The RNA-binding protein SERBP1 interacts selectively with the signaling protein RACK1

Abstract

The RACK1 protein interacts with numerous proteins involved in signal transduction, the cytoskeleton, and mRNA splicing and translation. We used the 2-hybrid system to identify additional proteins interacting with RACK1 and isolated the RNA-binding protein SERBP1. SERPB1 shares amino acid sequence homology with HABP4 (also known as Ki-1/57), a component of the RNA spicing machinery that has been shown previously to interact with RACK1. Several different isoforms of SERBP1, generated by alternative mRNA splicing, interacted with RACK1 with indistinguishable interaction strength, as determined by a 2-hybrid beta-galactosidase assay. Analysis of deletion constructs of SERBP1 showed that the C-terminal third of the SERBP1 protein, which contains one of its two substrate sites for protein arginine N-methyltransferase 1 (PRMT1), is necessary and sufficient for it to interact with RACK1. Analysis of single amino acid substitutions in RACK1, identified in a reverse 2-hybrid screen, showed very substantial overlap with those implicated in the interaction of RACK1 with the cAMP-selective phosphodiesterase PDE4D5. These data are consistent with SERBP1 interacting selectively with RACK1, mediated by an extensive interaction surface on both proteins.

1. Introduction

The WD-repeat β-propeller protein RACK1 interacts with an expanding list of protein “partners” and therefore has numerous and diverse functions in cells (see refs. (1–3) for reviews). First isolated as a partner of specific protein kinase C isoforms, for which it was named receptor for activated C-kinase (4–8), RACK1 has also been shown to interact with a considerable number of other signal transduction proteins, including the cAMP-specific phosphodiesterase PDE4D5 (9–17), the tyrosine kinase oncoproteins SRC (18–21), FLT1 (22), and FAK (23), protein serine phosphatase PP2A (24–26), the MAPKs JNK1 and ERK1/2 (27), VARP (28), phospholipase C (29), the NMDA receptor (30–33), cytokine and growth factor receptors (34–38), G_sβγ (39–42), selected 14.3.3 isoforms (43) and likely other signaling proteins (44–47). RACK1 also interacts with members of complexes that play important roles in protein degradation, such as the Cullin-2 based E3 ubiquitin ligases (48;48;49), as well as certain enzymes of unclear cellular function, such as inositol-requiring enzyme 1 (IRE1; refs, (50)). RACK1 also may interact with elements of the cytoskeleton, including β-integrins (51–56) and β-actin (21;21;57). In some of these contexts, RACK1 has been considered to be a scaffold protein, i.e., a protein that mediates the association of two other, separate proteins and thereby allows them to interact functionally (58).

Numerous investigators have shown that RACK1 has several important, possibly related, roles in mRNA splicing and in translation. RACK1 interacts directly with specific components of the 40S ribosomal subunit (specifically, the 18S rRNA and ribosomal proteins rpS16e, rpS17e, and rpS3e [new names: uS9, eS17 and uS3, respectively]; refs. (59;60)). It has been suggested that RACK1 may recruit signaling molecules, such as PKC and JNK, to the ribosome (61;62), although the precise function(s) of ribosomal RACK in complex eukaryotes remain an active area of investigation (3).

RACK1 also interacts with the HABP4 protein (synonym: Ki-1/57), which is involved in mRNA metabolism (63;64)). HABP4 interacts with the chromo-helicase-DNA-binding domain protein 3 (CHD-3), a nuclear protein involved in chromatin remodeling and transcription regulation (65). HABP4 is also a substrate of protein arginine N-methyltransferase 1, (PRMT1), which methylates two discreet regions on HABP4, both approximately 21 amino acids in length and consisting of repeats of the sequence RGG and/or RGR (66;67). HABP4 plays an important role in mRNA splicing, which is mediated by its interaction with the splicing proteins hnRNPQ and SFRS9 (68;69). It may also play a role in mRNA translation, as one group has shown that it interacts with ribosomal protein L38, although the physiologic significance of this finding is not clear (70). One group has also shown that RACK1 interacts with the polyA-binding protein LARP4 (71).

SERBP1 (synonyms: CGI-55, CHD3IP, HABP4L, PAI-RBP1), the focus of the present study, is an mRNA-binding protein (72). SERBP1 binds to a 134-nucleotide cyclic nucleotide-responsive sequence (CRS) located near the 3′ end of several mRNAs, most notably that encoding type-1 plasminogen activator inhibitor (PAI-1; refs. (72–74)). Studies in a number of cell lines that shown that elevation of intracellular cAMP levels markedly increases the rate of degradation of PAI-1 mRNA, which has been shown to be mediated by SERBP1 (73;75). Like HABP4, SERBP1 binds to CHD-3, which for SERBP1 involves two separate regions, located in its N- and C-termini (76;77). SERBP1 has also been shown to bind to the RNA splicing component U2AF (78). These data implicate SERBP1 in several nuclear functions, separate from its action on mRNA. Also like HABP4, SERBP1 is a substrate for PRMT1-mediated arginine methylation, which may regulate the nuclear-cytoplasmic distribution of the protein (79;80). SERBP1 and HABP4 target similar mRNA populations in cells, clearly suggesting some overlap in their physiological functions (81). SERBP1 also interacts with progesterone receptor membrane component 1 (PGRMC1), linking it to progesterone-signaling pathways (82).

Given the similarities between SERBP1 and HABP4 that have been reported to date, it would not be unexpected that they might share protein partners. As discussed above, both interact with CHD-3 (65;76;77). In the present study, we report that SERBP1, like HABP4, interacts with RACK1 in 2-hybrid assays. We also show that the interaction of SERBP1 with RACK1 requires an extensive region located in the C-terminal third of the SERBP1 protein. We also show that a large surface on RACK1, which overlaps substantially that needed for RACK1 to interact with PDE4D5, is necessary for it to interact with SERBP1. Our data provide additional impetus to search for additional functional roles of RACK1 in nuclear processes, such as chromatin remodeling, and possibly in mRNA splicing and stability.

2. Materials and Methods

2.1. Yeast 2-hybrid analyses

Yeast 2-hybrid techniques identical to those used previously by us were used to identify and analyze protein-protein interactions (10;14;17;83;84). In all experiments, one of the two interacting proteins was expressed as “bait”, in pLEXAN (a derivative of BTM116; refs. (10;85)), as a LexA DNA-binding domain fusion. The other interacting protein was expressed as “prey”, in pGADN (a derivative of pGAD.GH; ref. (85)), as a Saccharomyces cerevisiae GAL4 activation-domain protein fusion. For all experiments, the oncoproteins RAS and RAF1 (ref. (86)) were employed as standards. All proteins were targeted to the nucleus. Plasmids encoding the appropriate fusion pair were transfected into the S. cerevisiae strain L40 (ref. (86)). Yeast colonies expressing the fusions were obtained by selecting for the prototrophic markers encoded by the bait and prey plasmids. For filter β-galactosidase assay, colonies were streaked out as patches, transferred to nylon membranes (GE Healthcare), permeabilized with liquid nitrogen, and then incubated with 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (Xgal), as described previously (10;14;83;84). Positive patches were identified by their turning a blue color. Except where indicated otherwise, all experiments shown here have been repeated at least twice, in at least two different yeast transformants, with identical results.

2.2. Yeast 2-hybrid screen

The yeast 2-hybrid screen used, as bait, the full open reading frame of human RACK1 (GenBank M24194) cloned into the NotI site of pLEXAN, to express it as a LexA DNA-binding domain fusion (i.e., pLexARACK1). pLexARACK1 was used to screen a cDNA library derived from the HeLa human cervical cancer cell line S3, cloned into the EcoRI and XhoI sites of the pGAD.GH vector; we have described this library previously (10). All interactions were evaluated in the S. cerevisiae strain L40 (ref. (86)). Positives were isolated by their ability to grow as colonies on plates lacking histidine, without the addition of 3-amino-triazole. Positive colonies were streaked as patches on plates selecting only for prototropic markers encoded by the bait and prey plasmids and then subjected to the filter β-galactosidase assay described above. The prey plasmid was then isolated from HIS+, β-gal+ colonies and sequenced, using the Sanger sequence method, on an ABI 3870 sequencer. DNA sequence data was analysed using the basic local alignment sequence tool (BLAST) at www.ncbi.nlm.nih.gov/blast/. All positive clones were verified for their ability to interact with pLexARACK1 by re-transforming them back into the L40 strain, with pLexRACK1, and repeating the filter β-galactosidase assay. All prey were also tested for non-specific interactions by testing them against pLEXAN (empty vector) and a panel of test bait, which included several transcription factors and several kinases; no non-specific interactions were noted.

3. Results

3.1. Identification of SERBP1 and ribosomal proteins as RACK1-interacting partners in a 2-hybrid screen

Previously, we isolated RACK1 in a 2-hybrid screen where PDE4D5 was used as bait (10). To identify additional proteins that interact with RACK1, we used RACK1 as bait to screen a HeLa-derived 2-hybrid library, using procedures described in Materials and Methods. Of the approximately 500 positive (i.e., HIS+ β-gal+) colonies obtained in the screen, the prey plasmids from 192 colonies were isolated and analysed in detail by restriction enzyme mapping and sequencing. The 192 positive prey encoded GAL4 fusions with SERBP1 (148 isolates), PDE4D5 (2 isolates), ribosomal protein S13 (1 isolate; rpS13e, new name: uS15), ribosomal protein S17 (4 isolates, rpS17e, new name: eS17, ref. (59)), and 22 isolates, generally seen only 1–3 times each, of uncertain significance. Fifteen isolates encoded non-physiological fusions (i.e., the physiological open reading frame [ORF] of the insert cDNA was not in frame with that of GAL4).

3.2. SERBP1 isoforms define regions required for its interaction with RACK1

The human SERBP1 gene encodes a large number of protein isoforms, by transcripts that are generated by alternative mRNA splicing. The longest known isoform has 474 amino acids (Fig. 1a). In our screen, we isolated cDNAs that encoded each of the 402-, 393- and 387-amino acid isoforms (Fig. 1b). Since some of the isolates that encoded the 393- and 387-amino acid isoforms isolates contained inserts that were truncated at the 5′ ends of their ORFs, it is possible that their cDNAs were derived from mRNAs encoding the 450- and 456-amino acid isoforms (Fig. 1b). The regions of sequence FVSSFSHYS and ESPKYIQKQISYNYS, which are encoded by variably-spliced exons and are of unknown physiological significance, are not present in at least one of 402-, 393- and 387-amino acid isoforms (Fig. 1b). Since all of these isoforms were capable of interacting with RACK1, these regions of SERBP1 clearly are not essential for it to bind to RACK1. Conversely, at least some of the “core” regions (i.e., those found in all the known isoforms) are essential for the interaction.

Fig. 1. SERBP1 protein isoforms and key functional features.

1a. Amino acid sequence of the longest known (474-amino acid) human SERBP1 isoform. The single-letter amino acid code is used. Regions that are encoded by alternatively-spliced exons and therefore not present in some of the shorter isoforms are shown in gold (i.e., the 63-amino-acid N-terminal region), red, blue or green (see Fig. 1b for more details). The two regions that are targets for arginine methylation by PRMT1 (typically, repeats of sequence RGG or RGR) are shown in grey. The C-terminal region required for interaction with RACK1 is underlined.

1b. Human SERBP1 variants. BLAST searches of amino acid sequence data in GenBank using the 474-amino acid SERBP1 isoform as a query yielded 9 additional isoforms, all encoded by SERBP1 and generated by alternative mRNA splicing. Each of these isoforms differs from the 474-amino acid isoform in that it lacks specific blocks of sequence, as shown.

3.3. Deletion constructs define an extensive area on SERBP1 required for its interaction with RACK1

To define further the region(s) on SERBP1 required for it to interact with RACK1, a series of SERBP1 deletion constructs were prepared in pGADN (see Materials and Methods) and tested in 2-hybrid filter β-galactosidase assays for their ability to interact with pLexARACK1 (Fig. 2). These data show that amino acids 354 to 474 (using coordinates based on the full-length isoform, Fig. 1a) of SERBP1 are necessary and sufficient for it to interact with RACK1. This region of SERBP1 has 65 of 121, or 53.7%, amino acid identity with the corresponding region of HABP4 (Fig. 3). It contains one of the arginine methylation sites (Figs. 3 and ​1a) but also extensive areas of undetermined function. It has no obvious amino acid sequence homology to the regions of PDE4D5 required for its interaction with RACK1 (10;11;14–16).

Fig. 2. Determination of the region of SERBP1 required for it to interact with RACK1.

Yeast 2-hybrid experiments were performed using RACK1 as bait and various regions of the SERBP1 cDNA as prey. For bait, RACK1 was expressed in S. cerevisiae L40 as a LexA fusion (i.e., as pLexARACK1). For prey, the SERBP1 clones were expressed as GAL4 fusions (i.e., in pGADN containing full-length SERBP1, or various deletion subclones thereof). The strength of the Interactions were determined by a filter β-galactosidase assay. Interactions that were indistinguishable from that of RACK1 and full-length SERBP1 were scored as positive (symbol to the right of each construct); those that were indistinguishable from the corresponding empty vectors were scored as negative. All data were obtained in a least 3 independent experiments, using separate transformations. Discrete regions of amino acid sequence encoded by each construct are shown, using the same color scheme as in Fig. 1a.

Fig. 3. Alignment of the amino acid sequence of the long SERBP1 isoform with that of the longest known HABP4 isoform.

BLAST searches of human amino acid sequence data in GenBank using the 413-amino acid HABP1 isoform as a query yielded at least 2 additional isoforms, all encoded by HABP4 and generated by alternative mRNA splicing. The longest HABP4 isoform (413 amino acids) was then aligned with the longest SERBP1 isoform (474 amino acids) using BLASTP at www.ncbi.nlm.nih.gov/blast/. The alignment is shown using the single letter amino acid code. Dashes indicate gaps inserted by the alignment program, amino acid identities are indicated by the presence of that amino acid in the line between the 2 sequences, and related amino acids are indicated by the (+) sign. The region of SERBP1 essential for its interaction with RACK1 (amino acids 354 to 474, Fig. 2) is underlined.

3.4. Specific amino acids in RACK1 are required for its interactions with both SERPB1 and PDE4D5

RACK1 is a WD-repeat, β-propeller protein that has extensive amino acid sequence homology to the trimeric G-protein G_sβ subunit (refs. (87–90); see ref. (12) for an alignment). The crystal structure of RACK1 from Arabidopsis thaliana demonstrates its β-propeller structure (91) and in particular shows the presence of highly conserved surface residues that could play critical roles in protein-protein interactions (91). Human RACK1 has high amino acid sequence homology to A. thaliana RACK1 (92) and its solution structure demonstrates high structural homology to A. thaliana RACK1, as well as overall similarity to G_sβ and other β-propeller domains (64;93). Therefore, it would be of interest to determine whether amino acid side chains in RACK1 that are critical for it to interact with one partner also play a role in interacting with other partner(s). We used 2-hybrid approaches to test this hypothesis for RACK1.

Previously, we have used a “reverse” 2-hybrid approach to identify amino acids of RACK1 essential for it to interact with PDE4D5 (12). Briefly, we used PCR with a low-fidelity polymerase to perform random mutagenesis of a cDNA encoding WD-repeats 5 through 7 of RACK1 and then used a 2-hybrid assay to isolate mutants that attenuated the interaction of the WD-repeats with PDE4D5. With this approach, we identified 11 amino acids in RACK1 that, when individually mutated, blocked its interaction with PDE4D5. Molecular modeling showed that all 11 mutations mapped to a single surface on the RACK1 protein, presumably a “face” necessary for it to interact with PDE4D5 (12).

We therefore tested the effects of 8 of these 11 mutants on the ability of RACK1 to interact with SERBP1, using a 2-hybrid assay (Fig. 4). The results show that a number of the mutations (e.g., H191R, I236T, F241S, C249R, W259R, and W310R) significantly attenuated the interaction of RACK1 with both PDE4D5 and SERBP1. The N237D mutant affected the interactions with either protein to a roughly similar extent (although to a lesser extent than other mutations), while the C288G mutant attenuated the interaction with SERBP1 much more than it did with PDE4D5. These results are consistent with the concept that a discrete set of amino acids on one surface of RACK1 is necessary for it to interact with multiple partners, with additional amino acids contributing to the specificity of interactions with individual partners, such as PDE4D5 or SERBP1. Our approach cannot rule out the possibility that that some of the mutations could destabilize RACK1 generally, rather than contribute directly to the binding surface.

Fig. 4. Analysis of mutations in RACK1 that attenuate its interaction with SERBP1 and PDE4D5.

Various mutations in RACK1 that attenuated its interaction with PDE4D5 were created by random mutagenesis and identified in a reverse 2-hybrid screen, as we have described previously (12). These mutations were then tested for their ability to attenuate the interaction of RACK1 with SERBP1, as described in Materials and Methods. Yeast S. cerevisiae L40 cells in any given row contained the same bait and those in any given column contained the same prey. Positive interactions, assessed with a filter β-galactosidase assay, produce blue patches, while negative interactions produce amber patches. Controls are vectors alone. Standards are the oncoproteins RAS^V12 and RAF1. This figure shows data typical of experiments performed at least 3 times.

4. Discussion

In this study, we show that RACK1 interacts with SERBP1, a protein implicated in diverse functions, including transcription, genome integrity, and the regulation of mRNA stability. Our observation should stimulate additional investigation into RACK1’s biophysical and biochemical role in these processes. Given that RACK1 has numerous protein partners in cells, it also potentially links SERBP1 to some of RACK1’s other partners and to their physiological roles. It also provides insights into how RACK1 interacts with its partners and, in particular, the precise structural features on RACK1 that provide specificity to its interactions and determine the number of partners with which it can interact at any given time.

SERBP1 has homology to several related proteins, most notably HABP4 (synonym: Ki-1/57), a protein that also interacts with RNA and has several potential nuclear functions. However, SERBP1 differs from HABP4 in that it interacts avidly with highly-specific regions of sequence in the 3′ termini of several mRNAs, regulating their stability in response to specific stimuli (72–74). Both SERBP1 and HABP4 have 2 sites for arginine methylation by PRMT (Figs. 1 and ​3) and interact with CHD-3, a nuclear protein involved in chromatin remodeling and transcription regulation (65;76;77). These common functions may provide an explanation for their significant amino acid sequence homology (Fig. 3); however, the precise region(s) of the 2 proteins responsible for many of their functions are not known, especially in the absence of structural clues. HABP4 is felt to be an intrinsically-disordered protein (94); to the best of our knowledge, there is no structural data on SERBP1.

We have shown that an extensive region in SERBP1 (amino acids 354 to 474, using co-ordinates based on the full-length isoform, Figs. 1a and ​2) is necessary for it to interact with RACK1; since this region has substantial homology to the corresponding region of HABP4 (Fig. 3) it is a reasonable hypothesis that this region of HABP4 is also essential for it to interact with RACK1. More generally, it appears that many of RACK1’s partners employ a large region(s) of their respective proteins in their interaction with RACK1; for example, we and our collaborators have identified 2 extensive regions in PDE4D5 that are required for it to interact with RACK1 (10–16;95). The large size of the interacting regions of PDE4D5 and RACK1 is consistent with the high avidity of their interaction; in this regard, we and our collaborators have shown previously that the interaction between RACK1 and PDE4D5 has an affinity in the low nanomolar range (10). Although there is no obvious amino acid sequence homology between the RACK1-interaction regions of PDE4D5 and that of SERBP1 (and, by extension, that of HABP4), it is possible that similarities in the overall protein conformation of these regions might be apparent upon appropriate structural study.

Given the large region(s) on RACK1’s partners that seem to be essential for many of them to interact with RACK1, a reasonable hypothesis would be that a corresponding large region on RACK1 is also essential for the interaction. We have shown previously that a large surface (WD-repeats 5 through 7) on RACK1 is essential for it to interact with PDE4D5 and have identified a number of amino acids on RACK1 that, when individually mutated, markedly attenuate their interaction (12). In the present study, we have shown that many of these RACK1 mutants also attenuate its interaction with SERBP1 (Fig. 4). These observations suggest, but do not prove, that a large region of RACK1 interacts with SERBP1 and that this region overlaps substantially with the region of RACK1 that is essential for it to interact with PDE4D5. Potential alternative explanations for our data could be that some of the mutations alter the conformation of RACK1 generally (rather than affecting the interaction face itself), or destabilize the protein in cells. Finally, we cannot exclude the possibility that some of our observations reflect artifacts of the yeast 2-hybrid system; however, our S. cerevisiae-based system has, in a very large number of cases, been shown to be a reasonably physiologic system for these studies.

One important issue in interpreting data on the interaction of RACK1 and its partners is the stoichiometry of the complex. In our 2-hybrid studies, RACK1 and its partners each interact as monomers. Some studies have suggested that RACK1 is dimeric in cells (96–99), while other data (100) are more consistent with it existing in solution primarily as a monomer. It is clearly monomeric in the ribosome (101). If RACK1 exists primarily as a monomer in cells, then our mutagenesis data are consistent with the surface on WD repeats 5 through 7 directly interacting with PDE4D5 and SERBP1; however, if RACK1 exists as a dimer, then the mutations could disrupt RACK1 dimerization and not affect the actual interaction site(s) with its partners. Additional structural or interaction data will be essential to distinguish these possibilities. Finally, we cannot exclude the possibility that many of RACK1’s partners could also dimerize or multimerize; in this regard, we, our collaborators, and a number of other groups have clearly shown that PDE4D5 can dimerize (83;84;102–105), although the regions of PDE4D5 that are essential for dimerization are quite separate from those necessary for it to interact with RACK1 (see ref. (17) for a more complete discussion).

Our 2-hybrid screen data also provide some additional insight into the interaction of RACK1 with the small (40S) ribosomal subunit. A large number of biochemical studies have shown previously that RACK1 interacts directly with the 40S subunit. A structural study employing X-ray diffraction has shown that RACK1 interacts directly with rpS16e, rpS17e, and rpS3e [new names: uS9, eS17 and uS3, respectively, ref. (59)] and the 18S rRNA of the human 40S subunit (101). In our 2-hybrid screen, we isolated fusions encoding the rpS13e and rpS17e proteins (new names uS15 and eS17, respectively; ref. (59)). Our data confirm the importance of rpS17e in the interaction of RACK1 with the 40S subunit; however, the significance of our rpS13e isolate is unclear, as this protein is located some distance away from RACK1 (101). The functional significance of the interaction of RACK1 with the ribosome, especially in complex eukaryotes, remains uncertain (3).

Collectively, the studies reported here provide additional impetus for the study of the role(s) of RACK1 in nuclear processes, such as gene expression, RNA processing and the maintenance of genome integrity, and also in ribosomal composition and the regulation of translation. They also provide additional insights into the interaction face(s) that mediate the interaction of RACK1 with its partners. Further structural and functional experiments should provide insights into the physiological significance of our observations. One possible experiment would be to over-express (or, alternatively knock down) RACK1 in an assay of SERBP1 action on PAI1 mRNA stability.

5. Conclusion

1. We have identified SERBP1, ribosomal protein S17 (rpS17e) and potentially ribosomal protein S13 (rpS13e) as partners for RACK1 in a 2-hybrid screen.

2. The interaction of RACK1 and SERBP1 is mediated by large region(s) on both proteins, located in the C-terminal third of SERBP1 and a large “face” on one side of WD-repeats 5 through 7 of RACK1, respectively.

3. These data provide additional impetus for investigation of the role(s) of RACK1 in nuclear processes, such as gene expression, RNA processing and the maintenance of genome integrity, and also in ribosomal composition and the regulation of translation.

Highlights.

1. SERBP1 and ribosomal protein S17 (rpS17e) are partners for RACK1 in 2-hybrid analyses.

2. The interaction of RACK1 and SERBP1 is mediated by large region(s) on both proteins.

3. The interaction region on SERBP1 is located in the C-terminal third of the protein.

4. The interaction region on RACK1 includes its WD-repeats 5 through 7.

5. The data provide evidence for roles of RACK1 in nuclear processes and RNA processing.

Abbreviations

RACK1

Receptor for activated C-Kinase

PRMT1

Protein arginine methyl transferase 1

CHD-3

Chromo-helicase-DNA-binding domain protein 3

PGRMC1

Progesterone receptor membrane component 1