Emerging perspectives on multidomain phosphatidylinositol transfer proteins

Abstract

The phosphatidylinositol transfer protein domain (PITPd) is an evolutionarily conserved protein that is able to transfer phosphatidylinositol between membranes in vitro and in vivo. However some animal genomes also include genes that encode proteins where the PITPd is found in cis with a number of additional domains and recent large scale genome sequencing efforts indicate that this type of multidomain architecture is widespread in the animal kingdom. In Drosophila photoreceptors, the multidomain phosphatidylinositol transfer protein RDGB is required to regulate phosphoinositide turnover during G-protein activated phospholipase C signalling. Recent studies in flies and mammalian cell culture models have begun to elucidate functions for the non-PITPd of RDGB and its vertebrate orthologs. We review emerging evidence on the genomics, functional and cell biological perspectives of these multi-domain PITP_d containing proteins.

Introduction

The evolution of specialized membranes in living cells came with the requirement to populate these membranes with unique proteins and lipids that serve specific functions. In the case of lipids, given their hydrophobic nature, mechanisms to deliver and remove lipid molecules from specific membranes are a specific requirement. The movement of lipids in cells can be achieved through several mechanisms, including spontaneous fusion of membranes, vesicular transport and the transfer of hydrophobic lipids through aqueous cytosol bound to proteins. This latter function is performed by lipid transfer proteins (LTP). The earliest reports on LTP identified it as a ‘cytosolic factor’ capable of transferring Phosphatidylcholine (PC) between mitochondria and microsomes in vitro. Since this factor was found to be heat-labile and sensitive to trypsin mediated digestion, it earned the name of a lipid transfer protein [1]. Since then, using biochemical, genetic, molecular cloning and genomics approaches, multiple LTPs have been identified across a range of organisms. These LTPs have been classified into 27 protein families [2]. These include distinct LTPs that transfer specific classes of lipids including phospholipids, sterols, glycerides, glycosphingolipids, etc, with each LTP demonstrating a degree of specificity for the lipids they transfer. Most LTPs have a hydrophobic cavity where the lipid is bound, and is shielded from the aqueous cytosol by a short mobile protein stretch acting as a ‘lid’.

Phosphatidylinositol transfer proteins (PITPs) are a class of LTPs that are able to transfer phosphatidylinositol (PI) between membranes. The first PITP, capable of transferring PI from rat liver microsomes to liposomes, was identified and biochemically characterized from bovine brain cytosol [3] and the gene underlying it subsequently cloned [4]. In yeast a protein a with similar in vitro PI transfer function was also identified later [5]. It was found to be identical to the secretory protein Sec14 which had earlier been reported to transport secretory proteins from the Golgi [6–8]. PITPs have also been identified in other fungi, D. discoideum, C. elegans, D. melanogaster, plants and mammals. However, it is important to note that though sec14 like PITPs show similar function, they do not share any structural or sequence similarity with the PITPs identified in other organisms. The PITP_d is present in all sec14-unrelated PI transfer proteins; and based on the sequence homology of the PITP_d, PITPs have subsequently been classified into two groups: the Class I group of PITPs comprise of PITPα and PITPβ which have only the single PITP_d and Class II PITPs. The Class II proteins are further sub-classified into two groups: (i) Class IIA which represent multi-domain PITPs, which in addition to the PITP_d contain additional motifs and domains (ii) Class IIB PITPs which in addition to the PITP_d have a short C-terminal tail of 60-70 amino acids (reviewed in Husain and Cockcroft, 2001). The discovery and functions of Class I PITPs are reviewed elsewhere in this issue (Cockcroft 2021). Here we focus on the multidomain Class IIA PITPs.

Discovery of multidomain PITPs

The founding member of the family of multidomain PITPs is the Retinal Degeneration B (RDGB) protein in Drosophila, and hence this group is also annotated as RDGB family of proteins. The RDGB protein (encoded by the rdgB gene) was initially found in a screen which identified genes necessary to support phototransduction in Drosophila photoreceptors [9]. Mutation of the rdgB gene in Drosophila led to retinal degeneration of the photoreceptors, hence the name rdgB. The rdgB gene encodes a protein of 160 kDa with the N-terminal region that shows strong sequence homology to PITPα (reviewed in [10]). Independently, the first mammalian orthologs of rdgB were identified by Lev and colleagues as proteins that could interact with the N-terminal domain of protein tyrosine kinase PYK-2, and are hence also named as Nir (N-terminal domain-interacting receptors) proteins [11]. The defining feature of multidomain PITPs is the presence of at least one domain in addition to the PITP domain. In the founding members of this multidomain PITP family, in addition to the PITP domain, an FFAT motif, DDHD and LNS2 domains have been described. Since then its orthologs in other species including humans were identified by using the Drosophila gene as a template and screening for similar sequences in cDNA libraries [12,13]. While the Drosophila and C.elegans genome [14] contain a single gene for a multi-domain PITP, the human genome encodes two of these: PITPNM1 /RdgBαI/Nir2 and PITPNM2/RdgBαII/Nir3. The approved name for these two genes as per the Human Genome Nomenclature Committee (https://www.genenames.org/) is PITPNM1 and PITPNM2.

Domain structure of the classic multidomain PITPs

The rdgB orthologs cloned from the established model organisms such as Drosophila, C.elegans, rodents and humans encode proteins that contain an N-terminal PITP domain (PITP_d); C-terminal to this lie the FFAT motif, the DDHD and LNS2 domains. The C-terminal DDHD and LNS2 domains are separated from the N-terminal PITP_d and the FFAT motif by a large unstructured region (Figure 1A).

Figure 1. Domain structure of multidomain PITPs and USR region.

(A) The figure represents domain architecture of human PITPNM1, PITPNM3, RDGB (Drosophila melanogaster) and PITP1 (C.elegans). The domains have been labelled on the protein and are length adjusted. The figure has been generated using IBS (version2.0). (B) USR region is defined as the unstructured region between the FFAT motif and DDHD domain in RDGB. Three disorder prediction methods were used to predict the probability of disorder in the USR region. The figure represents the prediction by DisEMBL (green), PrDOS (cyan) and IUPred (pink).

PITP_d

The initial classification of PITPs was based on the sequence similarity of the N-terminal PITP_d (https://www.ebi.ac.uk/interpro/entry/InterPro/IPR001666/); on this basis Class I and Class II PITPs were defined [15]. More recently biochemical and functional studies have examined the lipid transfer activity of the PITP_d. While both classes of PITP can transfer PI, Class I PITPs can additionally transfer PC, whereas Class II are not able to do so [16,17]. Instead the PITP_d of Class IIA [17] and Class IIB [16] PITPs is able to transfer PA in vitro.

Although there is no experimentally determined structure for a Class II PITP_d, the structures of human and rat PITPα with PC or PI show the PITP_d is comprised of three parts: a lipid-binding core, a regulatory loop, and the C-terminal region. In the lipid binding hydrophobic pocket, the polar head group is oriented towards the centre of the pocket, while the acyl chains radiate towards the surface [18,19]. The hydrophobic core can accommodate either of the two lipids at a time. Residues which bind the phosphate group are the same for either PI or PC (Q22, T97, T114, and K195), while unique contacts with the inositol ring of PI are mediated by amino acids T59, K61, E86 and N90. These residues are conserved in the PITP_d of RDGB and mutation of any of these results in loss of ability to bind PI, the ability of the protein to support inositol 1,4,5 trisphosphate (IP₃) generation in vivo and the ability to rescue any of the phenotypes of Drosophila rdgB mutants in vivo [17]. While mutation in the PI binding residues of RDGB also affect PA binding [17], without any impact on PC binding, no residues that specifically mediate PA binding by the PITP_d of Class IIA have been identified thus far. It may be interesting to identify such residues; analysis of these residues will help resolve the molecular mechanism of PA binding by the PITP_d. It will also facilitate an analysis of the contribution of PA binding and transfer by Class IIA PITP to physiological function in vivo uncoupled from the contribution of PI binding and transfer.

It has been noted that expression of the PITP_d of RDGB is sufficient to rescue key photoreceptor phenotypes of the rdgB mutant [17,20] and this has also been reported for C.elegans pitp-1 [14]. Interestingly, reconstitution with a Drosophila Class I PITP (vib) [21] could not rescue the phenotypes of rdgB mutants [17]. This observation likely reflects the distinctive lipid transfer properties of the PITP_d of Class IIA PITPs in relation to their in vivo function.

FFAT motif

The FFAT [diphenylalanine (FF) in an acidic tract] motif is a short peptide sequence that was initially identified in oxysterol binding protein-related proteins (ORPs) [22]. This motif is characterized by the presence of the amino acid sequence EFFDAxE, that acts as an ER targeting signal. The FFAT motif is present not only in multidomain PITPs, but also in several other groups of lipid transfer proteins such as the oxysterol binding protein (OSBP) and ORPs [23], Ceramide Transfer Protein (CERT) [24] and Vps13A [25]. A large number of proteins contain FFAT motifs which allow them to interact with the endoplasmic reticulum (ER) through the MSP domain of the ER resident protein VAP [26]. A very recent study identified a non-conventional FFAT motif present in the cholesterol transfer protein StarD3 which functions at ER-endosome contact sites [27,28]. The non-conventionality arises due to the presence of a serine/threonine residue in place of an acidic amino acid which can be phosphorylated, and hence this motif is also named as phospho-FFAT motif. The phosphorylation of the FFAT motif allows the interaction of StarD3 with the MSP domain of VAP thus facilitating sterol exchange at ER-endosome contacts.

DDHD domain

The DDHD domain (https://www.ebi.ac.uk/interpro/entry/InterPro/IPR004177/) was first noted in mammalian PITPs [29] and subsequently in some members of the phospholipase A1 (PLA1) family of enzymes [30]. The nomenclature of this domain is derived from the 4 conserved amino acids D,D, H and D; present in both PITPs and PLA1 proteins. These amino acids are predicted to a site for binding divalent metal similar to what is observed in phosphoesterase. Although the exact role of this domain in regulating cellular functions is presently unclear, it has been proposed to regulate cell signalling, membrane trafficking, lipid metabolism and lipid transfer based on the activity of the proteins in which it is found. While there have been limited studies on the role of the DDHD domain in PITPs, studies done on the PLA1 proteins have suggested that it binds phosphatidylinositol 4 phosphate (PI4P) and may play a role in protein localization [31,32].

LNS2 domain

The nomenclature of the LNS2 domain (https://www.ebi.ac.uk/interpro/entry/InterPro/IPR031315/) is based on the proteins where this domain was initially identified: Lipin, Ned1, and Smp2 proteins. LNS2 domain was first discovered in lipin proteins which function as PA phosphatases converting PA to DAG [33]. Other proteins such as Ned1 is found to regulate nuclear morphology and chromosome stability in yeast, while Smp2 is also a lipin homologue in yeasts. Importantly, this domain is also present at the C-terminal end of Class II PITPs. However, in contrast to lipins, the LNS2 domain in Class II PITPs can bind PA but not hydrolyse it. The conserved Asp residue in lipins is substituted with Ser in the LNS2 domain of Class II PITPs, likely rendering the latter phosphatase inactive.

Unstructured Region

Although most regions of a polypeptide have a secondary structure, some proteins have regions that are intrinsically disordered or unstructured. These regions are important for interaction with other proteins and nucleic acids, e.g RNA. Unstructured regions are predicted as coils by secondary structure prediction methods and do not show homology to any known secondary structure domains in proteins. The amino acid residue arrangement tends to take up a disordered conformation in the lower stable energy state of protein. The region between FFAT motif and DDHD domain in RDGB protein appears to be unstructured (USR). The secondary structure prediction methods predict coils in that region. Three independent methods (DisEMBL[34], PrDOS [35] and IUPreD [36]) that predict disordered regions suggest that most of the region marked as USR are disordered (Figure 1B).

Diverse domain composition of multidomain PITPs

Previous studies have noted that the lipid transfer domain of LTPs is variably associated with other domains [37,38]. The Pfam family for PITP_d (https://pfam.xfam.org/family/PF02121) contains 2785 sequences classified into 34 different domain architectures. After removing obsolete entries and sequences with no taxonomic classification, there are 2611 sequences. Of these sequences, for seven of the proteins, the structure of the PITP_d is available in PDB (Protein Data Bank). We analyzed these sequences for the presence of 8 residues experimentally determined to be required for PI binding to the PITP_d. Out of the 8 residues, 4 residues (E86, K61,T59 and K195) are more than 90% conserved, 2 of them (N90 and Q22) are conserved above 80%, 1 of them (T97) is 69% conserved and 1 (T114) is 38% conserved.

We classified the sequences based on the associated domains (Figure 2A) and taxonomy (Figure 2B). PITP_d as single domain exist in 1865 sequences. Among proteins that contain at least one other domain in addition to the PITP_d, we observed that the most common domain combination is presence of the DDHD and LNS2 domains C-terminus to PITP_d. There are 661 sequences with this domain architecture (Table 1). PITP, DDHD and LNS2 domain also exist in combination with other domains but such combinations are fewer in number. The second most common domain combination is PITP_d associated with DDHD at the C-terminus; there are 28 sequences in this class. The other domains commonly associated with PITP_d are PH domain, UCH domain and Oxysterol-OSBP domain. The PH domain helps to target the proteins to different membranes by binding to lipids on the membranes [39]. OSBP domain containing proteins are lipid transfer proteins that controls the transfer of lipids between organelles [40]. UCH domains are present in proteins involved in the ubiquitination of cellular proteins [41]. Thus it would appear that the domains most commonly associated with the PITP_d are also lipid binding in nature and can help in localization or membrane targeting. Interestingly, among sequenced genomes there is a set of genes that encode proteins that contain all of the domains in a classical Class IIA PITP (e.g RDGB) except the PITP_d itself, prominent examples include PITPNM3 (Nir1) and zebrafish PITPNM3 (pl-RdgB)[13]. Studies in zebrafish have indicated that despite the lack of a PITP_d, this type of protein has physiological functions [42].

Figure 2. Distribution of multidomain PITPs in various branches of the tree of life.

(A) Pie chart to represent the number of different domain combinations that exist along with PITP domain in various sequenced genomes. (B) The figure represents the distribution of various domain combinations in metazoan phyla. ‘*’ represents presence of the domain combination in the phyla as obtained from the Pfam database. For sequences which were not deposited in Pfam, Blastp was used to search for sequences in respective phyla. ‘*’ represents sequences obtained using Blastp.’*‘ represents domain combinations not found.

The taxonomic distribution of the PITP_d containing sequences is very diverse (Figure 2B). The single PITP_d consisting sequences are present in unicellular organisms, plants, most invertebrate and vertebrate phyla. Sequences encoding proteins with an architecture that includes the PITP_d, DDHD and LNS2 domains exist in Porifera, Rotifera, Cnidaria, Tardigrada, Arthropoda, Placazoa, Brachipoda, Mollusca, Nematoda, Annelida and Chordata, i.e only in metazoan genomes. The PITP_d followed by DDHD domain at C-terminus exist in fewer lineages such as Porifera, Platyhelminthes, Nematodes, Arthropoda and Chordata. These observations clearly suggest the importance of co-existence of domains for functional requirement along evolution of metazoans. The associations of other domains with the PITP_d are typically restricted to single phylum and thus may have evolved for organism specific function.

Signalling functions of multidomain PITPs

The role of PITPs in regulating cellular signalling was demonstrated when it was first identified as a cytosolic factor that could support G protein-stimulated phospholipase C (PLC) activity in permeabilized HL60 cells [43]. Sec14p, and mammalian PITPα and PITPβ could all reconstitute PLC signalling during G-protein coupled receptor (GPCR) activation by specific agonists [44]. Indeed, this reconstitution of PLC activity is dependent on the PI transfer ability of PITP_d, since versions of PITP_d that cannot bind PI are unable to support PLC activity during GPCR signalling [19]. PI synthesized in the ER is transferred to the PM by PITPs where it is phosphorylated to form PI4P and then phosphatidylinositol 4,5 bisphosphate (PIP₂) during signalling. PITP dependent production of PI4P and PIP₂ has been shown to regulate other processes like secretory granule formation, regulated exocytosis, vesicle budding, etc. [reviewed in [15]].

Hydrolysis of PIP₂ into IP₃ and diacylglycerol (DAG) will also require the resynthesis of PIP₂. Similar to Class I PITPs, the members of the Class II PITPs might be expected to support PIP₂ turnover during receptor mediated PLC activation. To date the only in vivo, physiological system where the requirement for Class IIA PITPs in supporting phospholipase Cβ (PLCβ) PLCβ signalling is Drosophila photoreceptors [45]. The founding member of the Class IIA PITP family, RDGB is highly expressed in these cells where phototransduction is mediated by G-protein mediated PLCβ activation. In these cells loss of RDGB result in a reduction of light response, defective PIP₂ resynthesis during phototransduction and light-dependent retinal degeneration (reviewed in [10]). The PITP_d of RDGB has been shown to transfer PI in vitro and mutations in the PITP_d that abolish binding and PI transfer activity fail to support the function of RDGB in regulating PLCβ signalling in these cells [46]. These findings remain the strongest evidence of a requirement for Class IIA PITPs in regulating PLC signalling in vivo.

The requirement of mammalian Class IIA PITPs in regulating PLC signalling has also been studied in cultured mammalian cells. In this setting, depletion of PITPNM1 is reported to slow the resynthesis of PIP₂ [47–49] following activation of PLC through the stimulation of cell surface receptors. Some studies have also reported that overexpression of PITPNM1 can also accelerate the resynthesis of PIP₂ following GPCR-PLCβ activation. The PITP_d of PITPNM1 appears to be required for this process [49]. It has also been noted that in mammalian cells, PITPNM1 is able to regulate the accumulation of PA and the synthesis of CDP-DAG during PLC stimulation [50], both of which appear to depend on the presence of its PITP_d. In mammalian genomes two genes PITPNM1 and PITPNM2 that encode Class IIA PITPs are found; it is reported that while PITPNM1 regulates PIP₂ levels during high rates of PLC activity, PITPNM2 controls this process during resting conditions [49]. Collectively these studies provide evidence of a role for Class IIA PITPs in regulating the turnover of several lipids following PLC mediated PIP₂ hydrolysis.

Cell biological perspectives of Class IIA PITP function

Studies in Drosophila have shown that the endogenous RDGB protein from heads extracts is a membrane associated protein [51]. This study also described the localization of the RDGB protein in photoreceptors where it was found to be enriched in the photoreceptors. Immunoelectron microscopy analysis found that the RDGB protein was specifically localized to the base of the microvilli that make up the apical, light sensitive plasma membrane of the photoreceptor [51] and subsequent studies using independent antibodies and protein null alleles as controls have also confirmed these findings [17]. Other proteins localized at this location include two enzymes involved in PA generation, the diacylglycerol kinase (RDGA) [52] and phospholipase D [53,54]. In the region of the photoreceptor where the RDGB protein is localized, the endoplasmic reticulum (ER) and the plasma membrane (PM) come into close proximity with each other (< 10 nm) forming what is referred to as a membrane contact site (MCS)[55] (Figure 3).

Figure 3.

Representative image of a Drosophila compound eye, composed of repeating units called ommatidia. Each ommatidium is a collection of photoreceptor cells which can be seen in a transverse section. Magnified image of a photoreceptor cell is shown where the PM (rhabdomere) forms contacts with the ER (sub-microvillar cisternae). The multi domain Class IIA PITP, RDGB is localized at the ER-PM junction in photoreceptors.

At this MCS, on the PM side is the microvillar light sensitive membrane where PLCβ is activated leading to consumption of PIP₂ and the production of diacylglycerol which is then converted to PA and on the ER side is a specialization of the ER known as the sub-microvillar cisternae (SMC), marked by reticulon [56] where PI is synthesized (reviewed in [55]). Importantly, the RDGB protein remains stable at this location and does not show any redistribution even during high rates of light induced PLC activity that are a feature of these cells. Thus RDGB, a PI/PA transfer protein is positioned at this interface of organelle membranes where it could act as a protein that can supply PI from the SMC to the microvillar PM and also remove PA from the microvillar PM to the ER.

In cultured mammalian cell lines, RDGB is a membrane associated protein [57]. There are only limited studies on the localization of endogenous PITPNM1 and PITPNM2 in mammalian cells. Studies in cultured HeLa cells have reported through immunocytochemistry that the endogenous PITPNM1 protein is localized to the Golgi apparatus [58], cytokinesis furrow [59] lipid droplets and endoplasmic reticulum [57]. By contrast, several studies have reported, also in cultured mammalian cells, that the tagged PITPNM1 protein is diffusely distributed throughout the cell body [47,48,50]; these studies have also reported that after stimulation of the cultured cells with growth factors or agonists that activated PLC, PITPNM1 translocates to the plasma membrane and in some studies reported that they co-localize to ER-PM contact sites. It is unclear if endogenous PITPNM1 translocates in mammalian cells during PLC activation. This dynamic behaviour of PITPNM1 in cultured mammalian cells is fundamentally different from that in the Drosophila photoreceptor where RDGB localization at ER-PM contacts sites remains unchanged under both resting conditions and also during PLC activation. This translocation of mammalian PITPNM1 is reminiscent of that seen for other contact site proteins such as extended synaptotagmin-1 (E-Syt1) and STIM1 (reviewed in [60]) observed on studies in cultured mammalian cells. For instance, the mammalian Extended synaptotagmin-1 (E-Syt1) protein which transfers glycerophospholipids in a Ca²⁺ dependent manner [61,62] translocates to ER-PM contact sites upon elevation in levels of cytosolic Ca²⁺.

Mechanisms of sub-cellular localization

Given its precise localization of endogenous RDGB to ER-PM contact sites in Drosophila photoreceptors, the mechanism underlying this process have been analysed. Conceptually any protein that need to localize to ER-PM contact sites must require a molecular determinant to anchor it to the ER interface and one that anchors it to the PM interface. The RDGB protein has an FFAT motif and mutation of the FF/AA results in mis-localization of the protein away from the ER-PM MCS into the photoreceptor cell body [63]. Mislocalization of RDGB away from the ER-PM MCS by disrupting the FFAT motif phenocopies the key phenotypes of RDGB depletion in photoreceptors, including a reduced electrical response to light and reduced PIP₂ resynthesis following light stimulation [63]. Conversely, depletion of dVAP-A that in Drosophila photoreceptors is enriched at ER-PM contact sites, results in mislocalization of RDGB [63]. Thus the interaction between the FFAT motif of RDGB and dVAP-A is an essential determinant in anchoring RDGB to the ER-side of the ER-PM MCS. Likewise, an intact FFAT motif on overexpressed PITPNM1 is required for its recruitment to ER-PM contact sites following PLC activation in cultured mammalian cells [47] and depletion of VAP is reported to impact PITPNM1 function at the Golgi [23]. Although the FFAT motif is required for interaction with VAP and thus localization of RDGB, it is found to be insufficient. A large unstructured region present between the FFAT motif and the DDHD domain is essential for the formation of a stable interaction of RDGB with VAP [64].

The signal mediating the anchoring of RDGB to the PM aspect of the ER-PM contact site in photoreceptors has also been studied. A recent study reported that an intact FFAT/VAP interaction is insufficient for accurate localization but requires the presence of the C-terminal domains DDHD and LNS2 [65]. A truncated RDGB protein without the two C-terminal domains; RDGB^(DDHD-LNS2) _Δ, is not only mislocalized from the ER-PM contact site but also fails to complement RDGB function in photoreceptors. Further analysis revealed that each of these domains, individually makes contributions to RDGB protein localization. Deletion of the C-terminal LNS2 domain alone leads to mislocalization of RDGB and loss of function and the LNS2 domain when expressed by itself is able to localize to the plasma membrane in cultured Drosophila cells and to the apical membrane in photoreceptors [65]. Thus it appears that the LNS2 domain has a key role in localization of RDGB to the ER-PM MCS. The LNS2 domain of RDGB [65] and human PITPNM1 [48] binds PA and this may be the signal that mediates its interaction with the plasma membrane. Despite this compelling evidence on a role for the LNS2 domain of RDGB as a plasma membrane interaction signal, it has also been noted that mutation of key residues in the DDHD domain of RDGB also results in mislocalization of the protein and loss of function in vivo [65]. This was surprising since mutations in the DDHD domain leave the LNS2 domain fully intact. It also implies that while the LNS2 domain in necessary as a PM targeting signal, it is not sufficient. This study also found that the presence of the DDHD domain may influence the behaviour of the LNS2 domain and has proposed that interdomain interactions may play a role in the localization and function of RDGB [65]. Thus collectively multiple domains in RDGB, a multidomain PITP help to localize it to ER-PM contact sites (Figure 4).

Figure 4.

Representative image depicting the arrangement of each annotated domain in the multidomain PITP, RDGB. The N-terminal supports the PI transfer domain (PITP_d) activity, the FFAT motif and unstructured region (USR1) are shown. The interaction between the FFAT motif, dVAP-A and USR1 is depicted. The LNS2 domain interacts with the rhabdomere membrane. The physical inter-domain interaction between the DDHD and LNS2 domains is shown.

Extended synaptotagmins (E-Syt) have recently been described as proteins that localize to ER-PM contact sites and may mediate lipid transfer or tethering functions at this location [66]. In mammalian cultured cells, where tagged, overexpressed PITPNM1 has been shown to translocate to ER-PM contact sites on PLCβ activation, E-Syt1 function has been shown to be required for this translocation [47]. In Drosophila, whose genome encodes a single E-Syt (dESyt), photoreceptors show progressive loss of ER-PM contact sites and this is preceded by mis-localization of RDGB from this contact site [67]. In addition, dESyt is a genetic enhancer of rdgB, double mutants of rdgB;dESyt show enhancement of all photoreceptor phenotypes of rdgB mutants [67]. Together with the mislocalization of RDGB seen in dEsyt mutants, these observations suggest that in photoreceptors, dESYT plays a role in the localization of RDGB; the mechanism by which is does so remains to be determined. Interestingly, in this study, it was noted that rdgB mutant photoreceptors show very few ER-PM MCS even at eclosion, in sharp contrast to dESyt knockout flies; this implies that RDGB itself, apart from being localized to MCS is required for the establishment of ER-PM MCS; the reason for this is not known. It is possible that in addition to intrinsic targeting sequences in Class IIA PITPs additional proteins such as E-Syts may contribute to their localization.

Physiological functions of Class IIA PITPs

Given the strong evidence that Class IIA PITPs regulate PIP₂ turnover during PLC signalling, one might expect that these proteins regulate key physiological processes in animal models. These have been explored in animal models including Drosophila, C.elegans and rodents. In Drosophila RDGB is mainly enriched in the adult head [68,69], although it is also highly expressed in the embryonic nervous system, male and female germline [70]; the gene is primarily expressed very early during embryogenesis, during pupal development and young adults mainly in the nervous system (data curated at http://flybase.org/reports/FBgn0003218). Drosophila mutants for rdgB show defects in phototransduction and undergo retinal degeneration (reviewed in [10]) underscoring the role of this protein in regulating G-protein coupled PLC activation which is essential for phototransduction [71]. In addition rdgB mutants also show defects in olfactory transduction with abnormal electrical responses at the peripheral sensory organs [72,73]. Null alleles of rdgB are homozygous viable as adults [68,74].

In C. elegans the single Class IIA PITP, pitp-1 is expressed in the intestine and various neurons in the head region (https://wormbase.org/species/all/expr_pattern/Expr6405#021--10) and in the sensory neurons localizes to the presynaptic regions where it plays an essential role in neurotransmission and behavioural plasticity. Mutations in pitp-1 gene affects the chemosensory behaviour of the worms causing them to show reduced chemotactic attraction towards salt. Other sensory activities that were affected due to the mutations are the odour perception and detecting osmolality [14]. Interestingly these phenotypes could be rescued in the mutant worms by expressing just the PI transfer domain of the PITP-1 protein. This implied that the transfer of PI mediated by the PITP-1 is central to controlling the chemosensory responses of the worm. A recent study has also implicated pitp-1 in the behavioural response of worms subjected to hypoxia [75]. Overall, it appears that in worms Class IIA PITPs regulate neuronal function in worms.

In the mouse, each of the two genes encoding a Class IIA PITP-pitpnm-1(PITPNM1) and pitpnm-2 (PITPNM2) have been studied. Both genes are widely expressed although it has been reported that pitpnm-1 expression is enriched in the eye, inner ear and brain [12,76,77] whereas pitpnm-2 is primarily expressed in the retina and dentate gyrus. Of these pitpnm-1 when expressed in the Drosophila rdgB mutant is able to rescue its photoreceptor phenotype [78]. The physiological functions of pitpnm-1 and pitpnm-2 have been examined using mouse knockouts of these genes. A protein null allele of pitpnm-2, gave homozygous viable adults with no discernible phenotype [79] although an independent study using knockout mice reported that pitpnm-2 function is required for the activity of intrinsically light sensitive retinal ganglion cells under low light conditions [80]. By contrast the pitpnm-1 knockout, although homozygous viable and fertile, showed reduced plasma cholesterol levels and hypocalcemia as well as reduced leukocyte counts. Although pitpnm-1 is expressed in a clearly defined spatial and temporal pattern, detailed analysis of auditory function revealed no defects [77]. This lack of phenotypic effects has been attributed to redundancy between some defects in pitpnm-1 and pitpnm-2; a double knockout of both genes has not been studied and remains a priority to understand the relevance of Class IIA PITP in mammalian systems. Identification of tissues or cell types in mammalian systems will allow an analysis of the signalling function of Class IIA PITP in situ and will complement studies in cell culture models.

Some phenotypes have also been reported for Class IIA PITPs in other model organisms, although these have not been analysed in detail. In the zebrafish D.rerio two genes pitpnm-2 and pitpnm-3 are found. The expression pattern of these genes is not known; however morpholino mediated depletion of pitpnm-3 results in several arterio-venous malformations in the vasculature of the brain [42]. Likewise, in the flatworm S. mediterranea, depletion of the RDGB ortholog results in reduced regeneration and a curling phenotype [81]. Further analysis of these interesting but unexplained phenotypes may provide additional cellular models in the context of which Class IIA PITP function can be understood.

In the context of human biology, large scale expression analysis has defined the distribution of PITPNM1 and PITPNM2 (Figure 5). While both genes are widely expressed, in almost all tissues PITPNM1 is expressed at higher levels; the only expression is the thymus where PITPNM2 is expressed at much higher levels (Figure 5A). In the brain, both genes are widely expressed with PITPNM1 expressed at higher levels (Figure 5C). Most recently, the availability of single-cell gene expression data has allowed an assessment of Class IIA PITPs in nearly fifty cell types. While PITPNM2 was highly expressed in spermatids, whereas PITPNM1 is expressed at higher levels in most other cell types (Figure 5B). A number of studies indicate a role for PITPNM1 and PITPNM2 in the control of normal cell division and growth in human cells in the context of cancer. In the highly invasive human breast cancer cell line MDA-MB-231 and hepatic carcinoma cell lines, depletion of PITPNM1 shows attenuation of cell migration in trans-well chambers indicating that PITPNM1 is required for tumour cell migration. PITPNM1 depletion reduce the phosphorylation of AKT in response to EGF that results in reduced rates of cell migration. Similar observations were made in mouse lung metastasis, where in upon PITPNM1 depletion, there was reduction in the number of both metastatic and micro-metastatic lesions [82]. PITPNM3 and PITPNM1 have been identified as targets for transforming growth factor-β (TGFβ) interacting factors and control the growth of cancer cells [83]. In zebrafish in-vivo studies it is seen that downregulating PITPNM3 shows dilation and deformation of the basal communicating artery, a developmental process regulated by TGFβ signalling [42]. In pancreatic ductal adenocarcinoma, the chemokine CCL18 protein is elevated leading to altered cell adhesion and angiogenesis. These phenotypes seem to depend on PITPNM3 which has been proposed as a CCL18 receptor [84–87][88].

Figure 5. Expression pattern of PITPNM1 and 2 in humans.

Expression in different tissues (A) single cell types (B) and brain regions (C) from transcriptome analysis. (B) Y-axis represents the normalized expression value. Each bar represents a specific tissue.

PITPNM1 has been implicated in the infection and life-cycle of viruses in human cells. Studies have reported that depletion of PITPNM1 results in reduced infection of human cells with HCV and viral replication [89] and also in the egress of virions of HSV from human cells [90]. Some of these functions have been attributed to the ability of PITPNM1 in regulating phosphoinositide synthesis and virion trafficking. Downregulation of PITPNM2 has been implicated as a risk factor for multiple sclerosis and allergic reaction which are linked to rheumatoid arthritis [91]. PITPNM1 was up-regulated in the cartilages from patients with osteoarthritis compared to those from matched controls. In human adipose-derived stem cells (hADSC) from patients with rheumatoid arthritis, downregulation of PITPNM1 showed differentiation of cells into chondrogenic microspheres and thereby proposed to be used for stem cell based therapy [92]. Finally, based on analysis from the European population of 1000 Genomes Project, the PITPNM2 gene was identified as a candidate gene associated with schizophrenia [93] [94]. The gene is found to be located in locus that harbours multiple genes linked to Schizophrenia and bipolar disorder [95]. The signalling context and functional significance of these observations remains to be established using suitable models [96].

What is the function of the additional domains of Class IIA PITPs?

The Class IIA PITPs are proteins in which a PITP_d is present along with additional domains. The PITP_d of Class II PITPs have also been shown to act as PI transfer proteins when expressed and studied in vitro. Further two studies have noted that in the case of Drosophila RDGB [17,20] that reconstitution of rdgB mutants with only the PITP_d of this protein is sufficient to rescue key phenotypes. Likewise, in the case of Class IIA PITPs in C. elegans, the expression of just the PI transfer domain of the PITP-1 protein could restore normal olfaction and osmosensation in pitp1^-/- worms [14]. These findings raise the question of what the function of the additional domains of Class IIA PITPs might be?

Studies on Drosophila RDGB in photoreceptors summarised above show that each of the three key regions C-terminal to the PITP_d, the FFAT motif, DDHD and LNS2 domain when mutated result in mislocalization of the RDGB protein [63,65] and in mammalian cells, equivalent mutations in the FFAT motif [47] and LNS2 domain result in altered translocation of the PITPNM1 protein [48,50]. Thus it is possible that the most likely function of the additional domains of Class IIA PITPs is the accurate localization of these proteins and hence the PITP_d where presumably it can effectively transfer

PI during PLC signalling. Why then can the PITP_d alone rescue phenotypes in Class IIA PITP_d mutants? The answer to the conundrum has been proposed by careful quantitative studies in Drosophila photoreceptors. In this study [63], the authors found that while reconstitution with the PITP_d could rescue the light response as previously reported [17,20], this domain by itself could not rescue PIP₂ resynthesis following depletion by high intensities of light [63]. Rescuing the defect in PIP₂ resynthesis following stimulation with very bright light, that triggers high rates of PLC activity required either expression of full length RDGB, which is accurately localized to the contact site or by expression of the PITP_d at high protein levels [63]. Presumably under the latter condition, the effective concentration of the PITP_d at the contact site is much greater allowing sufficient lipid transfer to rescue the defect in PIP₂ resynthesis. Additional studies that allow the PITP_d to be concentrated at the ER-PM MCS, independent of the C-terminal domains of RDGB will provide insights into whether this model is accurate.

The sufficiency of only the lipid transfer domain to carry out physiological functions, has been reported not only for PITPs, but also in other groups of lipid transfer proteins. The N-terminal domain of the lipid transfer protein Atg2 was found to be sufficient for rescuing stalled autophagosome biogenesis in Atg2^-/- cells. This was directly dependent on the ability of the domain to transfer glycerophospholipids at the ER-autophagosome interface [97]. Likewise, members of the ORP family and their yeast homologues (Osh proteins) consist of proteins that either have only the lipid transfer OSBP-related domain (ORD) or also have additional domains (PH domain, FFAT motif, ankyrin repeats) (reviewed in [98]). Although both single or multi-domain ORP or Osh proteins are able to transfer sterols at contact sites, a recent study has shown that overexpression of ORP2, which only has the ORD domain, can restore PM cholesterol transfer in HeLa cells deleted of the multidomain ORP1 protein [99]. However, in case of CERT protein, while the lipid transfer START domain alone can carry out ceramide transfer across artificial phospholipid bilayers [100], the Golgi targeting PH domain and the ER targeting FFAT motif are required for efficient ceramide transfer at ER-Golgi contact sites [24]. Thus an emerging principle might be that while the lipid transfer domain of a multi-domain protein might be sufficient to rescue phenotypes, it may not be able to do so under some conditions, where for example ongoing cellular physiology demands that the lipid transfer activity be concentrated at an MCS. The identification of suitable in vivo mammalian signalling systems will help test this idea.

Future Perspectives

Emerging studies from genomic sequencing studies indicate that Class IIA PITPs are distributed widely with diverse domain architectures in which a PITP_d is encoded in along with one or more additional domains. Experimental studies of these in tractable genetic model systems indicate that several of these domains play an important role in the accurate localization of the PITP_d, for example at ER-PM MCS. This may therefore be the raison d’être for the evolution of multidomain PITPs where additional domains help to position the PITP_d at specific sub-cellular locations such as individual MCS to mediate localized lipid transfer reactions. Once localized accurately, the PITP_d can mediate localized lipid transfer at these MCS. It is presently unclear whether, once correctly localized, the lipid transfer reaction of a multidomain PITP is constitutive or regulated. The mechanisms by which the PI transfer reactions of the PITP_d are regulated remain to be determined. In addition, the identification of additional in vivo model systems and cell types where the function of Class IIA PITPs can be mapped to ongoing physiological processes will hugely assist further analysis of this class of proteins.