Activation of Hypoxia-Inducible Factor Signaling Modulates the RNA Protein Interactome in Caenorhabditis elegans

Summary

The cellular response to hypoxia is crucial to organismal survival, and hypoxia-inducible factors (HIF) are the key mediators of this response. HIF-signaling is central to many human diseases and mediates longevity in the nematode. Despite the rapidly increasing knowledge on RNA-binding proteins (RBPs), little is known about their contribution to hypoxia-induced cellular adaptation. We used RNA interactome capture (RIC) in wild-type Caenorhabditis elegans and vhl-1 loss-of-function mutants to fill this gap. This approach identifies more than 1,300 nematode RBPs, 270 of which can be considered novel RBPs. Interestingly, loss of vhl-1 modulates the RBPome. This difference is not primarily explained by protein abundance suggesting differential RNA-binding. Taken together, our study provides a global view on the nematode RBPome and proteome as well as their modulation by HIF-signaling. The resulting RBP atlas is also provided as an interactive online data mining tool (http://shiny.cecad.uni-koeln.de:3838/celegans_rbpome).

Graphical Abstract

Highlights

• •RNA interactome capture in wild-type C. elegans and vhl-1 loss-of-function mutants
• •Identification of 1,354 nematode RBPs, 270 of which can be considered novel RBPs
• •The modulation of the RBPome by vhl-1 is primary explained by differential RNA-binding
• •The resulting RBP atlas is provided as an interactive online data mining tool

Biological Sciences; Molecular Biology; Molecular Interaction; Molecular Network; Integrative Aspects of Cell Biology; Proteomics

Introduction

RNA-binding proteins (RBPs) play an important role in cell biology, regulating expression, stability, and localization of all known RNA species (Halbeisen et al., 2008, Hasan et al., 2014, Lee and Lykke-Andersen, 2013, Wilkie et al., 2003). The importance of these proteins is underlined by the increasing body of evidence linking several hereditary diseases, developmental disorders, and cancer with mutations in genes encoding RBPs (Fredericks et al., 2015, Kechavarzi and Janga, 2014, Lukong et al., 2008). By using sequence motifs, many RBPs could be predicted by their classical RNA-binding domain (RBD)—e.g. RRM, ZINC finger domain, or PUF domain—and studied individually (Hall, 2005, Query et al., 1989, Zhang et al., 1997). However, their global characterization became possible only in the last years and was facilitated by novel techniques such as RNA interactome capture (RIC) combined with mass spectrometry to identify proteins co-precipitating with RNA (Baltz et al., 2012, Castello et al., 2012). As a result, the list of known and putative RBPs has been increasing in size and complexity across species with more than 2,000 proteins showing an interaction with RNA (Hentze et al., 2018). Interestingly, a significant number of these proteins lack classical RBDs—a finding that was the basis to the term enigmRBPs (Beckmann et al., 2015), showing that mere prediction by amino acid sequence and domain is not sufficient for an exhaustive identification of proteins capable of RNA binding. RBPs can be regulated not only on the level of protein abundance but also by modulation of their association with RNA, e.g. through differential RNA-binding capacity of the protein itself or availability of the actual RNA interaction partners. Consequently, the comparison of different environmental conditions and genetic mutants is crucial to gain a better understanding of the RBPome landscape (Jiang et al., 2014). As an example, this aspect was addressed regarding the induction of apoptosis in the only Caenorhabditis elegans (C. elegans) RIC dataset published to date (Matia-Gonzalez et al., 2015). In cell culture, a recent study from our group found differences in RBP-binding to RNA upon exposure to hypoxia (Ignarski et al., 2019). Key genes involved in sensing hypoxia are the hypoxia-inducible transcription factors (HIFs). HIF-1 is a heterodimer composed of an oxygen-sensitive α subunit and a constitutively expressed β subunit. HIF-1α is regulated by oxygen-dependent proline hydroxylation. Upon hydroxylation of HIF-1α, pVHL as the substrate recognition subunit of an E3 ubiquitin ligase binds to HIF-1α resulting in its proteasomal degradation. Upon hypoxia the HIF-1α subunit gets stabilized and translocates to the nucleus where it can exert its transcriptional activity as a heterodimeric transcription factor (Kaelin, 2005, Luo et al., 2014). Stabilization of HIF-1α can also be gained by a loss-of-function of the VHL gene, which is the basis to von Hippel-Lindau disease, an autosomal-dominant multitumor syndrome (Maxwell et al., 1999). A small number of studies have shown a potential role of specific RBPs on HIF-signaling (Cho et al., 2015, Galban et al., 2008). Yet, the global impact and regulation of RBPs in this pathway has not been sufficiently delineated. HIF-signaling and its regulation through pVHL are highly conserved with activation of HIF mediating longevity in nematodes (Mehta et al., 2009, Muller et al., 2009). Consequently, we chose this model organism to examine the impact of genetic activation of HIF-signaling on the RBP landscape to complement previous data at organismal level. In the study at hand, we performed RIC in wild-type (WT) and vhl-1(ok161) (from now on referred to as vhl-1) loss-of-function worms. We identified more than 1,000 bona-fide RBPs in WT and compared these results with vhl-1 mutants to identify hypoxia-signaling induced changes in the nematode RBPome. This analysis was then combined with the whole proteome quantification in both nematode strains to distinguish changes in RBP abundance from differential binding events. An interactive online interface to visualize and interrogate these datasets is provided at http://shiny.cecad.uni-koeln.de:3838/celegans_rbpome. Taken together, our study provides the first global atlas of HIF-signaling-induced changes in the nematode RBPome.

Results

Global Analysis of RBPs in WT and vhl-1 Mutant Worms

We performed RIC using UV-C crosslinking and oligo(dT)-bead-based RNA pulldown coupled with mass spectrometry (MS) to obtain a global view on the nematode RBPome in WT and vhl-1 mutant worms (Figure 1A). Protein concentration measurements of lysates obtained from 500 worms in both strains revealed a significantly lower protein yield in the vhl-1 mutant strain (Figure S1A). In line with this result, the RNA concentration obtained from 500 worms from both strains also showed a significantly lower RNA yield in the vhl-1 mutant strain (Figure S1B). In order to rule out that this would influence the results of the RIC we also measured the RNA concentration upon pooling the three eluates resulting from oligo(dT)-bead-based pulldown. These measurements revealed a similar amount in both strains and conditions (crosslinked and non-crosslinked) (Figure S1C). As expected, a higher amount of protein was co-precipitated in crosslinked samples from both strains (compared with the non-crosslinked samples) as shown by SDS-PAGE and silver staining (Figure 1B). The samples were then analyzed by MS. Principal component analysis (PCA) and hierarchical clustering of these data revealed that two of the non-crosslinked samples were outliers due to an unexpected high number of proteins identified (one sample for each genotype, Figures S1D and S1E, see also the section on Limitations of the Study). Consequently, we excluded these two samples from further analyses. Reanalysis of the remaining samples showed a clear separation of the crosslinked samples versus the non-crosslinked in the PCA (Figure 1C). Similar results were obtained by hierarchical clustering (Figure S1F). We measured a total of 2,473 proteins co-precipitating with RNA in WT worms and 2,219 proteins in the vhl-1 mutant, respectively. 721 (WT) and 530 (vhl-1 mutant) proteins were significantly enriched (student's ttest; FDR<0.05) in crosslinked over non-crosslinked samples (red dots in Figure 1D and blue dots in Figure 1E). Of note, we found an additional 371 (WT) and 305 (vhl-1 mutant) proteins that were exclusively detected in crosslinked samples but never detected in non-crosslinked samples.

Figure 1.

Identification of C. elegans RBPs Using RNA Interactome Capture

(A) Schematic overview of the RNA interactome capture protocol. Worms grown in liquid culture were UV-C (254 nm) irradiated; a non-irradiated sample was used as control. RNA-protein complexes were captured using oligo(dT) beads and analyzed by mass spectrometry after treatment with RNase I and Benzonase.

(B) Protein samples from WT and vhl-1(ok161) mutant worms were analyzed by SDS-PAGE and silver staining. Input and eluate of both crosslinked and non-crosslinked (−) samples (+) were directly compared. The band corresponding to RNase I and benzonase is indicated by an arrow. M: molecular weight marker (kDa).

(C) Principal component analysis (PCA) of the RIC mass spectrometry data (on the basis of iBAQ intensities). WT samples are indicated in red and vhl-1(ok161) mutants in blue. Crosslinked samples are indicated by circles and non-crosslinked samples by squares.

(D) Volcano plot depicting the t-test comparison of protein abundance in the crosslinked and non-crosslinked RIC dataset of WT worms. x-axis: log₂ difference; y-axis: corresponding -log₁₀ p-values. Seven hundred twenty-one significantly enriched proteins are shown in red (FDR<0.05). Proteins not reaching significance are shown in gray. FC: fold change; +CL: crosslinked; -CL: non-crosslinked; vs: versus.

(E) Corresponding volcano plot for the vhl-1(ok161) worm RIC dataset [for details see (D)]. Five hundred thirty significantly enriched proteins are shown in blue (FDR<0.05).

Identification of RNA-associated Proteins in WT and vhl-1 Mutant Worms

In order to classify the proteins identified by levels of confidence we defined two different classes of RBPs. Proteins detected three times in the crosslinked samples in either genotype and never measured in the non-crosslinked samples are considered as class I RBPs. These proteins are not depicted in the volcano plots, as they do not have intensity values for the non-crosslinked samples (Figures 1D and 1E), but a list is provided in the online repository (“RBPome” tab, class I pulldown menu, http://shiny.cecad.uni-koeln.de:3838/celegans_rbpome) and in Table S1. Proteins with an FDR lower than 0.05 for enrichment in crosslinked samples (student's ttest) are defined as class II RBPs. Proteins not reaching the criteria for either class I or class II in our study that had been identified as RBPs in previous RIC experiments are defined as “other RBPs” (Hentze et al., 2018, Ignarski et al., 2019, Queiroz et al., 2019, Tamburino et al., 2013, Trendel et al., 2019, Urdaneta et al., 2019). All remaining proteins are summarized under the term “no evidence.” We found that 45% of the proteins co-precipitated with RNA in WT and 38% of the proteins in the vhl-1 mutant fall into either class I or class II (Figure 2A, “RBPome” tab http://shiny.cecad.uni-koeln.de:3838/celegans_rbpome and Table S1). In addition, we found that 788 proteins in WT and 963 proteins in vhl-1 mutants were classified as “other RBPs,” whereas the remaining 593 (WT) and 421 proteins (vhl-1) belong to the “no evidence” group (Figure 2A). For a general view on the nematode RBPome, the lists of proteins identified in either strain were pooled for enrichment analyses. For this purpose, the class II was assigned to RBPs classified differently between WT and vhl-1 mutant (e.g. class I in vhl-1 and class II in WT). Proteins classified as class I or class II in one strain and assigned to “other RBPs” or “no evidence” in the other strain remained in their respective class (class I or class II) (Figure 2A, third bar). Gene ontology (GO) enrichment analyses of class I and class II revealed a striking overrepresentation of molecular function (MF) terms associated with RNA-binding underlining the validity of our dataset (Figure 2B). Additionally, other terms clearly linked to RNA metabolism were enriched in biological processes (BP) and cellular compartments (CC) (Figures S2A and S2B). In line with this finding, Pfam and SMART analyses of overrepresented protein domains showed RNA recognition motifs as the most enriched domain followed by other classical RNA-binding domains (Figures S2C and S2D).

Figure 2.

Comparison of the C. elegans RIC Dataset to Published RBPomes Reveals Novel RBPs

(A) Bar diagram showing the percentage of proteins contained in each class (class I: yellow, class II: orange, other RBPs: light gray, no evidence: dark gray). The total numbers of proteins measured are shown on top of each bar. The first two bars show data for each genotype only, whereas the third bar depicts combined data from both strains. Proteins measured by MS after RIC are classified depending on the level of confidence regarding their association with RNA.

(B) GO-term enrichment analysis (molecular function) of the combined RBPome from WT and vhl-1 using the whole proteome as a background. Bar diagram depicting the top ten significant terms showing the highest and lowest enrichment factors (Fisher exact test; p-value<0.001). The numbers next to each bar indicate proteins contained in the RBPome followed by the size of the category.

(C) Comparison of the 1,354 RBPs (class I and II) identified in the pooled analysis from both C. elegans genotypes to published datasets. Dark gray: RBPs identified by previous RIC; light gray: previously not identified. Protein numbers are indicated in the respective bars.

(D) Comparison of the 1,354 RBPs (class I and II) identified in the pooled analysis from both C. elegans genotypes to the published worm RIC dataset (Matia-Gonzalez et al., 2015) and to a combined dataset of RBPs (previous RIC in human, mouse, yeast, fly [Hentze et al., 2018]). The total number of proteins contained in each dataset is indicated in brackets.

(E) Comparison of the 1,354 RBPs (class I and II) identified in the pooled analysis from both C. elegans genotypes to proteins identified by three novel techniques called protein-crosslinked RNA extraction (XRNAX), orthogonal organic phase separation (OOPS), and phenol toluol extraction (PTex) (Queiroz et al., 2019, Trendel et al., 2019). In brackets are reported the numbers of RBPs identified in the respective study.

(F) Bar diagram showing the percentage of known and novel RBPs in our dataset when pooling all comparisons to published studies (as specified in Figures 2C, 2E, and S2F).

Comparative Analyses with Published Datasets Reveal Novel RBPs in C. elegans

To further characterize the C. elegans RBPome, we performed an in-depth comparison to published datasets from different model organisms summarized in a recent study (Hentze et al., 2018) and complemented it with an RBPome of murine cells cultured under hypoxia (Ignarski et al., 2019). More than half of class I and class II proteins identified in our study had previously been identified by RIC screens (Figure 2C). In order to check whether our new dataset provided additional information to the only previously published C. elegans RBPome (Matia-Gonzalez et al., 2015), we compared both RBPomes with all RBPs identified in other species. This analysis revealed that our dataset confirmed 364 proteins as RBPs previously identified in the nematode and additionally 610 proteins previously identified as RBPs in other species (Figure 2D). Interestingly, most of the RBPs identified in our dataset were not predicted before in a study identifying putative nematode RBPs in silico (Figure S2E); this is in line with the fact that many RBPs do not contain classical RNA-binding domains (Beckmann et al., 2015, Tamburino et al., 2013). To complete this characterization, we compared our data with three recently published RBPome datasets using a novel methodology that—instead of RNA-pulldown targeting polyadenylated transcripts only—is based on purification of proteins crosslinked to the total RNA by organic extraction (Queiroz et al., 2019, Trendel et al., 2019, Urdaneta et al., 2019) (Figure 2E). Finally, taken together all these different comparisons, we can conclude that our dataset contains 1,084 previously described and 270 novel RBPs (Figure 2F, “RBPome” tab http://shiny.cecad.uni-koeln.de:3838/celegans_rbpome and Table S1).

Analysis of the Proteome of vhl-1 Mutant Worms

To move our study toward a characterization of RBPs differentially regulated upon mutation of vhl-1, we performed MS on whole worm lysates obtained from WT and vhl-1 mutants (RICs input). PCA and hierarchical clustering of these data showed a clear separation by genotype (Figures 3A and S3A). In total, we identified 5,759 proteins, 153 of which were differentially expressed between vhl-1 mutant and WT worms (student's ttest; FDR<0.05) (Figure 3B, “Proteome vhl-1 vs WT” tab http://shiny.cecad.uni-koeln.de:3838/celegans_rbpome and Table S1). Importantly, both HIF-1 itself as well as known HIF-1 target genes are more abundant in vhl-1 mutants (Figure 3B, black dots) (Dengler et al., 2014, Ortiz-Barahona et al., 2010, Semenza, 2012, Shen et al., 2005). A GO term enrichment analysis of the significantly regulated proteins revealed that biological processes known to be modulated by HIF-1 such as defense, immune response, and CoA desaturase activity were overrepresented (Figures 3C, S3B, and S3C) (Krzywinska and Stockmann, 2018, Palazon et al., 2014, Zhang et al., 2013). Comparing the differentially expressed proteins with our RBPome dataset, we found only 12 RBPs that differ in abundance on the protein level (student's ttest; FDR<0.05) (Figure 3D).

Figure 3.

RBPs Differentially Regulated in vhl-1 Mutant Worms Proteome

(A) PCA of the proteome mass spectrometry data (on the basis of iBAQ intensities). WT samples are indicated in red and vhl-1(ok161) mutants in blue.

(B) Volcano plot illustrating the differentially expressed proteins between the proteomes of vhl-1(ok161) mutant and WT worms. The -log₁₀ p-value is plotted on the y-axis. The log₂ fold change (vhl-1(ok161) vs WT) is indicated on the x-axis. Proteins above the cutoff line are considered significant (student's ttest; FDR<0.05). Black dots: known HIF-1 targets (Dengler et al., 2014, Ortiz-Barahona et al., 2010, Shen et al., 2005), differentially expressed HIF-1 targets are indicated by name; HIF-1 itself is indicated by the arrow; gray dots: other proteins. Total number of proteins = 5759.

(C) Bar diagram shows the top five GO-terms with the highest enrichment factor (Fisher exact test; p-value<0.001). Biological processes (GOBP) and molecular function (GOMF) of the significantly regulated proteins in the proteome are depicted. The numbers indicate the proteins in the category followed by category size.

(D) Volcano plot illustrating the differentially expressed proteins between the proteomes of vhl-1(ok161) mutant worms and WT worms. The -log₁₀ p-value is plotted on the y-axis and the log₂ fold change (vhl-1(ok161) vs WT) on the x-axis. Proteins above the cutoff line are considered significant (student's ttest; FDR<0.05). Class I and II RBPs from WT and vhl-1(ok161) mutant worms are shown in black, differentially expressed RBPs are indicated by name.

Modulation of the RBPome by vhl-1 Loss-of-Function

The classification of proteins identified by RIC in the two different strains as shown in the bar diagram in Figure 2A depends on arbitrary thresholds. Consequently, proteins detected only in one of the strains based on these thresholds are not necessarily specific to this condition. To allow for a more exhaustive view on hypoxia-signaling-associated RNA-protein binding events, we performed an in-depth analysis of this aspect using the following strategy. Analyzing the data from crosslinked samples in RIC, we found five proteins that were measured in all three vhl-1 crosslinked samples but never measured in WT (both + and − crosslinking) (Figure 4A). Importantly, the abundance of these proteins in the proteome is not affected by mutation of vhl-1, suggesting that the different efficiency in the pulldown observed indeed depends on differential binding of RNA molecules (Figure S4A). Additionally, we found 24 RBPs that were measured exclusively in all WT crosslinked samples but never in vhl-1 (Figure 4B). Again, the abundance of these 24 proteins was not affected by genotype on the protein level (Figure S4B). To extend the comparison to proteins detected in both genotypes we calculated for WT (crosslinked vs non-crosslinked) versus vhl-1 (crosslinked vs non-crosslinked), the linear regression, and the 95% prediction interval from log₂ fold changes (for detailed information see Transparent Methods). Proteins outside the 95% prediction interval were considered to be more strongly enriched in either WT or vhl-1 mutant worms (Figure 4C and Table S1) (“Proteome WT vs. vhl-1” tab http://shiny.cecad.uni-koeln.de:3838/celegans_rbpome). Performing this analysis, 25 RBPs were more enriched in vhl-1 mutants (blue) and 26 RBPs in WT (red) worms. The abundance of the 25 RBPs enriched in vhl-1 mutants was not affected by genotype on the protein level and only 1 (H28G03.1) out of the 26 proteins enriched in WT showed a difference in protein abundance (Figures S4C and S4D). Interestingly, this protein—H28G03.1, an orthologue of human HNRNPA proteins—was enriched in the pulldown from crosslinked WT samples but showed higher protein abundance in vhl-1 mutant worms pointing toward opposite modes of regulation regarding protein levels and RNA-binding capacity.

Figure 4.

Loss of VHL-1 Leads to Quantitative and Qualitative Changes in the RBPome

(A) Table of RBPs detected exclusively in all three vhl-1(ok161) crosslinked samples but never detected in WT. Depicted are: wormbase (WB) gene ID, gene name, human ortholog name, RBPome class (indicated by I [class I] or II [class II]), novel RBP (indicated by +), and the function; n.a., not available.

(B) Table depicting RBPs detected exclusively in all three WT crosslinked samples but never detected in vhl-1(ok161) (for table details see A.).

(C) Scatterplot showing the correlation of log₂ fold changes of vhl-1(ok161) RBPs (crosslinked vs non-crosslinked) on the x-axis and WT (crosslinked vs non-crosslinked) samples on the y-axis. FC, fold change; +CL, crosslinked; -CL, non-crosslinked; vs, versus. The linear regression was calculated with R (black line, deming() method, formula: y = 0.7806137 + 1.095646*x). Proteins outside the calculated 95% prediction interval (gray lines) are considered to be more strongly enriched after RIC in one of the two strains, suggesting a regulation of the binding to target RNAs in WT or vhl-1(ok161). Blue dots: proteins more enriched in vhl-1(ok161); red dots: proteins more enriched in WT; gray dots: not in class I or class II.

Discussion

Recently, the global landscape of RBPs has been addressed in many organisms from yeast to mammals using RIC (Hentze et al., 2018). Whereas studies on mammalian RBPs—due to availability and feasibility—focused on cultured cells (Baltz et al., 2012, Beckmann et al., 2015, Castello et al., 2012, Kwon et al., 2013, Liepelt et al., 2016, Schueler et al., 2014), the RBPome of multicellular organisms has been described for the fruit fly, nematode, Arabidopsis thaliana, and zebrafish so far (Despic et al., 2017, Marondedze et al., 2016, Matia-Gonzalez et al., 2015, Reichel et al., 2016, Sysoev et al., 2016, Wessels et al., 2016). Data regarding the C. elegans RBPome is available from a study published by Matia-Gonzalez et al. that used UV-C crosslinking and RIC in mixed-stage worms as well as L4 larvae after induction of apoptosis (Matia-Gonzalez et al., 2015). In our study, we describe the first C. elegans RBPome from young adult worms using two different genotypes. Employing stringent filtering criteria this approach identified 1354 RBPs, around 26% of which had been described by Matia-Gonzalez before. This finding underlines the importance of performing RIC in different biological and technical conditions to obtain a global view on the RBPome of a specific organism. However, the majority of our RBPs is not entirely novel and had been identified in cells from other species before showing the validity of our dataset. Two hundred seventy proteins had not been described in any other published RIC study leading to their classification as novel RBPs. Their novelty may be attributed to a couple of different reasons. First, only about half of them are conserved in mammals, making their identification impossible in most of the previous screens. Second, technical differences, especially regarding MS and data analysis, may account for this finding. Third, and very importantly in our view, both expression and RNA-binding capacity of specific RBPs can be context specific, leading to their first description in synchronized young adult and vhl-1 knockout worms. In this context, dynamic modulation of the RBPome is a highly interesting research question. The use of vhl-1 mutants allowed us to obtain a first view on changes in the RBP landscape upon genetic activation of HIF-signaling. There are several lines of evidence linking HIF-signaling to RNA-protein binding events. On the one hand, RBPs can have an impact on HIF expression itself. The human antigen R (HuR) binds to the 5′untranslated region (UTR) of HIF-1α mRNA and thereby promotes it translation (Galban et al., 2008). Another study showed that the RBP RBM38 is able to bind HIF1α mRNA via binding to HIF1α 5′ and -3″UTRs. Moreover, knockdown of RBM38 increased the level of HIF-1α protein under hypoxic conditions (Cho et al., 2015). On the other hand, it is known that hypoxia can lead to repression of cap-mediated translation, involving RBPs (Uniacke et al., 2012). This phenomenon can be overruled for transcripts that are important to the response to hypoxia through an HIF-2α-RBM4-eIF4E2 complex that binds to these mRNAs and targets them to polysomes for translation (Uniacke et al., 2012). Furthermore, recently published work from our group showed modulation of the RBPome by hypoxia in cultured cells (Ignarski et al., 2019). Consequently, hypoxia signaling appeared as an attractive target to be studied in a genetic model of C. elegans. Here, it should be noted that we observed a reduced RNA and protein yield in vhl-1 mutant worms. Mutation of vhl-1 has been shown before to lead to a smaller size of the nematode (Wen et al., 2015), which may be the underlying reason for the unexpected lower yield. However, due to the similar amount of RNA after pulldown in both strains and conditions we do not expect this to have an impact on the results of our RIC. Analysis of the proteome confirmed HIF-1 to be stabilized and significantly upregulated in vhl-1 mutant worms. It is important to note that our study cannot dissect RBPs affected by activation of HIF-signaling from those that may be affected by loss of vhl-1 directly. Considering that most changes in these mutants are generally assumed to be mediated by HIF-1, it is likely that this is the case for the majority of changes we observed as well. However, final proof of this will require future experiments using either vhl-1; hif-1 double mutant worms or different means of HIF activation, e.g. expression of a stabilized version of HIF-1. To allow for a first insight into evolutionary conservation of hypoxia-signaling-associated modulation of RBPs we compared the results in this study with our previous findings in cultured cells under hypoxia (Ignarski et al., 2019). Based on the results in Figure 4 of the study at hand, we did not find any obvious overlap regarding the modulation of RNA-binding. This is well explained by several key differences in experimental design. In Ignarski et al., HIF activation was performed by short-term exposure of cultured mIMCD3 cells to hypoxia, whereas HIF-signaling is permanently activated in vhl-1 mutants. However, when comparing all RBPs that show a hypoxia-modulated RNA-binding capacity in cultured cells with the C. elegans RBPomes, we found two RBPs showing similar changes in both studies. On the one hand, CCT-1—a component of the TCP1 chaperonin complex—reaches statistical significance as class II RBP only in WT but not in the vhl-1 mutant (see “RBPome” tab http://shiny.cecad.uni-koeln.de:3838/celegans_rbpome and Table S1). In line with this finding, the mouse ortholog, TCP1, is identified as an RBP only in normoxic cultured cells (Ignarski et al., 2019). TCP1 is well known to be dedicated to the folding of actin and tubulin (Sternlicht et al., 1993, Vallin and Grantham, 2019). Interestingly, TCP1 mediates also the folding and assembly of VHL into a complex with its partner proteins (Feldman et al., 1999), showing its importance in HIF-singling. On the other hand, PRO-3 only reaches significance as an RBP in vhl-1 mutant worms (see “RBPome” tab http://shiny.cecad.uni-koeln.de:3838/celegans_rbpome and Table S1). Its mouse orthologue—SDAD1, a protein required for 60S pre-ribosomal subunit export to the cytoplasm—did only reach the criteria of an RBP in cells after hypoxia (Ignarski et al., 2019). SDAD1 is implicated in regulation of tumor progression and metastasis (Ding et al., 2018, Zeng et al., 2017). There is an indication that HuR, an RBP targeting HIF-1 mRNA (Galban et al., 2008), also binds to SDAD1 mRNA (Jing et al., 2019). It will now be of great interest to focus on the function of specific hypoxia-modulated RBPs such as SDAD1 and TCP1 and to further characterize their differential RNA-binding, e.g. using crosslinking and immunoprecipitation protocols. This will not only allow for distinguishing differential RNA-binding capacity from differential availability of the actual mRNA targets as being the reason for the observed hypoxia-associated changes but also be the first step toward elucidating the biological consequences of hypoxia-induced alterations of these RNA-protein binding events.

Limitations of the Study

In the present study, we found significantly lower RNA and protein yields in vhl-1 mutants using equal numbers of worms. The reasons for this finding are not examined here but can be hypothesized based on published data. Firstly, Wen et al. described that vhl-1 mutants are shorter compared with WT (Wen et al., 2015). Secondly, it is well known that loss of vhl-1 induces longevity (Mehta et al., 2009, Muller et al., 2009, Zhang et al., 2009). Ewald et al. reported a direct link between collagen abundance and longevity (Ewald et al., 2015). Furthermore, the mammalian orthologue of VHL-1 is known to have an impact on extracellular matrix formation (Kurban et al., 2008). Changes in the worm cuticle could lead to differences in the efficiency of RNA and protein extraction. Although this issue does not affect the RNA yield after pulldown, we cannot exclude a resulting bias introduced into the comparison between WT worms and vhl-1 mutants. More experiments will be required to address this aspect conclusively. Our analysis of RBPs in C. elegans started with three biological replicates of crosslinked and non-crosslinked WT and vhl-1 mutant samples. However, one non-crosslinked sample of each genotype showed a much higher number of proteins identified by MS than expected (even more than in the corresponding crosslinked samples), potentially due to contamination with whole worm lysate. These samples were excluded from the analysis. The RBPome analysis is therefore based on three biological replicates of crosslinked and two non-crosslinked WT and vhl-1 mutant samples. Notably, previously published RIC studies also gained reliable data using two replicates and even pooled non-crosslinked data from different conditions due to high similarity (Baltz et al., 2012, Castello et al., 2012, Liepelt et al., 2016, Mitchell et al., 2013).

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.

Published: December 20, 2019

Data and Code Availability

The mass spectrometry data (Raw data and MaxQuant [version 1.5.3.8] output) have been deposited to the ProteomeXchange Consortium (http://www.ebi.ac.uk/pride) via the PRIDE (Perez-Riverol et al., 2019) partner repository with the dataset identifier PXD014469.

An interactive online repository was created and is provided at http://shiny.cecad.uni-koeln.de:3838/celegans_rbpome.

Supplemental Information

Table S1. Table of WT and vhl-1 RNA Interactome and Proteome, Related to Figures 1–4

Proteome dataset (blue filling): vhl-1 versus WT ttest results (FDR<0.05) and mean IBAQ intensities are presented. RBPome datasets (red filling): the results for the t-tests (vhl-1 +CL vs -CL and WT +CL vs -CL) (FDR<0.05) and the mean iBAQ intensities are presented. For detailed information see Table S1 Info tab.