3’ PROCESSING OF ANIMAL REPLICATION-DEPENDENT HISTONE mRNAs

INTRODUCTION

Each time a eucaryotic cell divides, it must first replicate its DNA, and at the same time also synthesize large amounts of the 5 classes of histone proteins to package the newly replicated DNA into chromatin. These two processes are tightly coupled in all eucaryotes, with histone mRNAs being primarily expressed in S-phase. Metazoans contain a specialized set of genes that encode replication-dependent RD-histone mRNAs, and the metazoan RD-histone mRNAs differ from all other eucaryotic cellular mRNAs. These mRNAs are not polyadenylated, but end instead in a conserved stemloop sequence (Fig. 1). The RD-histone genes do not contain introns, and the only processing reaction, other than capping of the 5’ end, required for their biosynthesis is cleavage of the nascent transcript to form the 3’ end of the histone mRNA. In this review, I first give a history of the critical discoveries that resulted in elucidating the mechanism of formation of the 3’ end of metazoan histone mRNAs. I then discuss the contribution that regulation of 3’ processing plays in cell-cycle regulation of histone mRNA, and the role of the histone locus body (HLB), a nuclear body containing the histone genes, in coupling transcription to 3’ processing of histone mRNA and coupling processing with transcription termination, in addition to a critical role of the HLB in cell-cycle regulation of histone mRNA expression.

Fig. 1.

Structure of replication dependent histone mRNAs.

The 3’ end of a replication-dependent histone mRNA is formed by endonucleolytic cleavage directed by two cis-elements: a conserved stemloop which binds the SLBP protein, and a Histone Downstream Element (HDE) which base pairs with the 5’ end of U7 snRNA. There are no introns in the transcript. Cleavage occurs 4 (invertebrates) or 5 nts (vertebrates) after the stemloop to form the 3’ end of the histone mRNA. Histone mRNAs have short 5’ UTRs (10-50 nts) and 3’ UTRs, with the SL 45-75 nts after the stop codon. This distance is required for proper regulation of histone mRNA half-life.

Metazoan histone genes.

The genes encoding the RD-histone mRNAs, which encode the bulk of the histone proteins that package the newly replicated DNA, are a unique class of genes in metazoans. The genes for the 5 different classes of histone proteins (H1,H2a,H2b,H3,H4) are physically linked and coordinately expressed, as large amounts of each protein is required for rapid assembly of chromatin after replication.

The early embryonic histone genes from two sea urchins, Strongylocentrotus purpuratus and Psammechinus milaris were the first protein coding genes from metazoans that were isolated, cloned and sequenced(Cohn, Lowry, & Kedes, 1976; Schaffner et al., 1978; Sures, Lowry, & Kedes, 1978), by the laboratories of Max Birnstiel and Larry Kedes. This was possible because the genes were present in a 2 megabase repeat containing 400 copies of each histone gene(Marzluff, Sakallah, & Kelkar, 2006), which comprises about 1% of the sea urchin genome. The histone genes were also isolated shortly after that from Drosophila melanogaster where there are 100 copies of each gene which are present in a single 500 kB cluster(Bongartz & Schloissnig, 2019) representing about 0.5% of the genome(Lifton, Goldberg, Karp, & Hogness, 1978). In both these species, the five histone genes are arranged in a tandem repeat, with each repeat unit (~5 kB), containing one copy of each histone gene. In Drosophila there is a bidirectional H2b-H2a and H3-H4 gene pair and a histone H1 gene, while in sea urchins all 5 histone genes are transcribed from the same strand of DNA (Fig. 2). The Drosophila repeat contains all the replication-dependent histone genes in the organism, while sea urchins have a second set of histone genes which are expressed starting in the mesenchyme blastula stage of development when the early histone genes are silenced, and a different set of histone genes is activated. In mammals there are two large clusters of histone genes, containing more than 10 non-allelic genes for each nucleosomal histone and 6 histone H1 genes (Fig. 2), and these clusters are conserved across mammals (Seal et al., 2022). The stem-loop at the 3’ end of histone mRNAs was recognized as a conserved feature of animal histone genes(Hentschel & Birnstiel, 1980), and the mRNAs did not bind to poly-dT cellulose suggesting that they were not polyadenylated

Fig. 2.

The genes for all five classes of histone proteins are linked. The histone genes in Drosophila are in a 5 kB tandem repeat containing one copy of each of the five histone genes and the genes expressed in the early sea urchin embryo are also arranged in a tandem repeat present in 400 copies in the genome. A second set of histone genes in sea urchins, expressed after the blastula stage and in adults are not in a repeat unit but are linked like the mammalian genes. In mammals there are two clusters of histone genes, each encoding a distinct mRNA (with different promoters and 5’ and 3’ UTRs), but coding for the same (H4) or very similar (H1, H2a, H2b, H3) proteins.

Components of histone 3’ processing machinery.

Max Birnstiel’s lab pioneered the early work on histone pre-mRNA processing. Like cleavage/ polyadenylation the formation of the 3’ end requires cleavage of the transcript between two cis-elements; the SL and a downstream sequence termed the histone downstream element (HDE), and specific factors recognize each element (Fig. 1). The initial evidence that the 3’ end of histone mRNAs was formed by cleavage and not by transcription termination was the demonstration by Krieg and Melton that synthetic chicken histone H2b pre-mRNA injected into Xenopus oocytes was cleaved to form a mature histone mRNA(Krieg & Melton, 1984). Subsequently Birnstiel and coworkers took advantage of the fact that the sea urchin histone H2a pre-mRNA was not processed in Xenopus oocytes. They complemented the oocytes with different biochemical fractions from sea urchin lysates(Galli, Hofstetter, Stunnenberg, & Birnstiel, 1983), and subsequently identified U7 snRNA as a component essential for histone pre-mRNA processing. They showed that the 5’ end of U7 snRNA could base-pair with sequences 3’ of the SL and these sequences, termed the histone downstream element (HDE), were required for processing(Georgiev & Birnstiel, 1985; Strub, Galli, Busslinger, & Birnstiel, 1984) (Fig. 1). U7 snRNA, like the spliceosomal snRNAs, has a trimethyl G cap, and was bound to proteins recognized by anti-Sm antibodies(Strub & Birnstiel, 1986) to form the U7 snRNP.

Birnstiel and coworkers subsequently showed that they could process mammalian histone pre-mRNAs in a nuclear extract from mammalian cells(Gick, Krämer, Keller, & Birnstiel, 1986). Mammalian U7 snRNA was subsequently identified by Mowry and Steitz(Mowry & Steitz, 1987), Soldati and Schumperli(Soldati & Schümperli, 1988) and Birnstiel’s group(Cotten, Gick, Vasserot, Schaffner, & Birnstiel, 1988). They identified specific base-pairing interactions between U7 snRNA and the HDE in histone pre-mRNA essential for processing(Cotten et al., 1988). The SL at the cleavage site was shown to interact with a factor distinct from U7 snRNP(Mowry, Oh, & Steitz, 1989; Vasserot, Schaufele, & Birnstiel, 1989). The processing reaction can occur in high concentrations of EDTA, ruling out any role for Mg⁺² in the reaction.

Scharl and Steitz showed that a critical parameter for efficient processing in vitro in mammals is the distance between the SL and the HDE (U7 snRNP binding site). Cleavage occurs 5 nts after the SL in vertebrates, and 4 nts after the SL in invertebrates, with the HDE starting about 10 nts after the SL. Lengthening the distance between the SL and the HDE in a human histone gene moved the position of the cleavage site further 3’ of the SL indicating that in mammals the position of the U7 snRNP determined the cleavage site, although the efficiency of cleavage decreased as the length between the SL and the HDE increased(Cho, Scharl, & Steitz, 1995; Scharl & Steitz, 1994, 1996). Schumperli and coworkers further showed that if the mammalian HDE binds very tightly to U7 snRNP (due to extensive base complementarity with the HDE) then the pre-mRNA could be cleaved in vitro independent of the presence of the stemloop(Spycher et al., 1994; Streit, Koning, Soldati, Melin, & Schümperli, 1993). Together these studies demonstrated that the mammalian U7 snRNP was likely able to bind the cleavage complex, independent of the factor that bound the SL, and that the position of the U7 snRNP on the pre-mRNA and not the position of the stemloop determined the site of cleavage in vitro.

In contrast, we found that in Drosophila the factor that binds the stemloop is required for processing in vitro even if the U7 snRNA base-pairs perfectly with the HDE. The cleavage site is 4 nts from the stemloop, and it is not changed when one moves the HDE further 3’ of the stem-loop(Dominski, Yang, Purdy, & Marzluff, 2005). This suggests that in Drosophila the stem-loop determines the position of the cleavage site, and is a major difference between the Drosophila and vertebrate 3’ processing system.

Identification of SLBP (HBF) a protein that specifically binds the stemloop.

The major challenges to understand how histone pre-mRNAs were processed were to identify the factor that bound the SL, the composition of U7 snRNP, and the nuclease that cleaved the pre-mRNA. U7 snRNP was clearly similar to the spliceosomal snRNPs, but must have distinct components, because the Sm binding site in U7 snRNA differed from the Sm binding site in spliceosomal RNAs, and this difference was essential for assembly of an active U7 snRNP(Grimm, Stefanovic, & Schümperli, 1993). Dani Schumperli showed that changing the U7 Sm-site to the sequence present in spliceosomal mRNA resulted in formation of an snRNP that was inactive in histone pre-mRNA processing(Grimm et al., 1993).

The factor that binds the stemloop was detected by EMSA, using a radioabeled stemloop as a probe. UV-crosslinking of the complex revealed that a protein which migrated at about 50 kDa on SDS-PAGE bound the stemloop and that a similar protein was present both in the nucleus and on polyribosomes in the cytoplasm(Dominski, Sumerel, Hanson, & Marzluff, 1995; Pandey, Sun, & Marzluff, 1991). The EMSA assay allowed us to use mutations in the SL RNA to map the specificity of binding and provide initial evidence mutations in the SL of histone mRNA that affected binding also affected processing in vivo(Pandey et al., 1991; Pandey et al., 1994; Williams & Marzluff, 1995). However we were not successful in purifying the protein. In 1994 when I went to the RNA Processing Meeting at Cold Spring Harbor, the last talk of the first evening was given by Marv Wickens. His abstract was on a paper that had come out just before the meeting, so he talked about a new method he had developed with Stan Fields; the yeast 3-hybrid system, which was designed to identify sequence specific RNA binding proteins. After the session, many of us asked Marv for the reagents for this system. He agreed but wanted to have the paper in press first. Of course the first journal rejected his paper, slowing our progress. When their paper was finally published(SenGupta et al., 1996), Dani Schumperli’s lab and our lab got the reagents and we both rapidly cloned the cDNA for the factor that bound the SL, which we called stemloop binding protein (SLBP) and they called hairpin binding factor (HPF), and showed that this protein was essential for histone pre-mRNA processing in vitro (F. Martin, Schaller, Eglite, Schümperli, & Müller, 1997; Wang, Whitfield, Ingledue, Dominski, & Marzluff, 1996). We also found SLBP was bound to histone mRNA in the histone mRNP (Whitfield et al., 2004). Using the same 3-hybrid approach, we also cloned the Xenopus SLBP and Drosophila SLBP (E. Sullivan et al., 2001; Wang, Ingledue, Dominski, Sanchez, & Marzluff, 1999). From the Xenopus oocyte 2-hybrid library we also cloned a second SLBP, xSLBP2, which has a very similar RNA binding domain but is not active in processing, but functions in storing maternal histone mRNA in the Xenopus oocyte (Wang et al., 1999). A similar protein is found in zebrafish (He et al., 2018) and likely in other vertebrates that store large amounts of histone mRNA in oocytes. Surprisingly the 3-hybrid system has not been successful in identifying many sequence-specific RNA-binding proteins. SLBP has an extremely high affinity for its target, which is also very large (26 nts), and these factors likely accounted for our success.

At about the same time the SLBP cDNA was cloned, Dr. Zbig Dominski joined our lab, initiating a productive collaboration that has continued for 30 years. Zbig led the biochemical studies of histone pre-mRNA processing starting with determination of regions in SLBP required for processing, and culminating in the determination of the components of the processing complex, followed by reconstitution and structure of the active complex in collaboration with Liang Tong’s lab. A 73 amino acid region in the center of vertebrate SLBP was sufficient for binding the stemloop(Wang, Whitfield, et al., 1996). This novel RNA binding domain (RBD) is present only in SLBPs, and RD-histone mRNAs are the only mRNAs bound by SLBP(Brooks et al., 2015). Using nuclear extracts from either mammalian or Drosophila from which SLBP was depleted using biotinylated stemloop RNA, and SLBP proteins expressed in baculovirus, we showed an additional 20 amino acids after the stem loop in mammalian SLBP is required for processing activity, and that one function of SLBP is to stabilize binding of U7 snRNP to histone pre-mRNA(Dominski, Zheng, Sanchez, & Marzluff, 1999). In Drosophila, the RNA binding domain is located near the C-terminus of SLBP and the C-terminal 90 amino acids are sufficient for processing(Dominski et al., 2002). Thus the 70 aa RBD, which includes the SL binding site plus about 20 additional amino acids C-terminal to the stemloop, is necessary and sufficient for processing in vitro in both SLBPs.

Structure of SLBP bound to the 3’ end of histone mRNA.

Using EMSA assays we showed using nuclear extracts that SLBP binds very tightly to a 26 nt RNA containing 5 nts before the 16 nt stemloop and 5 nts after the stemloop, and demonstrated that the 5’ flanking sequences and specific nts in the loop were required for high-affinity binding(Williams & Marzluff, 1995). Further studies by Battle and Doudna using recombinant protein highlighted an important role for the GC base pairs at the base of the stem(Battle & Doudna, 2001). In a subsequent study in collaboration with Howard Chang’s lab we screened a large number of stemloop mutants(L. Martin et al., 2012). These studies showed a critical role for nucleotides 1 and 3 (uridines) in the loop, and the second G nt at the base of the stem, as well as the 2 adenosines conserved at positions 2 and 3 before the stem. In collaboration with Liang Tong, we determined the crystal structure of human SLBP RBD bound to the stemloop. The human SLBP has only been crystallized as a ternary complex containing 3’hExo, which binds the 3’ end of histone mRNA and is responsible for initiating histone mRNA degradation(Tan, Marzluff, Dominski, & Tong, 2013). The histone mRNP likely contains SLBP and 3’hExo bound to the SL. Traci Hall determined the structure of the Drosophila SLBP RBD bound to the stemloop (J. Zhang et al., 2014). The structures were essentially identical (Fig. 3). SLBP binds to the 5’ side of the stem- loop with two long a-helices (A and C) extending from the loop to the base of the stem, including the conserved A’s, 2 and 3 nts before the stem. A third smaller helix B protrudes from the structure near the loop and does not contact the RNA, but is required for processing.

Fig. 3.

Structure of SLBP bound to the SL at the 3’ end of histone mRNA. A. The structure of the RNA binding domain of human SLBP bound to the 3’ end of histone mRNA together with the 3’ to 5’ exonuclease, 3’hExo, which is responsible for initiating degradation of histone mRNA. The SLBP was expressed in baculovirus and T171 is phosphorylated. B. Zoom in on the role of the phosphorylated TPNK. The phosphate makes contacts with all three helices, and all the aa in the loop are resolved. Phospho-T171 makes contacts with multiple amino acids, including K146 in helix A, Y151 in helix B and W190 in helix C, and R160 and R163 in the loop joining helix B to helix C. B. The structure of the SLBP/SL/3’hEXO complex, when SLBP was expressed in bacteria. T171 is not phosphorylated and portions of the loop joining helix B and C is disordered, and the binding affinity for SLBP to the stem-loop is reduced. (Fig. adapted from Tan et al., 2013 and Zhang et al., 2014)

Xenopus SLBP2 bound the histone 3’ end with similar affinity but did not support processing. We were able to determine specific sequences in the RNA binding domain that were required for processing but not for RNA binding. We identified one region, YDRY in SLBP1 and YQCY in SLBP2 present in helix B, where changing two amino acids in SLBP1 to the amino acids in SLBP2 inactivated processing but not RNA binding (Dominski, Erkmann, Greenland, & Marzluff, 2001). The YDRY does not directly contact RNA (Fig. 3), but likely contacts another factor required for processing. Note that the C-terminal 20 aa required for processing starts immediately after helix C and the C-terminal end of helix C is located close to helix B. Thus helix B and the C-terminal extension likely interact with a factor(s) required for processing.

Phosphorylation of SLBP is required high-affinity binding of SLBP to the SL RNA.

The 73 aa RNA binding domain of Drosophila SLBP is very similar to the vertebrate SLBP, although Drosophila SLBP does not support processing in mammalian cell extracts and vice-versa, likely due to the large differences in the 20 aa’s after the RBD (Dominski, Yang, Purdy, et al., 2005). Top-down proteomics showed that both mammalian and Drosophila SLBP are phosphorylated on the TPNK in the RNA binding domain, and that all the molecules of SLBP are phosphorylated on this residue in both the human and Drosophila protein expressed in baculovirus (Borchers et al., 2006). Liang Tong also determined the crystal structure of the phosphorylated human SLBP(J. Zhang et al., 2014). In the unphosphorylated protein expressed in E.coli, there is a region which is disordered in the linker region between helices B and C, and the 20 aa after the RBD required for processing is also disordered (Fig. 3A). In the phosphorylated SLBP all the residues in the RBD are resolved as a result of the phosphothreonine forming salt bridges with basic amino acids in helices A and C, stabilizing the structure, but the SLBP C-terminal extension is still disordered (Fig. 3). The phosphorylated SLBP binds more tightly than the unphosphorylated protein, as evidenced by its very slow off-rate, with the complex remaining stable for several hours in vitro in the presence of a large excess of competitor RNA (Borchers et al., 2006). Changing the T to A in Drosophila TPNK is lethal suggesting that the phosphorylation of the TPNK may be essential for proper folding of the Drosophila SLBP in vivo (Lanzotti et al., 2004).

Identification of proteins in the U7 snRNP.

While it was clear the U7 snRNP contained some Sm proteins found in spliceosomal snRNPs, there is a significant difference in the sequence of the Sm-site which binds the ring of Sm-proteins suggesting there might be a difference in the Sm proteins that associate with U7 snRNA. U7 snRNP is present in very low abundance relative to the spliceosomal RNPs, approximately 1% of the levels of the spliceosomal snRNPs, which provided a challenge for biochemical studies of the U7 snRNP. Ramesh Pillai, when he was a graduate student in Dani Schumperli’s lab, identified two proteins they named Lsm10 and Lsm11 that replace SmD1 and SmD2 in the mammalian U7 snRNP(Pillai et al., 2003; Pillai, Will, Lührmann, Schümperli, & Müller, 2001). Lsm11 is the largest of the proteins in the Sm ring, with an extended N-terminus that plays a critical role in histone pre-mRNA processing (see below).

The composition of the U7 snRNP is conserved between mammals and Drosophila. The Drosophila U7 snRNA was identified by Zbig Dominski (Dominski, Yang, Purdy, & Marzluff, 2003) and the Drosophila orthologues of human Lsm10 and Lsm11 were identified by Azzouz and Schumperli (Azzouz & Schumperli, 2003). Together with the cloning of Drosophila SLBP, these studies opened the way for genetic and biochemical studies of histone pre-mRNA processing in Drosophila.

Components of the cleavage complex.

Once the basic components required for definition of the cleavage site were identified, our focus shifted to the identification of all the factors required for cleavage, and particularly the nuclease that carried out the cleavage. Early studies had suggested that processing was regulated during the mammalian cell cycle and Birnstiel and Schumperli’s groups each provided evidence from in vitro processing experiments, as well as in vivo in a temperature sensitive mammalian cell mutant(Stauber & Schümperli, 1988). that there was a “heat-labile” regulatory factor required for processing(Gick, Krämer, Vasserot, & Birnstiel, 1987; Lüscher & Schümperli, 1987). The first direct evidence identifying other factors involved in cleavage, came from Zbig Dominski who carried out UV-crosslinking experiments in nuclear extracts using a histone pre-mRNA substrate with a radiolabeled adenosine thiophosphate at the cleavage site using the method of Moore and Sharp (Moore & Sharp, 1992). A single protein of about 70 kDa crosslinked to the substrate and Zbig identified it by antibody precipitation as CPSF73, strongly suggesting that CPSF73 was directly bound to the cleavage site(Dominski, Yang, & Marzluff, 2005), and was the enzyme that cleaved the pre-mRNA, a role it also plays in cleavage/polyadenylation. Shortly afterward Kolev and Steitz showed that the heat-labile factor required for in vitro processing, previously described by Gick and Birnstiel(Gick et al., 1987), was a second polyadenylation factor, symplekin(Kolev & Steitz, 2005), which is present in a complex with CPSF73/CPSF100 in both mammals and Drosophila (Kolev, Yario, Benson, & Steitz, 2008; K. D. Sullivan, Steiniger, & Marzluff, 2009).

To identify additional factors potentially required for processing in vivo, Eric Wagner performed a genome-wide in vivo RNAi screen for processing factors in Drosophila. The screen identified a number of factors known to be involved in processing; SLBP, Lsm10, Lsm11, CPSF73 and symplekin, as well as CPSF100, as well as a number of potential in vivo regulatory factors(Wagner et al., 2007). A critical role for CPSF73, CPSF100 and symplekin in histone pre-mRNA processing was further confirmed by Kolev and Steitz in mammalian cells by expression of a fusion protein containing all 3 polypeptides, as well as a fusion protein with specific mutations in the active site of CPSF73 or CPSF100. The wild-type protein but not the protein with mutations in either CPSF73 or CPSF100 could restore processing histone pre-mRNA processing in a heat-treated extract, suggesting that a complex of these three proteins was required to assemble the cleavage activity(Kolev et al., 2008). We obtained a similar result in Drosophila, where Sullivan and coworkers showed that symplekin, CPSF73 and CPSF100 formed a complex which was present at the 3’ end of histone genes (presumably bound to nascent histone mRNAs) in Drosophila cells by CHIP(K. D. Sullivan et al., 2009). These two sets of data suggested a core cleavage complex shared by histone pre-mRNA processing and cleavage polyadenylation of polyadenylated mRNAs.

How are the polyadenylation factors required for histone pre-mRNA processing recruited to histone pre-mRNA?

We hypothesized that the U7 snRNP was required to recruit the cleavage factors to the histone pre-mRNA, becuse cleavage of histone pre-mRNA in vitro in mammalian cell extracts could occur in the absence of SLBP if there was extensive base-pairing between the HDE and U7 snRNA, either in natural substrates or synthetic substrates (Streit et al., 1993). Mammalian Lsm11 is the largest Sm protein in the ring, and unique to U7 snRNP. There is a long N-terminal extension of ~180 aa in all metazoan Lsm11 proteins, as well as loop of variable size between the Sm1 and Sm2 domains. We used the N-terminal domain of Lsm11 as a bait in a yeast 2-hybrid screen. The major positive hit in the 2-hybrid screen was a fragment which contained the first 200 amino acids of FLASH(X.C. Yang, Burch, Yan, Marzluff, & Dominski, 2009). FLASH is a large >2000 aa protein, which previously had previously been reported to colocalize with NPAT in the nucleus (and hence was concentrated at the histone genes)(Barcaroli, Bongiorno-Borbone, et al., 2006; Barcaroli, Dinsdale, et al., 2006) and was necessary for histone gene expression. Zbig mapped the binding site on Lsm11 to a highly conserved region around amino acid 50, which bound to aa 80-120 of human FLASH (Fig. 4B). Many nuclear extracts we prepared from HeLa cells had low histone pre-mRNA processing activity. Addition of as little as 5 ng of the N-terminus of FLASH to inactive nuclear extract greatly increased activity (Fig. 4C). Using this assay Zbig showed that a conserved sequence in FLASH adjacent to the Lsm11 binding site was essential for processing activity. He identified a homologue of FLASH in Drosophila which had a similar N-terminal sequence, and showed it was also required for histone pre-mRNA processing in Drosophila cultured cells using a histone pre-mRNA processing reporter(Burch et al., 2011; X.C. Yang, Burch, et al., 2009).

Fig. 4.

FLASH stimulates histone pre-mRNA processing in vitro. A. U7 snRNA forms a complex with a different set of Sm proteins than the spliceosomal snRNAs. The Sm site in U7 snRNA differs from the spliceosomal Sm site. B. Schematic of binding of FLASH to Lsm11. A conserved LDLY in FLASH is required for processing but not for binding to Lsm11. C. FLASH is limiting for processing in nuclear extracts. A labeled histone pre-mRNA (lane 1) shows a low processing a Hela cell nuclear extract. Addition of the indicated amounts of aa 29-139 of FLASH (lanes 2-4) stimulates processing in vitro. Fig. Adapted from Yang et al., 2009.

The FLASH/Lsm11 complex binds the cleavage factor.

To determine whether FLASH directly recruited the cleavage complex to the U7 snRNP, Zbig formed a heterodimer between the MBP-tagged Lsm11 and GST-tagged FLASH N-termini and tested whether this heterodimer bound any polyadenylation factors suspected to be involved in histone pre-mRNA processing from nuclear extracts. Both Symplekin and CPSF73 were bound to the heterodimer as analyzed by Western blotting (Fig. 5). He then used a biotinylated antisense 2’OCH3 RNA oligonucleotide to pull down the U7 snRNP from mammalian nuclear extracts, after the addition of the N-terminus of FLASH to the extract. Analysis of these complexes by Western blotting and proteomics showed that symplekin, CPSF73, CstF64 and CPSF100 were pulled down by U7 snRNP bound to FLASH, but that other polyadenylation factors tested were not enriched(Sabath et al., 2013; X.C. Yang et al., 2013).

Fig. 5.

The FLASH/Lsm11 Complex binds to a subset of polyadenylation factors including CPSF100-CPSF73-Symplekin as part of the U7 snRNP.

A. The N-terminus of FLASH and an inactive mutant of FLASH with the LDLY changed to AAAA fused to GST and the N-terminus of Lsm11 fused to MBP were expressed in bacteria. The purified proteins were mixed to form FLASH/Lsm11 complexes and incubated in HeLa cell nuclear extract. The GST-bound proteins were purified and analyzed by Western blotting for CPSF73, CPSF100 and symplekin. Lane 1: 40% of the input. Lane 2: proteins bound to GST. Lane 3: proteins bound to FLASH/Lsm11. Lane 4: proteins bound to the FLASH LDLY mutant bound to Lsm11.

B. A antisense oligonucleotide containing a 3’ biotin complementary to the 5’ end of U7 snRNA was incubated in a nuclear extract from mouse myeloma cells and the bound proteins analyzed by Western blotting for the indicated polyadenylation factors. Lane 1: 1% of the input; Lane 2: proteins bound the anti-sense oligo/U7 snRNP complex. Lane 3: proteins bound to a control biotinylated oligonucleotide. Note that >95% of Lsm11 (endogenous U7 snRNP) was bound but less than 1% of the total polyadenylation factors were bound.. Adapted from Yang et al. 2013.

Essentially all the Lsm11 in the extract was pulled-down by the anti-sense U7 snRNA but only a small fraction (1-2%) of the polyadenylation factors were pulled down, consistent with the low amounts of U7 snRNP present in the cell. In contrast the synthetic Lsm11/FLASH heterodimer was able to pull down much larger amounts of the same polyadenylation factors, demonstrating that the Lsm11/FLASH heterodimer alone is very effective at recruiting the polyadenylation factors(X.C. Yang et al., 2013). These results suggested that the polyadenylation factors required for cleavage bind directly to U7 snRNP together with FLASH.

Jim Manley had shown previously that CstF64 bound to symplekin(Takagaki & Manley, 2000) and that symplekin binding to CstF64 prevented binding of CstF64 to the CstF complex involved in polyadenylation. Dani Schumperli had provided evidence that the CstF64/symplekin complex was involved in histone pre-mRNA processing in vivo in mammalian cells (Romeo, Griesbach, & Schumperli, 2014; Ruepp, Schweingruber, Kleinschmidt, & Schumperli, 2011), suggesting that CstF64 was also a component of the cleavage complex. Thus it is likely that one molecule of CstF64, is also present in the mCF complex and interacts with symplekin. Two molecules of CstF64 are also present in the dimer of the CstF 50/64/77 trimer.

To identify all the components of the U7 snRNP, Zbig and Aleksandra Skrajna purified the processing complex from both mammalian and Drosophila extracts using biotin-labeled substrates which were resistant to cleavage by introducing 2’OCH3 groups at 5 nts around the cleavage site. The biotin could be attached to the 5’ end of the substrate by a photocleavable linker This one-step method allows extensive washing of the bound U7 snRNP followed by recovery of the complex by cleavage of the native U7 snRNP from the beads with the UV light, without releasing material non-specifically bound to the beads. The purified complexes wer analyzed by quantitative mass spectrometry (Skrajna, Yang, Dadlez, Marzluff, & Dominski, 2018), identifying all the components of the U7 snRNP required for activity in both human and Drosophila cells.

Structure and function of the processing complex.

Liang Tong’s lab expressed the 7 Sm proteins, including Lsm10 and Lsm11, allowing Zbig to assemble a complex from synthetic U7 snRNA, the Sm proteins and the N-terminus of FLASH. CPSF has recently been found to contain two subcomplexes mPSF, that binds the AAUAAA, containing CPSF160, CPSF30 and WDR33(Sun et al., 2018) ; and mCF, containing CPSF73, CPSF100 and symplekin (Y. Zhang, Sun, Shi, Walz, & Tong, 2020). When Liang expressed CstF64, CPSF73, CPSF100 and symplekin, they formed a structure similar to the mCF complex. This complex was incubated with the U7 snRNP/FLASH complex, and the complete processing complex purified by gel filtration. Strikingly the initial complex they assembled showed no activity in vitro (Fig. 7C, lane 1). Not all the proteins in the complex were full-length and in particular the N-terminal 300 amino acids of symplekin had been removed to enhance protein expression from the MultiBAC system. This N-terminal domain from humans (Xiang et al., 2010) and Drosophila (Kennedy et al., 2009) had previously been crystallized and the structure determined. Steiniger and coworkers had provided evidence that in Drosophila the N-terminal domain of symplekin played a role in histone pre-mRNA processing in vivo (Michalski & Steiniger, 2015). When Zbig added this fragment in trans to the inactive complex and substrate, it restored robust processing activity (Fig. 7C, lane 3), although it did not form a stable complex with the other processing factors in the absence of substrate RNA. Liang then expressed the complex containing the full-length symplekin which was active and determined the structure of the complex with the substrate RNA by cryo-EM (Fig. 7). A detailed review of this structure and the mechanism of cleavage has recently been published by Dominski and Tong(Dominski & Tong, 2021) and won’t be discussed here in detail. A major unexpected finding was that the N-terminal region of symplekin was bound to the U7 snRNA/HDE duplex (Fig.7, red arrow). The C-terminal region of symplekin was bound directly to the conserved region of FLASH required for processing (Fig. 7, black arrow), although the interaction was not resolved well enough to identify the precise amino acids on symplekin which interacted with FLASH. The detailed interactions between the SLBP/SL complex and the Lsm11/FLASH complex were also not resolved (Fig. 7, green arrow), although specific residues in helix B of SLBP are essential for processing as is the 20 aa region C-terminal of the RNA-binding domain (Dominski et al., 2001). SLBP also is required for stable binding of U7 snRNP to the histone pre-mRNA (Dominski et al., 1999; Skrajna et al., 2017). Despite the well-characterized binding of FLASH and Lsm11 determined by biochemical studies(Aik et al., 2017; Skrajna et al., 2017; Skrajna et al., 2016), the N-terminal region of Lsm11, which binds to the coiled-coil in FLASH, was not resolved in the structure, nor was any part of CstF64, which was present in the complex.

Fig. 7.

Structure of histone pre-mRNA processing complex.

The structure of the histone pre-mRNA processing complex containing full-length symplekin bound to substrate RNA (orange) was solved by cryo-EM. CstF64 was present in the complex but was not detectable in the structure. The N-terminus of the FLASH dimer interacts with the C-terminal domain of symplekin (black arrow) region and the FLASH dimer extends to the SLBP/SL complex (green arrow). The N-terminus of Lsm11 interacts with the FLASH dimer, although it is not visible in the structure. The N-terminal domain of symplekin interacts with the HDE/U7 snRNA duplex (red arrow), helping position the histone pre-mRNA into the active site of CPSF73. Fig. adapted from Sun et al., 2020.

Structure-function studies of U7 snRNP: The U7 snRNA Sm site is required for cleavage.

To probe the requirements for cleavage in U7 snRNA or the Sm proteins, Zbig developed a method to assemble a semi-recombinant U7 snRNP, using s ynthetic U7 snRNA, recombinant Sm proteins and the N-terminus of FLASH, allowing assembly of U7 snRNP with altered biotinylated U7 snRNAs or Sm proteins. The core complex was incubated with nuclear extract and the active U7 snRNP purified by binding to streptavidin beads, and eluted from streptavidin agarose with UV light. The complex was active in histone pre-mRNA processing and the activity was stimulated by recombinant SLBP (Fig. 8). This system allowed purification of large amounts of “semi-recombinant” U7 snRNP, and allowed making mutations in the U7 snRNA or Sm proteins in the semi-recombinant complex, to determine the effect on processing in vitro.

Fig. 8.

A. Preparation of an active semi-recombinant U7 snRNP. A. The core U7 snRNP was assembled from a biotinylated U7 snRNA and recombinant Sm proteins, and the N-terminus of FLASH added. The core U7 snRNP was incubated in a nuclear extract, bound to streptavidin agarose, eluted with UV-light, and then analyzed by gel electrophoresis and proteomics (B) and assayed for activity (C). The activity was inhibited by anti-sense U7 oligonucleotide and stimulated by SLBP. Fig. adapted from Bucholc et al. (2020).

By changing the core of the HDE in the substrate pre-mRNA from AAAGAG to UUUCUC and the binding site for the HDE in U7 snRNA from CUCUUU to GAGAAA, processing of the synthetic histone pre-mRNA could be studied in the presence of any wild-type U7 snRNP present in the nuclear extract. In vitro the 7 Sm/Lsm U7 snRNP proteins provided in excess assembled on synthetic U7 snRNA with the spliceosomal Sm site or with point mutants in the U7 Sm site, and the core complex still recruited the cleavage complex. Both Sm sites start with AAUUU but the spliceosomal site ends in G/U U U/G G/A and the U7 Sm site ends in GUCU/C A/U G and it is 11 nts long rather than 9 nts. The conserved C at position 8 and G at position 11 form a base pair in the cleavage complex (Fig. 9).

Fig. 9.

A. The role of the Sm site. The Sm site of U7 snRNA is structured in the processing complex, with a base pair between the conserved C at position 8 and the conserved G at position 12. B. We assembled a U7 snRNP with either a mutant preventing formation of the C-G base pair or with consensus splicesosomal Sm site. These mutant U7 snRNPs recruit the processing complex when incubated in nuclear extract C. The purified U7 snRNPs were incubated with the H2a pre-mRNA substrate. Disrupting the C-G base pair greatly reduced activity, while the complex formed on the spliceosomal Sm site had very little activity, and the small amount of cleavage occurred 2-3 nts after the normal cleavage site (arrow). Fig. adapted from Bucholc et al., 2020.

The complex assembled on the spliceosomal Sm site had minimal activity (<5% of the U7 Sm site) and cleaved primarily 2 nts after the normal cleavage site; Fig. 7D, lane 8) and a single point mutant in the invariant C to a G in the U7 Sm had ~ 10-fold reduced activity (Fig. 7D, lanes 3 and 7). These mutant U7 snRNPs formed a stable complex with the FLASH N-terminus. Thus proper binding of Lsm 10 and Lsm11 to U7 snRNA must help form the active conformation of the cleavage complex, which allows the opening of CPSF73 RNA binding site to allow substrate binding.

Scharl and Steitz (1996) had shown that insertion of 4-12 nts into histone pre-mRNA between the SL and the HDE moved the cleavage site an equivalent further 3’ of the stem-loop in vitro, demonstrating that the position of the HDE (the U7 snRNA binding site), determined the site of cleavage. Using the semi-recombinant U7 snRNP, Bucholc and coworkers also showed that insertion of 4 nts into the U7 snRNA, base-pairing of those nts with the H2a+4 substrate, restored the cleavage to the normal site. In U7 snRNA there is only 3 nts between the end of the UUUCUC that base pairs with the AAAGAG in the HDE and the Sm site. Deletion of the 3 nts between the HDE and the Sm site shifted the cleavage site closer to the SL(Bucholc et al., 2020). Thus in mammalian cells U7 snRNP is the molecular ruler proposed by Scharl and Steitz(Scharl & Steitz, 1996) that determines the cleavage site based on the position of the HDE. Note that in Drosophila, cleavage requires SLBP and moving the position of the HDE has no effect on the cleavage site but just reduced activity, a major difference between the vertebrate and insect histone genes (Dominski, Yang, Purdy, et al., 2005; Skrajna et al., 2017).

U7 snRNP degrades the 3’ cleavage product: CPSF73 is also a 5’ to 3’ exonuclease.

After cleavage to form the 3’ end of histone mRNA, the U7 snRNP remains bound to the downstream cleavage product. It is likely that removal of the U7 snRNP is necessary to rapidly recycle U7 snRNP. The existence of a U7 snRNP dependent 5’ to 3’ exonuclease associated with histone pre-mRNA processing was first suggested by Walther and Schumperli(T. N. Walther, Wittop, Schümperli, & Muller, 1998). When we added the synthetic cleavage product to the nuclear extract, the cleavage product was degraded in a U7 snRNP dependent manner and CPSF73 was also crosslinked to the 3’ cleavage fragment in a U7 snRNP-dependent manner(Dominski, Yang, & Marzluff, 2005). Subsequent biochemical experiments defined the relationship between the distance between the 5’ end of the RNA and the HDE required to initiate 5’ to 3’ exonuclease digestion, rather than endonucleolytic cleavage in the nuclear extract(X.C. Yang, Sullivan, Marzluff, & Dominski, 2009). A striking feature of the exonuclease activity is that it is processive and while it requires U7 snRNP to initiate degradation, degradation continues through the U7 snRNP binding site. Degradation of the downstream fragment was blocked by 2’OCH3 nts inserted into the 3’ fragment in vitro, and 2’OCH3 nucleotides at the cleavage site of the pre-mRNA also inhibit processing in vitro.

All these studies were performed in nuclear extracts. Proof that the exonuclease was actually CPSF73 and not some other associated exonuclease, was provided by study of the activity of the recombinant processing complex. The 3’ fragment was degraded in a U7 snRNP dependent manner and required active CPSF73 (X. C. Yang et al., 2020). Using a processing substrate with different fluorescent labels on the 5’ and 3’ end, Sun et al. demonstrated both endonucleolytic cleavage to give the proper stable product and subsequent exonucleolytic degradation of the 3’ fragment, with the exonuclease activity resulting in a 3’ labeled mononucleotide or a very short fragment (Sun et al., 2021) (Fig. 11). Degradation goes through the HDE and can continue at least 15 nts further to completely degrade the cleavage product. U7 snRNP must be displaced from the substrate as the HDE is degraded, and because there is no accumulation of degradation intermediates, the 5’ to 3’ degradation by CPSF 73 is processive Thus in addition to endonucleolytic cleavage of histone pre-mRNA, CPSF73 subsequently acts as a 5’ to 3’ exonuclease to degrade the 3’ fragment in a U7 snRNP dependent reaction(Sun et al., 2021). This reaction is likely essential to rapidly recycle U7 snRNP and it may also play a role in efficient transcription termination.

Fig. 11.

Reducing the concentration or activity of FLASH in the HLB results in some read-through transcription. A. Diagram of FLASH mutant proteins expressed to complement a mutant in Drosophila FLASH. B. Analysis of the 3’ end of histone H2a mRNA by S1 nuclease mapping, which resolved properly processed, and misprocessed polyadenylated histone mRNAs and unprocessed, read-through RNAs (RT). Fig. adapted from Tatomer et al., 2016.

A link between 3’ processing and transcription termination.

Following cleavage and polyadenylation of mRNAs, transcription is terminated at a variable distance (often several kB) 3’ of the cleavage site over a range of nts(Eaton et al., 2018; Eaton & West, 2020). Failure to process the pre-mRNA results in continued transcription for much longer distances. Termination of transcription from genes encoding polyadenylated mRNA requires degradation of the RNA downstream of the cleavage site by the processive 5’ to 3’ exonuclease Xrn2 and it is thought that termination ultimately results from Xrn2 “catching up” with the RNA polymerase, and acting as a “torpedo” to induce termination(Eaton et al., 2018; Eaton & West, 2020). Termination also requires that the polymerase slow its elongation rate, both to allow efficient processing as well as termination. The elongation rate is slowed by the action of PNUTS, a protein phosphatase that dephosphorylates the elongation factor Spt5(Cortazar et al., 2019) and the combination of the slowed transcription rate, with cleavage and degradation of the 3’ cleavage product 5’ to 3’ by Xrn2 results in termination. Removal of Xrn2 results in extensive continued transcription after cleavage(Eaton et al., 2018) as does failure to dephosphorylate Spt5(Cortazar et al., 2019).

In contrast to transcripts that produce polyadenylated mRNAs, histone transcripts (and snRNA transcripts) terminate close to the processing site (<200 nts) and termination is not affected by loss of Xrn2(Eaton et al., 2018), but does require PNUTS and slowing of the RNA polymerase(Cortazar et al., 2019). This result is consistent with the possibility that the 5’ to 3’ exonuclease activity of CPSF73 promotes termination of transcription on histone genes by degrading the 3’ fragment product in vivo. Note that after cleavage of the pre-mRNA prior to polyadenylation, the mCF complex containing CPSF73 likely remains bound near the polyadenylation signal 5’ of the polyadenylation site, and is not in a position to carry out 5’ to 3’ degradation of the downstream fragment. mCF is associated with U7 snRNP is positioned 3’ of the cleavage site in histone mRNA allowing it to readily access the 5’ end of the cleaved transcript.

Does histone pre-mRNA processing have to be co-transcriptional?

Transcription from histone genes normally terminates shortly after the stem loop, suggesting that histone pre-mRNA processing is likely cotranscriptional under normal conditions in vivo o, with processing occurring shortly after transcription of the HDE. However this does not have to be an obligatory coupling. In extracts from Drosophila cells, Adamson and Price showed that processing by cleavage/polyadenylation occurs much more efficiently on a nascent in vitro transcript than on an RNA substrate, and this efficient processing is dependent on the CTD on RNA pol II(T.E. Adamson, Shutt, & Price, 2005), consistent with findings that the CTD interacts directly with some polyadenylation factors and that the pol II CTD stimulates processing in vitro(McCracken et al., 1997; Ryan, Murthy, Kaneko, & Manley, 2002). Adamson and Price carried out a similar study of histone pre-mRNA processing in a Drosophila nuclear extract capable of both transcription and processing of histone pre-mRNA. They found that the rate of processing was similar on both synthetic pre-mRNA and RNA formed during transcription. There is no evidence that any of the histone processing factors interact directly with the CTD. However they did find that there was strong transcription pause site 3’ of the histone genes shortly after the HDE, suggesting that this pausing may promote transcription-coupled processing(T. E. Adamson & Price, 2003), which then could be followed by transcription termination as a result of 5’ to 3’ degradation of the 3’ cleavage product by CPSF73 in vivo.

It is also possible that histone pre-mRNAs could be processed after transcription continues past the termination site. Our in vivo experiments in Drosophila are consistent with this possibility. In each Drosophila histone gene, there is a polyadenylation site after the histone processing site and before 3’ end of the next histone gene (which are about 300 nts apart in the tandem repeat (Fig. 2). If the processing factors, SLBP or U7 snRNP, are mutated or present at very low concentrations, the histone mRNAs are not processed at the normal site, but polyadenylated histone mRNAs are formed (Lanzotti, Kaygun, Yang, Duronio, & Marzluff, 2002; E. Sullivan et al., 2001). These polyadenylated histone mRNAs support cell growth in cultured cells, which actually grow faster when SLBP or U7 snRNP are knocked down by RNAi to very low levels.

Regulation of Expression Levels of histone mRNAs: role of 3’ processing.

In mammalian cells there are multiple (10-15) copies of the genes for each histone protein. These copies encode the same protein but have different 5’ and 3’ UTRs and 5’ and 3’ flanking sequences encoding the promoter and the HDE. The levels of each transcript differs widely among the different gene copies for each of the histones(Brooks et al., 2015). All these histone mRNAs are expressed primarily in S-phase with very low levels of histone mRNA present in cycling cells in G1- or G2-phase cells. The relative amounts of each histone mRNA present in an S-phase cell depends on both the strength of the promoter 5’ of the genes and efficiency of the processing of the transcripts. Histone transcripts that are not processed are likely rapidly degraded.

We showed that the efficiency of 3’ processing plays a role in the level of expression of the different histone gene copies. Each of the rodent H3 histone genes have almost identical nucleotide sequences in the coding region(Debry & Marzluff, 1994), allowing us to make a series of chimeric genes, containing the promoter from one gene and the 3’ processing signal of another gene taking advantage of common restriction sites in the coding region. The nucleotide sequence of the coding region was identical for each histone H3 gene, allowing us to measure the amount of total H3 mRNA, as well the amount of individual H3 mRNAs using S1 nuclease mapping(Graves, Wellman, Chiu, & Marzluff, 1985). The results were striking; changing either the 5’ UTR or the 3’ UTR on one histone gene for the 5’ or 3’ UTR of another histone gene resulted in a change in the expression of the mRNA, with the total change resulting from changes in both the transcription rate (5’ flanking sequence) and processing efficiency (3’ flanking sequence)(Levine, Liu, Marzluff, & Skoultchi, 1988; T.-J. Liu, Levine, Skoultchi, & Marzluff, 1989). The activity of the 3’ flanking region in promoting expression correlates with the ability of the HDE to interact with U7 snRNA, consistent with rapid cotranscriptional 3’ processing. If processing does not occur rapidly, then the pre-mRNA is likely rapidly degraded.

Thus the level of expression of an individual histone gene is determined by the strength of its promoter as well as the efficiency of 3’ processing, and these steps are independent of each other. If a gene had a promoter that was 4 times stronger, and a 3’ end 5 times more efficient in processing, the steady-state level of the mRNAs produced differed by 20-fold(Levine et al., 1988).

Regulation of 3’ processing during the mammalian cell cycle.

In mammalian cycling cells histone mRNA levels increase rapidly after starting just before entry in S-phase, peaking up to 35-fold higher than in G1, and decrease rapidly at the end of S-phase(Harris et al., 1991). Studies from Schumperli showed that most of the regulation of histone mRNA was due to elements at the 3’ end of the histone genes, including the processing signals(Lüscher & Schümperli, 1987; Lüscher, Stauber, Schindler, & Schümperli, 1985; Stauber, Lüscher, Eckner, Lotscher, & Schümperli, 1986; Stauber & Schümperli, 1988), and suggested that the heat-labile factor described by Gick and Birnstiel was involved in cell cycle regulation of expression(Lüscher & Schümperli, 1987). To determine the relative importance of changes in histone gene transcription compared with posttranscriptional regulation, we constructed a histone gene driven by the constitutive U1 snRNA promoter. The levels of this mRNA increased 10 fold as cells entered S-phase compared to the total of 35 fold for the same gene driven by the histone promoter, suggesting that a portion of the increase was due to histone gene transcription and most of the increase in histone mRNA was due to increased efficiency of 3’ processing(Harris et al., 1991). In contrast the rapid decrease in histone mRNA at the end of S-phase can only be accomplished by rapid degradation of histone mRNAs reducing the levels of histone mRNA to the levels in G1 cells before mitosis(Harris et al., 1991), and also stop biosynthesis of new histone mRNA. Both processing and degradation of histone mRNA require the stemloop, so the stemloop in histone mRNA is the most important cis element for histone gene regulation during the mammalian cell cycle(Pandey & Marzluff, 1987).

How is processing regulated during the cell cycle?

Using both cells released into the cell cycle after mitotic shake off, or synchronized by double thymidine block, Mike Whitfield found that SLBP protein levels are cell-cycle regulated in mammalian cells. There are low levels of SLBP immediately after mitosis and during G1, and SLBP levels increased rapidly as cells approached and entered S-phase(Whitfield et al., 2000). At the end of S-phase SLBP was rapidly degraded as a result of phosphorylation of SLBP by cyclin A/Cdk1(Koseoglu, Graves, & Marzluff, 2008), and this rapid degradation prevents further histone pre-mRNA processing(Koseoglu et al., 2008; Whitfield et al., 2000; Zheng et al., 2003). Note the histone mRNA is also rapidly degraded at the end of S-phase, even if the SLBP is not degraded. This suggests that histone mRNA is degraded to release SLBP, which then returns to the nucleus where it can be phosphorylated and degraded following activation of cyclin A/cdk1 as cells enter G2. When SLBP is stabilized by mutating the phosphorylation sites, histone mRNAs are still rapidly degraded at the end of S-phase(Whitfield et al., 2000).

SLBP is bound to histone mRNA throughout its entire life cycle. As a result, the maximal levels of total histone mRNA in the cell cannot be higher than the amount of SLBP. In double-thymidine blocked cells, which are blocked just after the initiation of S-phase, SLBP is present at maximal levels, consistent with the levels of SLBP rising just prior to the start of S-phase, allowing rapid histone mRNA and protein synthesis at the start of S-phase(Whitfield et al., 2000). SLBP mRNA levels only increase 2-fold during the cell cycle in thymidine blocked cells or cells released from mitosis, suggesting that SLBP protein levels are also regulated posttranscriptionally. SLBP levels may be regulated by regulating translation(Whitfield et al., 2000) or by continued degradation of SLBP during G1(Dankert, Pagan, Starostina, Kipreos, & Pagano, 2017) or a combination of the two.

A second mechanism for regulating histone pre-mRNA processing during the cell cycle could be by directly regulating the processing reaction, possibly by the heat-labile factor described by Gick and Birnstiel(Gick et al., 1987). Kolev and Steitz identified the heat-labile factor as symplekin, which we now know is a critical component for assembly of active U7 snRNP(Kolev & Steitz, 2005). As described below, one possibility is that assembly of the active form of U7 snRNP, containing mCF, may only occur during S-phase.

Histone mRNAs are synthesized in a nuclear body, the Histone Locus Body (HLB), which forms on the histone genes.

The replication-dependent histone genes in animal cells are linked in the genome and are present in large clusters containing multiple copies of genes for each of the 5 histone proteins (Fig. 2). In mammalian cells all but four of the RD histone genes, are present in two clusters, a large cluster of >50 genes on mouse chromosome 13 (human chromosome 6) and a small one (~10 genes) on mouse chromosome 3 (human chromosome 1), which are conserved in mammals (Albig & Doenecke, 1997; Seal et al., 2022; Wang, Krasikov, et al., 1996; Wang, Tisovec, et al., 1996) U7 snRNA was shown to be present near the histone genes by in situ hybridization by Frey and Matera(Frey & Matera, 1995), in what were initially thought to be a subset of Cajal bodies. Subsequently, NPAT was identified by Ed Harlow’s lab as a Cyclin E/cdk2 substrate that was required for entry into S-phase(Zhao, Dynlacht, Imai, Hori, & Harlow, 1998) and NPAT was localized to the two mammalian histone gene clusters(Ma et al., 2000; Zhao et al., 2000). In Drosophila there are 100 copies of a 5 kB tandem repeat containing all five histone genes. In 2006, Joe Gall and coworkers showed that Drosophila U7 snRNP was present in a nuclear body at the histone genes, which he named the Histone Locus Body (HLB) (J. L. Liu et al., 2006).

There is a large HLB on the human chromosome 6 cluster and a smaller HLB on the human chromosome 1 cluster, consistent with the number of genes in each cluster (Fig. 2). In the same year, Barcaroli et al. identified FLASH, a large protein of unknown function, that was present in the NPAT containing Cajal bodies in mammalian cells(Barcaroli, Bongiorno-Borbone, et al., 2006) and was necessary for histone gene expression. Subsequently we discovered that FLASH was an essential component of the histone pre-mRNA processing complex in both mammals and Drosophila (X.C. Yang, Burch, et al., 2009), and directly binds to NPAT. Together with Bob Duronio, we determined that the known homeotic gene, multi-sex combs (Mxc), was the Drosophila orthologue of NPAT, and that Mxc and FLASH were critical proteins required for assembly of the HLB in Drosophila (Terzo et al., 2015; White et al., 2011).

Regulation of histone mRNA biosynthesis within the HLB.

The HLB assembles at the histone genes immediately after mitosis, and in Drosophila and cycling mammalian cells the HLB is present throughout the cell cycle(Armstrong, Passanisi, Ashraf, & Spencer, 2023; Hur et al., 2020). Histone gene expression is activated by phosphorylation of NPAT/Mxc by cyclin E/cdk2 as cells commit to entering S-phase(Armstrong et al., 2023; Ye, Wei, Nalepa, & Harper, 2003; Zhao et al., 2000). NPAT, FLASH and U7 snRNP are all concentrated in the HLB. Note that the components of nuclear bodies readily exchange between the body and nucleoplasm, so while the factors are concentrated in the HLB, a substantial fraction is likely present dispersed through the nucleoplasm but not detectable by IF(Dundr, 2012). Using superresolution microscopy on Drosophila embryos, Bob Duronio’s lab showed that the active HLB has a core/shell structure with the N-terminus of Mxc, which can multimerize through polyvalent interactions to potentially form a network of Mxc molecules(Terzo et al., 2015), located in the core where the histone genes and pol II are also located and the C-terminus of Mxc is in the shell of the HLB(Kemp, Yang, Dominski, Marzluff, & Duronio, 2021). The synthesis of histone mRNA occurs within the core of the HLB, where nascent histone mRNA is detected by in situ hybridization (Kemp et al., 2021). Processing occurs rapidly after transcription of the 3’ processing signal and is followed by transcription termination.

There are four factors required for histone gene expression whose only known functions in gene expression are in histone mRNA biosynthesis: NPAT/Mxc, FLASH, U7 snRNP and SLBP. The first three are core components of the HLB, and SLBP binds the nascent transcript in the HLB. SLBP participates in processing and export from the nucleus, as well as translation of the histone mRNA in the cytoplasm. SLBP is an abundant protein and remains bound to the histone mRNA throughout its life cycle, and there must be one SLBP protein for every histone mRNA. NPAT/Mxc, FLASH, Lsm10 and Lsm11 are low abundance proteins, that are concentrated at the histone genes.

Active U7 snRNP may be present only in S-phase cells in Drosophila.

Regulation of histone mRNA biosynthesis occurs within the HLB, with the initial activating step being phosphorylation of NPAT/Mxc by cyclin E/cdk2, although there could also be other substrates of cyclin E/cdk2. Strikingly RNA pol II is present in the HLB throughout the cell cycle, and is essential for its formation, so recruitment of pol II to the HLB is not a regulatory step (Kemp et al., 2025).

The active form of U7 snRNP is bound to FLASH and the histone cleavage complex (mCF). Symplekin, a component of the mCF complex, interacts directly with the region of FLASH essential for processing (Sun et al., 2020), which is bound to the N-terminal region of Lsm11. Symplekin is readily detected in the HLB in S-phase in syncytial embryos by IF in Drosophila. In the mature polytene salivary gland cells that are not cycling symplekin is not detected in the HLB, although the core U7 snRNP is present in the HLB.

We can detect S-phase cells using an antibody that recognizes phosphorylated Mxc. We analyzed cycling ovarian follicle cells which are proliferating around the egg chamber as it grows. Symplekin is required for synthesis of all polyadenylated mRNAs, and as a result it is always present throughout the nucleoplasm. Symplekin is concentrated in the HLB in S-phase follicle cells (present at a higher concentration in the HLB then in the surrounding nucleoplasm), but is not concentrated in HLBs in G1 phase (Tatomer et al., 2014). This result strongly suggests that mCF may assemble with core U7 snRNP only after cyclin E/cdk2 is activated to activate DNA replication.

Effect of FLASH mutants in Drosophila on histone pre-mRNA processing.

Support of this interpretation comes from genetic experiments in Drosophila. The C-terminus of FLASH, which is not required for processing, is required to localize FLASH to the HLB. It binds directly to the C-terminus of NPAT(X.C. Yang et al., 2014) and near the C-terminus of Mxc(Kemp et al., 2021). A two-amino acid mutation in either FLASH or U7 snRNP prevents binding of FLASH to Lsm11 in vitro and this interaction is conserved in mammals and Drosophila (Burch et al., 2011). These mutations have no effect on localization of FLASH or U7 snRNP to the HLB, but abolish processing in vivo in Drosophila (Burch et al., 2011; D.C. Tatomer et al., 2016). We obtained a mutation in FLASH that reduces its expression by >90% results in flies that die in pupa stages with a small fraction becoming adults. The mutant flies express primarily polyadenylated histone mRNAs as determined by S1 nuclease mapping (Fig. 12). We tested test the ability of various FLASH proteins to rescue this mutation. Expression of full-length FLASH rescued the mutation and expression of FLASH with a 2 amino acid mutation that prevents binding to Lsm 11 did not have any effect on the mutation. Two mutants, FLASH 1-733, which is active in processing but isn’t concentrated in the HLB and miniFLASH, with the entire central region deleted (Fig. 12), rescued viability and most of the histone mRNA was properly processed. There were small amounts of polyadenylated histone mRNA was expressed and long read-through (unprocessed) transcripts were also detected

Fig. 12.

Delayed processing of histone mRNAs results in read-through transcripts. In situ hybridization of transcripts containing the coding region or transcripts extending past the normal termination site in endoreduplicating nurse cells in the Drosophila ovary. A. Diagram of the two probes used. B. First two columns: in situ hybridization of the coding probe green and DAPI (red) in different FLASH genotypes. Note about 50% of the cells are in S-phase indicating by the coding probe. Last three columns. Read -through transcripts hybridizing with the downstream probe. Polyadenylated histone mRNAs are in the cytoplasm and transcribing histone genes are in the HLB in the nucleus (arrows). Adapted from Tatomer et al., 2016

We assayed for expression of the coding region and the 3’ region of the gene after the termination site by in situ hybridization in endoreduplicating nurse cells in the ovaries. As expected the 3’ probe detected polyA histone mRNA in the cytoplasm in the mutant, and expression of the 3’ region in the HLB as nascent transcript. However when FLASH was not concentrated in the HLB, processed histone mRNA was present in the cytoplasm, but there was expression of the downstream region in the HLBs, indicating pol II did not terminate close to the 3’ end but nascent transcripts further down the gene are detected by in situ hybridization(D.C. Tatomer et al., 2016) (Fig. 13B, 1-733). Thus the failure to rapidly process the histone mRNA result in transcription past the normal termination site, but the histone mRNA was ultimately processed. We interpret these data to mean that while RNA pol II slows down and/or pauses after the HDE, if processing does not occur and transcription then continues past the normal termination site, the transcripts can still be processed, but processing is no longer coupled with termination(D.C. Tatomer et al., 2016).

Fig. 13.

Regulation of assembly of U7 snRNP during the cell cycle in Drosophila. In G1 cells, FLASH and core U7 snRNP are in the HLB, but the mCF complex is not. Activation of histone gene expression occurs following activation of cyclin E/cdk2, which phosphorylates NPAT/Mxc. This results in concentration of symplekin (mCF) in the HLB activating processing.

Activation of cyclin E, which phosphorylates NPAT/Mxc to activate histone mRNA expression may be the signal to allow binding of FLASH to U7 snRNP inside the HLB (Fig. 13). The complex would then bind mCF that diffuses into the HLB resulting in concentrating the mCF cleavage module in the HLB, and activating U7 snRNP. Symplekin was discovered as regulatory factor in mammalian cells, and it is likely this same pathway also exists in mammals. SLBP also a limiting factor for processing and is present in low levels in G1 and G2 mammalian cells, providing an independent mechanism for inhibiting histone pre-mRNA processing. Unlike mammalian cells, in Drosophila cells SLBP is present throughout the cell cycle(Potter-Birriel, Gonsalvez, & Marzluff, 2021), making this regulatory event critical in Drosophila cells

Note that cell-cycle regulation also occurs at the level of histone gene transcription with the recent finding that pause-release, activated downstream of cyclin E/Cdk2 activation, is the critical regulatory step for activating histone gene transcription Drosophila (Kemp et al., 2025).

Cyclin L/Cdk11 is required for histone pre-mRNA processing in vivo.

to efficient processing and proper termination of histone pre-mRNAs in vivo in mammals is cyclin L/cdk11, which phosphorylates Serine 2 of the Pol II CTD as well as likely other substrates, including FLASH(Gajduskova et al., 2020). Two other homologous kinases, Cyclin K/cdk12 and cyclin K/cdk13 phosphorylate the Serine 2 on the Pol II CTD and act near the 3’ end of transcripts which will be cleaved and polyadenylated(Chirackal Manavalan et al., 2019; Kohoutek & Blazek, 2012). Dolibor Blazek and coworkers found that knockdown of Cdk11 in mammalian cells, results in inhibition of progression through the cell cycle and reduction in histone mRNA. due to failure to efficiently process histone mRNA (Gajduskova et al., 2020). They found that Cyclin L crosslinks to the nascent histone transcripts, binding to the 3’ UTR before the stemloop on histone pre-mRNA(Gajduskova et al., 2020). Cyclin L/Cdk11 directly interacts with FLASH and phosphorylates it. A consequence of Cdk11 KD is read-though of histone gene transcription past the normal termination site. These results are consistent with a role for cyclin L/cdk11 promoting proper elongation of histone mRNAs through the 3’ half of the gene and promoting transcription-coupled histone pre-mRNA processing by interacting with the 3’ UTR of the histone pre-RNA. The Cyclin L/cdk11 RNA binding site is not sequence specific. Cyclin L/cdk11 could be positioned on the RNA by interacting with the SLBP/SL complex on the nascent mRNA to help activate processing and coordinate it with termination. Cyclin L/cdk11 is not required for processing in vitro.

Misprocessing of histone mRNAs.

Failure to process histone mRNAs normally results in rapid degradation of the unprocessed processed transcript, and unprocessed histone mRNAs don’t accumulate in the cell. “Misprocessing” of histone mRNAs occurs when the histone mRNA is not processed after the stem-loop but ultimately is polyadenylated and can then be exported to the cytoplasm and translated. Misprocessing occurs on all 5 Drosophila histone mRNAs when processing factors such as SLBP and U7 snRNP are mutated or depleted(Lanzotti et al., 2002).

Polyadenylated histone mRNAs have not been reported in wild-type Drosophila tissues. The presence of these polyadenylation sites in all 5 Drosophila histone genes, suggest they have been selected for. The Drosophila histone genes are close together and 4 of the histone genes (the H2a/H4 genes and the H2B/H1 genes have their 3’ ends <500 nts apart (Fig. 2). It is possible that, the polyadenylation sites may provide a mechanism to reduce the chance of RNA polymerases colliding with each other if they fail to process histone mRNAs after the stemloop.

In contrast in mammalian cells there are some RD histone genes which can produce two different transcripts, one that ends at the stemloop and the other that is longer and polyadenylated. The gene encoding histone H2a.X, a protein that is produced both in G1 and G2, as well as in S-phase, expresses two transcripts(Mannironi, Bonner, & Hatch, 1989). During S-phase it produces a transcript processed at the stemloop which is replication-dependent. In G1 and G2-phases it produces a polyadenylated transcript(Bonner, Mannironi, Orr, Pilch, & Hatch, 1993; Griesbach, Schlackow, Marzluff, & Proudfoot, 2021). The H2A.X gene is not present at the HLB, but in a gene rich region on human chromosome 11. Production of polyadenylated H2a.X mRNA in G1 is necessary for cells to efficiently enter S-phase after release from G0 or G1 arrest and to maintain efficient response to DNA damage throughout the cell cycle(Griesbach et al., 2021).

In addition to the H2a.X gene a subset of the RD histone genes produce polyadenylated histone mRNAs under some conditions. Art Skoultchi and coworkers found that when MEL cells were induced to differentiate a polyadenylated form of the histone H1-2 mRNA was produced(Cheng, Nandi, Clerk, & Skoultchi, 1989; Cheng & Skoultchi, 1989), and Moss and coworkers reported that histone H2AC16 mRNA was polyadenylated in mouse spermatids(Moss, Ferry, & Groudine, 1994). With the advent of high-throughput sequencing the Illumina Body Project sequenced total RNA from many human tissues. In all differentiated cells a subset of 9 histone mRNAs produced from the Hist1 and Hist2 clusters were polyadenylated (Lyons et al., 2016), including the H1-2 and H2AC16 mRNA, as well as several other H2a and H2b mRNAs and the H4X gene and the H3X genes produced polyadenylated histone mRNAs. The same histone mRNAs were polyadenylated in mouse tissues. Strikingly in several of the H2b mRNAs that are polyadenylated, the transcripts are also spliced with the 5’ splice junction in the 3’ UTR before the stemloop and is conserved in different H2b genes that can undergo polyadenylation(Lyons et al., 2016). As a result these mRNAs cannot bind SLBP, while the polyadenylation sites on the other histone genes are all after the stem loop, and the stemloop is present in the polyadenylated transcript of these genes. In the case of the H2a.X mRNA SLBP is found bound to at least some of the polyadenylated H2a.X mRNA(Brooks et al., 2015).

When mouse fibroblasts were arrested by serum starvation they did not induce polyadenylated histone mRNAs, but when the 3T3 L1 fibroblasts were arrested and induced to differentiate into adipocytes then polyadenylated mRNAs were induced(Lyons et al., 2016). Similar results were found for the H2b mRNAs in differentiation of stem cells(Kari et al., 2013). The function of the low levels of some polyadenylated histone mRNAs in terminally differentiated cells is likely to provide a mechanism for replacing defective histones as cells age. Fred Zweidler found 40 years ago that the core histone proteins changed dramatically as mice age, even in the absence of cell division (Zweidler, 1984). In particular, the H2b and H2a isoforms changed, even though all of these proteins are encoded by replication-dependent histone genes. The changes in these proteins are consistent with the expression of the polyadenylated forms of several H2b genes, and the H2a genes(Lyons et al., 2016). Thus polyadenylated mRNAs for the core histones likely provide continued expression of histone proteins at low levels in terminally differentiated cells to replace damaged histones as cells age.

Misprocessing of histone mRNAs results from interfering with transcriptional pausing in mammalian cells.

A surprising discovery made initially by Handa and coworkers was that knockdown of NELF, a negative elongation factor involved in promoter proximal pausing, resulted in polyadenylation of some replication-dependent histone mRNAs(Narita et al., 2007), H2AC16, H2BD CX and H1-2, which are also expressed as polyadenylated mRNAs in terminally differentiated cells. The amount of polyadenylated histone mRNA relative to properly processed histone mRNA was low, <5%. A number of subsequent studies have shown that similar results are obtained after knockdown of other factors involved in promoter proximal stalling. These include Ars2(Gruber et al., 2012), p-TEFb(Pirngruber, Shchebet, & Johnsen, 2009), and H2b monoubiquitination induced by Cdk9(Pirngruber, Shchebet, Schreiber, et al., 2009). The fold-increase in polyadenylated histone mRNA was large (10-fold or more), but because histone pre-mRNA processing is normally efficient and very little if any polyadenylated histone mRNAs was produced in normally growing cells, usually <5% of the total RNA from each histone gene was polyadenylated in these experiments. All of the factors reported to induce polyadenylated histone mRNAs are involved in the early stages of transcription. These results suggest that in vivo, proper pause-release of the RNA polymerase may be essential for efficient histone pre-mRNA processing. In addition as described above, Cyclin L/cdk11 is also required for efficient histone pre-mRNA processing in vivo. Cyclin L/Cdk11 may regulate elongation near the 3’ end of the transcript and is bound to the nascent histone mRNA just 5’ of the stemloop. In the absence of Cdk11 there is also misprocessing of histone mRNAs(Gajduskova et al., 2020). Note that these factors are not found associated with the processing complex, emphasizing a role for regulated transcription elongation for proper histone pre-mRNA processing in vivo.

Evolution of histone pre-mRNA processing.

The origin of the mechanism of histone pre-mRNA processing is not clear. With the evolution of the nucleosome to package DNA, it became imperative to coordinately synthesize the 5 different histone proteins (and especially the four core histones) to ensure rapid assembly of newly replicated DNA into chromatin. There are examples of diverse single cell eucaryotes that contain SLBP (Lopez & Samuelsson, 2008), and whose histone mRNAs contain stemloops, suggesting that the initial histone mRNAs may have had a structure different from other mRNAs. Lopez & Samuelsson, 2008). SLBP and the SL at the 3’ end of histone mRNAs are readily identified using informatics (Lopez & Samuelsson, 2008), whereas U7 snRNA, Lsm10 and Lsm11 or NPAT are not (Marz, Mosig, Stadler, & Stadler, 2007). These species with SLBP include, some green algae, including Chlamydomonas and Volvox; some Mycetozoans, including Dictyostelium and Physarum: and several members of the Apicomplexa. Many of these same species have stemloops at the 3’ end of their histone genes and some have Lsm10 and Lsm11 homologues(Lopez & Samuelsson, 2008) . Analysis of Physarum histone genes showed that each histone gene also has a polyadenylation site 3’ of the stem loop. Both mRNAs that end at the stemloop and polyadenylated mRNAs are expressed, and Lsm10 and Lsm11 homologues are present in the genome(Poulet et al., 2021). The specialized 3’ end of histone mRNA may have been lost from most lineages, but has been maintained in metazoans, which evolved an effective system for coordinate regulation of these critical mRNAsby localizing their genes in a subdomain in the nucleus.

In contrast in Chlamydomonas the histone mRNAs all end at the stem-loop and are cell-cycle regulated similar to the mammalian mRNAs (Z. Walther & Hall, 1995). This evolutionary data supports the idea that histone mRNAs ending in stemloops may be an ancient property of these mRNAs, which has likely been lost from most unicellular eucaryotic species as well as plants. In metazoans, a system localizing the histone genes in the subdomain of the nucleus and regulating the processing of histone mRNA, has evolved to play a critical role in cell cycle regulation.

Interestingly in C. elegans (and related nematodes), histone mRNAs end in a stem-loop and are bound by SLBP, but the processing pathway is altered. Instead of cleavage after the 3’ end, a longer polyadenylated transcript is produced(Mangone et al., 2010), which is processed by an RNAi-like mechanism. The transcript, which is bound by SLBP, is cleaved after the stemloop and before the polyA site and then trimmed by an exonuclease(Avgousti, Palani, Sherman, & Grishok, 2012) and the mRNA ends 5-6 nts after the stemloop(Keall, Whitelaw, Pettitt, & Muller, 2007). As a result of the evolution of this mechanism, C. elegans does not contain a U7 snRNA, Lsm10 and Lsm11 or NPAT. There is not a histone locus body and the histone genes, rather than being in one or two large clusters, are present in multiple small clusters.

Conclusions.

Metazoan replication-dependent histone mRNAs are likely the only eucaryotic cellular mRNAs that do not have a polyA tail, although many plant viral RNAs as well as some animal viral RNAs including flaviviral RNAs (dengue, yellow fever) are also not polyadenylated (while plant histone mRNAs are polyadenylated). There a number of factors whose sole function seems to be in histone mRNA metabolism (NPAT, FLASH, U7 snRNP, SLBP) in metazoans. SLBP is bound only to histone mRNAs in mammalian cells (Brooks et al., 2015; Van Nostrand et al., 2016) and is not homologous to any other known RNA binding protein. A complex system for expressing histone mRNAs has evolved in metazoans, with the histone genes concentrated in the histone locus body (HLB), where transcription and processing of the histone mRNA takes place. All known factors in histone mRNA metabolism are conserved between mammals and Drosophila. Histone 3’ end formation occurs by co-transcriptional cleavage to form the 3’ end, with CPSF73 in the mCF complex catalyzing cleavage as it does in cleavage/polyadenylation. Processing is activated by assembly of the active U7 snRNP, with FLASH, a core HLB component, and Lsm11, a U7 snRNP protein interacting within the HLB to recruit the mCF complex activating processing.

Fig. 6.

An active recombinant histone processing complex.

A. Schematic of the complex. B. color-coded diagram of the proteins. In the initial complex we tested, Symplekin was missing the first 300 amino acids because it was difficult to express large amounts. C. Activity of the complex. We then added the first 300 amino acids as a recombinant protein We had previously determined the structure of the N-terminal domain that was missing, and when we added that in trans activity was restored. Fig. Adapted from Sun et al. 2020.

Fig. 10.

CPSF-73 is both an endonuclease and a processive 5’ to 3’ exonuclease. Liang Tong used a substrate with fluorescent labels at both the 5’ end (red) and 3’ end (green). The purified complex cleaved the substrate leaving processed mRNA (26 nts, red) and the 3’ fragment (34 nts, green), which was degraded to a very short product. Note the degradation removes the complete HDE, so CPSF73 must degrade well past the U7 snRNA binding site. Fig. adapted from Sun et al (2021).