higher avidity for poly(U) consistent with recent 5'TOP binding¹³³ whereas the La module-containing fragment showed preference for poly(G)²⁶⁹. An approach to identify new RNA-binding proteins, employing bait constructs carrying 3' poly(A) either with or without a 5'- m⁷GpppG cap was used on native human cell extracts¹¹³. Mass spectrometry analysis showed that LARP1 was higher ranking over three other proteins, PARN, PATL1 and eIF4A1, that showed preference for poly(A) specifically at the 3' end of the bait¹¹³. Using non-capped RNA, LARP1 was bound preferentially by A(10) with specificity for A at the 3' end but not A(9) followed by a single U, C or G¹¹³. Although the assay as used cannot ensure direct binding to the RNA bait by LARP1, this specificity is reminiscent of RNA 3' binding during protection from 3' exonucleases by La protein although the mechanisms remain to be distinguished¹¹³. Accordingly, because of the noted relation of LARP1 to mRNA metabolism^{5, 6, 113, 269, 273–275}, one may suspect that its La module might protect against deadenylation (Fig. 9B). Cytoplasmic mRNA decay can be initiated by endonucleolytic cleavage or 3' exonucleolytic poly(A) shortening²⁷⁶, and followed by 5' and/or 3' exonucleolytic digestion^{174, 277, 278}. In plants, LARP1 associates with the 5′-3′ exonuclease XRN4 and is required for a heat-activated mRNA decay via 5'-decapping although if this pathway involves prior 3' deadenylation is yet unknown^{274, 275}.

Human LARP1, 5'TOP mRNA translation and mTOR

Affinity matrix of m⁷Gppp-sepharose was used to identify cellular proteins that associate with the m⁷G-cap in an mTOR-dependent manner⁴. This revealed PABP as the isolated protein most increased by insulin, followed by other translation factors as expected based on earlier studies. It also identified hLARP1, which was increased by insulin and decreased by mTOR inhibitors, in both cases concomitant with reciprocal changes in the 4E-BPs⁴. In this study, hLARP1 was complexed with PABP, and was associated in polysome profiles made from serum-stimulated cells with 40S–80S ribosomes as well as in the same fractions as PABP in polysomes. Also, in hLARP1 knock down cells, 5'TOP mRNA levels were substantially reduced⁴. The results provide strong evidence that LARP1 associates with and stabilizes 5'TOP transcripts, which comprise a substantial fraction of the total pool of actively translating mRNAs in stimulated cells. The data were also consistent with a positive role for LARP1 in translation of 5'TOP mRNAs.

Fonseca and colleagues also examined hLARP1 interactions in response to mTOR and its inhibitors⁶. In agreement with Tcherkezian et al.⁴, they found direct interactions with PABP and Raptor⁶ and also showed interactions with 5'TOP mRNAs. TOR inhibition decreased LARP1-Raptor and increased LARP1–5'TOP mRNA association, consistent with the identification of a TOR signaling (TOS) motif in the DM15^{6 and refs therein, 133}. Both studies showed most LARP1 in the RNP region as well as 40S–80S region, with less on polysomes with translating mRNAs and PABP^{4, 6}. In contrast to Tcherkezian et al. who propose a stimulatory role for hLARP1 in 5'TOP mRNA translation, data in Fonseca et al. led to a model in which LARP1 represses 5'TOP mRNA translation in response to mTOR activity. It has been proposed that LARP1 may function to either stimulate or inhibit translation, dependent on phosphorylation status or competitive binding to other components of the translational or mTOR machinery³⁶.

As noted previously for LARP1, distinct RNA recognition modes by DM15 and the La module could juxtapose the 5'TOP and 3' poly(A) and contribute to a closed loop configuration of mRNA, presumably facilitated by direct LARP1-PABP interactions³⁶. New evidence shows that in addition to binding to the 5'TOP sequence, the DM15 of hLARP1 also binds to the m⁷GpppC (cap-C) and sequesters the 5'TOP motif away from eIF4E, thereby preventing the formation of eIF4F complexes on 5'TOP mRNAs¹³⁴. In this capacity LARP1 would inhibit translation initiation by the eIF4F pathway¹³⁴ (Fig. 9B). In this model, the La module of LARP1 would protect the poly(A) 3' tail from exonucleolytic decay and DM15 would protect the 5' cap from exonucleolytic decay while the 5'TOP mRNA is in a state of repression¹⁸⁴.

Mitochondria-targeted mRNA translation and Drosophila LARP1

Recent work identified LARP1 as a translation stimulator for a set of embryo mRNAs whose protein products must be targeted to the inner mitochondria for D. melanogaster viability²⁶⁸. Earlier work identified dLARP as colocalized with mitochondria²⁷⁰. The mitochondrial outer membrane protein MDI, was found to bind LARP1 in association with PABP and eIF4G, and is required for insertion of cytoplasmic translated proteins that are essential for mtDNA replication into mitochondria²⁶⁸. Remarkably, targeting of LARP1 to the outer mitochondrial membrane in the absence of MDI restored protein synthesis to mitochondria and rescued mdi-mutant phenotypes²⁶⁸.

The 5'TOP motif is conserved among vertebrates and extends to Drosophila¹⁸⁴. A cursory examination revealed that the D. melanogaster cytoplasmic mRNAs that are targeted by LARP1 for mitochondrial outer membrane translation do not show evidence of enrichment of the 5'TOP motif (RJM, unpublished). The work on mitochondria²⁶⁸ provides biological evidence that D. melanogaster LARP1 functions to promote translation in a 5'TOP motif-independent manner.

LARP1 DM15 mediates m⁷GpppC-TOP mRNA interactions

As described above biochemical and biophysical analysis of DM15-RNA cocrystals are uncovering structure-function aspects of hLARP1 and insight into its translational control of 5'TOP mRNAs¹³⁴. Structural and biochemical analyses revealed that DM15 exhibits almost 100-fold greater binding specificity for TOP RNA bearing a 7mGpppC than a -G¹³⁴. Moreover, the DM15 domain could outcompete the cap-G binding factor, eIF4E, when the RNA bears a 7mGpppC and can displace eIF4E from RNA with 7mGpppC but not with a 7mGpppG¹³⁴. These results provide keen evidence that LARP1 is indeed a candidate for a most proximal translational regulator of 5'TOP mRNAs¹⁸⁴ and raise a bash of unanticipated questions. Do LARPs1 of other species exhibit similar activity in mTOR and nutrition control of 5'TOP mRNA translation according¹⁸⁴? It will also be interesting to see how this DM15 may be used to identify other cap-C -containing RNAs. Other questions arise. To what extent can LARP1 bind cap-C RNA that is less pyrimidine-rich than typical 5'TOP mRNAs? What is the nature of the cap-C transcriptome, is it tissue- and species-specific, and to what extent is it under the control of LARP1? What is its stoichiometry relative to the 5'TOP mRNAs; in practical terms, to what extent must mRNAs compete for LARP1? Does LARP1 protect mRNA poly(A) from destabilization during TOR inhibition?

It is interesting to note from phylogenetic perspective that the DM15 region as a 5'-pppCap-C-RNA binding motif is functionally evocative of the C-terminal region SBM in La that binds to the 5'-pppG-end of nascent pre-tRNAs and is sensitive to m⁷G-cap^{49, 132}.

LARP1 and cancer

There is strong association of LARP1 with cancer ^{reviewed in36 see 273}. In this capacity LARP1 appears to bind to a multitude of mRNAs including that encoding mTOR, the latter of which was shown to be elevated and stabilized by LARP1⁵. Although direct binding between LARP1 and PABP may account for much of the LARP1-interactome, it is also possible that LARP1 exhibits sequence-specific direct recognition of some of the mRNAs although this remains undetermined.

The TOR pathway regulates ribosome biogenesis to meet the needs of nutrition and factor-dependent cell growth and proliferation and includes transient repression of 5'TOP mRNA translation when appropriate^{184, 185}. Ribosome biogenesis is a hallmark of cancer^{185, 279}. It will be interesting and challenging to see how cancers not only overcome the otherwise inhibitory role of LARP1 in translation initiation of the 5'TOP RNAs^{6, 134}, but also use LARP1 toward their benefit³⁶.

Recent Advances on LARP4 Family Members

Protists and invertebrates have one LARP4 gene, whereas gene duplication appears to have established independent lineages for LARP4 and LARP4B (previously known as LARP5) paralogs in an ancestral vertebrate^{1, 3}. Zebrafish is the only known organism with more than two LARP4 genes³, designated LARPs 4aa, 4ab and 4b. No LARP4 homologs appear in plants or yeasts^{1, 3}.

Human LARPs 4 and 4B are cytoplasmic proteins encoded on chromosomes 12 and 10, respectively and although they share general overall architecture (Fig. 1), amino acid identity is 40% throughout including 74% in their La modules. They vary in degree to which their other motifs have been mapped^{11, 12}. Two PABP-interaction motifs flank their La modules and they also interact with the receptor for activated kinase C (RACK1), an integral component of the 40S subunit of the ribosome^280–282. Thus, LARPs 4 and 4B interact with PABP and RACK1, associate with polysomes, promote mRNA translation and localize to stress granules (SGs)^{11, 12}. LARP4 expression stabilized reporter mRNA and the cellular mRNAs for which it was examined and found to be associated, e.g., FAIM¹².

LARPs4 with distinctive LaM binding pocket sequence and RNA preference

The hLARP4 La module was examined for binding to different RNA homopolymers using gel-shift assays and isothermal titration calorimetry (ITC) which revealed specificity for poly(A) of ≥15 nt¹². As detailed in an earlier section, PAR-CLIP revealed that hLARP4B was bound to A-rich sequences; the same study identified numerous cellular mRNA targets that were stabilized by hLARP4B¹³ some of which were confirmed in other cells²⁸³. As noted previously and shown in Fig. 3, the LaMs of vertebrate LARPs4 differ from each other, La, and other LARPs in some of the six otherwise highly conserved key amino acids involved in UUU-3'OH recognition^{2, 3} (Fig. 4). Human LARPs 4 and 4B differ from La and the other LARPs by having a cysteine at position 24 (hLa numbering) whereas all others have the aromatic Tyr (or Trp, LARP7) (Fig. 4) contributing to the hydrophobicity of the binding pocket and making hydrogen bonds to the phosphate of U-1 (Fig. 3C)^{106, 108}. For hLa and TbLa, mutation of Y24 was severely detrimental to RNA binding^{44, 108}. Moreover, hLARPs 4 and 4B differ from each other at 2 other of the six LaM binding pocket residues, Q20 and F55 (Fig. 4). LARP4 has conserved Q20 whereas LARP4B has T at this position (Fig. 4) and notably Q20 in hLa and LARP7 make base-specific contacts to U-2 which resides in the cleft between the LaM and the RRM1^{106, 108} (Fig. 3B). Regarding the F55 position, hLARPs 4 and 4B contain M and L respectively in place of this otherwise nearly invariant residue (Fig. 4). LARPs 4 and 4B also differ in 10 amino acids in their RRMs. It would seem that these differences may contribute to differential RNA preferences of LARPs 4 and 4B.

LARP4 activity in mRNA and poly(A) metabolism

Unexpected insight into LARP4 function came from examination of its negative regulation which identified a central region in its mRNA as a codon-biased coding region determinant (CRD) of instability that when substituted with synonymous codons, increased its mRNA and protein levels¹¹⁴. This revealed activity of cellular LARP4 to affect mRNA poly(A) tail (PAT) length¹¹⁴ concomitant with mRNA stabilization ¹² (Fig. 9C).

RACK1 is a WD40 family member that is stably associated with the mRNA exit channel of ribosomes from protists, yeasts and humans, via contacts to three 40S proteins as well as 18S rRNA^{280–282, 284, 285}. RACK1 is a hub for interactions with several proteins in addition to protein kinase C, mediated through its WD repeats²⁸⁶. The yeast RACK1 homolog was shown to enhance translation of short ORF mRNAs according to the closed loop structure model²⁸⁷. Using yeast-2-hybrid assays, LARP4 was shown to interact with RACK1 propellers 6 and 7¹². Reciprocal yeast-2-hybrid and biophysical methods have recently mapped the RACK1-interaction region (RIR) to a C-terminal sequence of LARP4 (Fig. 1) (unpublished).

LARPs 4 and 4B links to cancer

LARP4 would appear to regulate cancer cell migration and invasion. Upon its knockdown, prostate cancer cells exhibited very elongated phenotype consistent with increased migration²⁸⁸. LARP4 knockdown increased cell migration and invasion of other cancer cells, whereas overexpression decreased elongation²⁶⁴. Moreover, more than 140 mutations in LARP4 have been identified in a multiple cancers, and six of these, chosen because they reside in the C-terminal regions of interaction with PABP and RACK1, were subjected to examination in the experimental cell migration system. Overexpression of some of the mutants were more effective than WT LARP4 toward reversing the anti-invasion phenotypes²⁶⁴. Of these, the S388* truncation mutant exhibited elevated association with PABP, which as noted had also been observed for other LARP4 C-terminal truncation constructs¹², suggesting that a downstream region is inhibitory to PABP binding²⁶⁴.

LARP4B has been found to act as a tumor suppressor and oncogene, possibly dependent on the type of cancer, e.g., solid or blood. LARP4B was identified as a tumor suppressor by a genetic screen in mice²⁸⁹ and more recently characterized in human glioma cells^{283 for more comprehensive review see 290}. Overexpression of LARP4B induced apoptosis and mitotic arrest, partially dependent on the La module suggesting that interaction with RNA and its other interacting proteins is important²⁸³. Knockdown of LARP4B in primary mouse astrocytes lacking p53 and Nf1 promoted cell proliferation, tumor size and invasiveness²⁸³. In a study of an Acute Myeloid Leukemia mouse model, LARP4B was found to act as an oncogene²⁹¹.

LARP4 regulation

The hLARP4 mRNA 3’UTR contains conserved AU-rich element (ARE) binding sites for Tristetraprolin (TTP) protein which negatively regulates mRNA stability and translation^{292 see 293}. LARP4 mRNA and protein but not LARP4B transiently decrease then increase reciprocally with TTP levels in response to tumor necrosis factor (TNFα)²⁹², an apical cytokine in initiation of inflammation that up-regulates other cytokines²⁹⁴. Unrestrained TNFα activity becomes pathogenic unless controlled in part by TTP^{295, 296}. Regulation by TNFα-TTP and evidence of function in mRNA stability^{12, 292} suggest that LARP4 may contribute to mRNA metabolism during fine tuning of the inflammatory response²⁹⁶.

Human LARP4 mRNA contains a translation-dependent coding region determinant of instability that may be sensitive to cellular tRNA levels¹¹⁴. Experimental over-expression of limiting tRNA cognate to codons in this region can increase LARP4 protein levels and its activity to mediate poly(A) length maintenance of heterologous mRNAs including those that encode ribosomal proteins¹¹⁴ (Fig. 9C). As increased tRNA production occurs in cancer as well as in response to growth promoting signals including insulin through mTOR ^{see 297} it is possible that this may contribute to the regulation of LARP4 and its activity.

CONCLUSIONS

When the La-related protein (LARP) family was discovered less than a decade ago, most of its members were largely unknown. Conserved features were eye-catching but it soon became evident that each family had a distinct profile and that in some cases species-specific members of the same family clearly had evolved to perform specialized tasks. The LARPs would appear to be phylogenetically rooted in an ancestral eukaryotic La protein which functions via 3' end binding and chaperone activities directed toward the intranuclear handling of precursor tRNAs and other nascent transcripts of RNA polymerase III. As investigations proceeded, commonalities among La and the LARPs have been emerging as discussed in this review, the most clear to date being links to the 3' end regions of their RNA targets, which for the cytoplasmic LARPs include 3' poly(A) binding and stability, mRNA translation and interactions with PABP.

A good deal of high quality and focused research has revealed the LARPs as a key superfamily of proteins that share a unique and versatile RNA-binding module, that surely play important functional roles in eukaryotic biology that appear to operate at fundamental points in RNA control. Collectively the LARPs tell a tale of 3’ and 5’ recognition of RNAs, details of which have revealed novel ways of interacting with nucleic acids. We believe that a lot more will be uncovered, from details of molecular interaction with RNA ligands as well as protein partners, to additional roles in translation and possibly extending back to tRNA. New challenges await that we and others will relish to take on.

Abbreviations

5'TOP

5'-terminal oligopyrimidine

AtLARP

Arabidopsis thaliana LARP

HCV

hepatitis C virus

IRES

internal ribosome entry site

LaM

La motif

LARP

La-related protein

LARP1

La-related protein-1

LARP4

La-related protein-4

LARP6

La-related protein-6

LARP7

La-related protein-7

m7G-cap

7-methyl-guanosine-cap

PAM2

PABP-interaction motif-2

PABP

poly(A)-binding protein

RRM

RNA recognition motif

RNAP

RNA polymerase

SL RNA

stem-loop RNA

snRNA

small nuclear RNA

tRNAome

the tRNA gene content of a genome