Post-transcriptional gene regulation (PTGR) is crucial for maintaining cellular proteome homeostasis [1
], disruption of which can cause severe diseases, such as cancer and infertility [3
]. PTGR requires the activity of RNA-binding proteins (RBPs), such as the widely studied pumilio (PUM) proteins, which are founding members of the PUF (pumilio and fem-3 binding factor) family of eukaryotic RBPs. PUM proteins are highly conserved and present in many organisms, from yeast to humans (for review, see [4
]). Simultaneous knockout of mouse PUM1 and PUM2 is lethal [5
], indicating their crucial role in development. Post-transcriptional regulation by PUMs is mediated by the conserved C-terminal RNA-binding PUF domain, which is composed of eight tandem repeats [6
], and binds a specific eight nucleotide sequence 5′-UGUAHAUA-3′ (H represents A, C or U, but not G), called the PUM-binding element (PBE) that is typically located in the 3′ untranslated regions (3′UTR) of target mRNAs. By binding PBEs, PUMs trigger the recruitment of protein cofactors, that together direct selected mRNAs towards post-transcriptional repression or activation (for review see [4
Each of the five PUMs in yeast (Puf proteins) contains a PUF domain that is different in structure from the others, contains between six and eight tandem repeats, and binds to a distinct PBE motif. In this way, each Puf co-ordinately controls the fate of multiple mRNAs sharing a specific PBE motif and which have been found to be functionally related [7
]. These findings became the basis for the so-called PUM/Puf RNA regulon model [8
]. Considering the high structural similarity of PUM1 and PUM2, it is still unresolved whether they form separate regulons in mammals. Although mammalian PUM1 and PUM2 contain nearly identical PUF domains [9
] that recognize the same PBE motif (UGUANAUA) [10
], there is some evidence for divergent modes of regulation. Examination of interactions between another RBP, ARGONAUTE2 (AGO2), and PUM proteins revealed a substantial fraction of nonoverlapping PUM1 and PUM2 mRNA targets [11
]. Therefore, it is possible that PUM1 and PUM2 paralogues are functionally nonredundant and function as distinct RNA regulons. We have recently demonstrated a specific example of functional nonredundancy between PUM1 and PUM2 by showing that while PUM2 induces PBE-dependent repression of the mRNA target SIAH1, PUM1 does so in a PBE-independent manner [12
]. Additionally, the regions N-terminal to the PUF domain, which are divergent between PUM1 and PUM2, were reported to contain three unique subregions with autonomous repressive activity that may represent an interface for binding protein cofactors since they were not demonstrated to bind RNA [13
A number of PUM protein cofactors, such as NANOS1, NANOS3, and DAZ family members, are associated with male or female infertility in humans [14
]. Therefore, establishing the mechanisms underlying functional divergence of PUM1 and PUM2, including identification of their protein cofactors, may help in understanding their particular roles in human germ cells as well as human infertility, a problem affecting 15% of couples world-wide [20
]. Male infertility in particular impacts 7% of the male population (for review see [21
]). Notably, male infertility is a risk factor for developing testis germ cell tumor (TGCT) [22
]. Testicular cancers are the most frequently diagnosed malignant tumors in young Caucasian males, and their incidence has increased [23
], highlighting the importance of the human male germ cell context in studying PUM1- and PUM2-controlled regulation. However, the available germ cell line which has the most in common with human germ cells among germ cell-mimicking cell lines is TCam-2. This cell line originates from human seminoma, a type of TGCT, and represents male germ cells at an early stage of prenatal development [24
]. The identification of PUM mRNA targets and PUM-interacting proteins had not been previously studied in human germ cells (which would help establishing the mechanisms underlying functional divergence of PUM1 and PUM2 in these cells) and therefore may help in understanding the reasons behind infertility in humans. To the best of our knowledge, the identification of PUM mRNA targets in germ cells has only been studied in the C. elegans
]. Therefore, the purpose of this study was to identify and characterize RNA regulons of PUM1 and PUM2 paralogues in TCam-2 cells, clarify whether these regulons are redundant, and if not, discuss potential functional consequences of their divergence in human reproduction. In this study, by RIP-Seq, RNA sequencing, and mass spectrometry (MS), distinct mRNA pools and interacting proteins were identified for PUM1 and PUM2 in human germ cells, thereby enabling understanding of the functional relevance of PUM to fertility.
2. Materials and Methods
2.1. RNA Immunoprecipitation and Sequencing
For RIP analysis, TCam-2 cells were grown in 37 °C and 5% CO2 in Roswell Park Memorial Institute (RPMI, Life Technologies 61870044, Paisley, UK) 1640 medium supplemented with 10% FBS (GE Healthcare HyClone SH30071, Logan, Utah, USA) and 1% penicillin/streptomycin (Lonza EE17-602E, Germany). RIP-Seq experiments with UV cross-linking were performed using the Magna RIPTM RBP Immunoprecipitation Kit (17-700 Merck, Darmstadt, Germany). Briefly, 100 μL of Magnetic A/G beads were coated with 12 μg anti-PUM1 (S-19, sc-65188 Santa Cruz Biotechnology, Santa Cruz, CA, USA), anti-PUM2 (K-14, sc-31535 Santa Cruz Biotechnology, Santa Cruz, CA, USA) antibody or IgG fraction from nonimmunized goat serum (G9759, Sigma Aldrich, Saint Louis, MO, USA) for 45 min at room temperature (RT) in Magna RIP Wash Buffer. TCam-2 cells were washed twice with ice-cold PBS and subjected to UV cross-linking at 254 nm on a HEROLAB CL-1 Cross-linker for 30 s (0.015 J). For one RIP-Seq reaction, 2–3 × 106 cells were lysed in 500 μL of Magna RIP Lysis Buffer for 30 min with rotation at 4 °C. Lysates were centrifuged (10 min at 10,000× g), and the supernatant was mixed with precoated beads suspended in washing buffer supplemented with protease and RNase inhibitors. The RIP reaction was held for 3 h at 4 °C on a rotator in a final volume of 1 mL. Then, magnetic beads were washed five times with Magna RIP Washing Buffer, followed by treatment with proteinase K at 55 °C for 30 min. Total RNA was isolated from magnetic beads using a QIAGEN RNeasy Plus Micro Kit according to the manufacturer’s protocol, and RNA quality was checked on an Agilent Bioanalyzer using an RNA 6000 Nano Kit. RNA with a RIN value >7 was used for further steps. cDNA libraries for RNA-Seq analysis were prepared using Illumina TruSeq RNA Sample Prep V2, and subsequent next-generation sequencing was performed on an Illumina HiSeq 4000 platform by Macrogen INC. Sequencing was performed under the following conditions: Paired-End reads were 100 nt long, and >70 million reads/sample were obtained. RIP-Seq with anti-PUM1, PUM2, and IgG (negative control) were performed in triplicate. For TCam-2 transcriptome analysis, total RNA was isolated from 80% confluent 10 cm2 dishes using a QIAGEN RNeasy Plus Micro Kit. RNA quality control and RNA-Seq were performed as described above. An mRNA level that was at least 2-fold enriched (with adjusted p-value < 0.05) in anti-PUM1/PUM2 co-IP, in comparison to the negative control (co-IP anti-IgG) and to the TCam-2 transcriptome level, was considered to be bound by PUM1 or PUM2.
2.2. Western Blot Analysis
To check for PUM1 and PUM2 binding efficiency, SDS lysates from beads after co-IP were resolved on 8% SDS polyacrylamide gels and transferred to nitrocellulose membranes (BioRad). Membranes were blocked with 5% low-fat milk in TBS buffer supplemented with 0.1% Tween 20 (blocking buffer) at RT for 1 h. Membranes were incubated with primary antibodies at 4 °C overnight in blocking buffer. On the next day, membranes were washed four times in TBS buffer with 0.1% Tween 20 and incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies at RT for 1 h in the same buffer. The following antibodies were used: goat anti-PUM1 (1:1000 Santa Cruz Biotechnology #S-19, sc-65188), goat anti-PUM2 (1:250 Santa Cruz Biotechnology #K-14, sc-31535), rabbit anti-actin beta (ACTB) (1:10,000 Sigma Aldrich, A2066), and HRP-linked anti-goat (1:50,000 Santa Cruz Biotechnology #sc-2020), as well as HRP-linked anti-rabbit (1:25000 Sigma Aldrich A0545). Next, membranes were washed twice in TBS buffer with 0.1% Tween 20, and then twice in TBS buffer. ClarityTM ECL Western Blotting Substrate (BioRad, 170—5061, USA) and the ChemiDoc Touch Imaging System (BioRad) were used for signal development and analysis. To check the silencing efficiency of PUM1 and PUM2, SDS lysates were prepared from cells 72 h post-transfection and analyzed in the same way as lysates from beads.
2.3. Bioinformatic Analysis of PUM1- and PUM2-Bound mRNAs
The Paired-End sequence reads obtained from the HiSeq4000 platform were trimmed using the Trimmomatic v0.35 tool with the following parameters: ILLUMINACLIP:TruSeq2:PE MINLEN:50, including quality filtration using SLIDINGWINDOW:10:25, MINLEN:50 parameter. Sequence reads that passed quality filters were mapped to the human reference genome (UCSC hg19) using TOPHAT(2.1.0) [26
] with default parameters. Then, reads were counted using CUFFLINKS (220.127.116.11), followed by merging replicates with CUFFMERGE and calculating differential gene expression with CUFFDIFF (18.104.22.168) using a p
-value and false discovery rate (FDR) of <0.05.
For selection of mRNAs potentially bound to PUM1, PUM2, or both, the following criteria were used: (1) only mRNAs enriched in all three replicates; (2) at least 2-fold enrichment in PUM1 or PUM2 IPs, compared to IgG; and (3) at least 2-fold mRNA enrichment in comparison to TCam-2 transcriptome (all enrichment were with p
-value and FDR < 0.05). To annotate mRNAs bound by PUM to their cell-specific functions and pathways, we performed GO analysis using BiNGO plug-in (version 3.0.3) [27
] on Cytoscape platform (version 3.6.1) with functional annotation of biological process and molecular function, searched against TCam-2 cell line gene expression (FPKM > 0.5) background derived from our RNA-Seq. Heatmaps were created using R (version 3.4.4) (R Core Team (2018). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL: https://www.R-project.org/
) and gplots R library. To identify PBEs in the 3′UTR of each PUM1- or PUM2-bound pool of mRNAs, we used the DREME motif discovery tool (v.4.12.0), which enables the identification of short uninterrupted motifs that were enriched in our sequences compared with shuffled sequences. To search for PBE in whole mRNA PUM targets or their 5′UTR, CDS, and 3′UTR, we used FIMO (v.4.12.0), which enables scanning for individual matches for an input motif aligned to individual sequences (with p
-value < 0.01). Whole mRNAs and their 3′UTR, 5′UTR, and CDS sequences were downloaded from the RefSeq Genes Genomic Sequence database (Table browser: assembly: Feb. 2009(GRCh37/hg19); track: NCBI RefSeq) [28
2.4. siRNA Silencing of PUM Proteins
TCam-2 cells were transfected with siRNA using PUM1 siRNA (sc-62912 Santa Cruz Biotechnology), PUM2 siRNA (sc-44773) containing three different siRNAs for each PUM gene (their sequences are in Table S1
) or control siRNA-A (sc-37007) at the final 40 nM concentration using the NEON transfection/electroporation system (Thermo Fisher Scientific). Transfections were performed in Buffer R using 10 µL NEON tips. Subsequently, after transfection, cells were cultured in antibiotic-free RPMI 1640 (Gibco) supplemented with 10% FBS (HyClone) at 37 °C and 5% CO2
for 72 h. To measure mostly the post-transcriptional effect of PUM knockdown, transcription was inhibited. To this end, the cells were treated with 5 µg/mL Actinomycin D (A1410 Sigma Aldrich) for 4 h before lysis. Transfection was performed in three biological replicates. RNA isolation was performed using a QIAGEN RNeasy Plus Micro Kit. RNA quality analysis was performed as described above, RNA with RIN > 9 was used for cDNA library preparation, and subsequent sequencing was performed as described above. The knockdown efficiency of each replicate was analyzed by western blot.
2.5. Bioinformatic Identification of mRNAs Under Regulation by PUM Proteins
More than 80 million reads per sample obtained from the Illumina HiSeq 4000 platform were analyzed as described above. If the mRNA level increased by at least 20% (adjusted p
-value and FDR < 0.05) under 70–90% PUM knockdown compared to negative siRNA control, the mRNA was considered to be under PUM repression. If the mRNA level decreased by at least 20%, it was considered to be significantly activated by PUM. We set the threshold at 20% as sufficient given that these changes were found in three biological replicates (adjusted p
-value and FDR < 0.05) and the protein silencing efficiency of PUM1 and PUM2 was high, (over 70% and 90%, respectively) (Figure 1
Cumulative distribution analysis was performed using log2 fold changes of mRNAs identified in RIP-Seq after PUM1 or PUM2 KD. We used mRNAs not bound in RIP-Seq as controls. A two-sided Kolmogorov–Smitnov test was used to assess statistical significance (using R software version 3.4.4).
2.6. RT-qPCR Analysis of mRNA Expression after PUM1 and PUM2 Knockdown
To confirm the targets regulated by PUM1 and PUM2, TCam-2 cells were transfected in three biological replicates with siRNA as described above. RNA from cells was isolated using TRIzol reagent (Gibco) according to the manufacturer’s protocol. Purity and amount of RNA was analyzed by Nanodrop 2000 (ThermoFisher). RNA integrity was determined by Bioanalyzer (Agilent Technologies). Approximately 1 µg of total RNA with RIN > 9 was treated with DNase I (D5307, Sigma-Aldrich) for 20 min at RT and reverse transcribed using the Maxima First-Strand cDNA Synthesis Kit (K1671, ThermoFisher Scientific) according to the manufacturer’s protocol. qPCR was performed on generated cDNA using Jump-Start Taq DNA Polymerase (D4184, Sigma-Aldrich), CFX96 Touch Real-Time PCR Detection System (BioRad) and SYBR Green dye (ThermoFisher Scientific) in three biological replicates with at least five technical replicates of each reaction. The list of primers used for RT-qPCR is shown in Table S2
. All changes in mRNA levels upon PUM1 or PUM2 knockdown were normalized to ACTB and GAPDH base on geometric averaging [28
]. For all RT-qPCR analyses we have used two-way unpaired t
-test (α = 0.05) to estimate statistical significance, since all our data was normally distributed, according to a Shapiro–Wilk test (α = 0.05). A p
value < 0.05 (*) in the t
-test was considered statistically significant. All our statistical analyses were performed using GraphPad Prism version 8 software.
2.7. Mass Spectrometry Analysis after Anti-PUM1 and Anti-PUM2 Immunoprecipitation
Six biological replicates of co-IPs (three performed without RNase A treatment, and another three with 100 mg/mL RNase A) with anti-PUM1, anti-PUM2 antibodies (including anti-IgG negative control) were performed as described above. The same specific antibodies as in RIP were used in MS/co-IP and RIP experiments. MS protein identification analysis was performed by MS Laboratory, IBB PAS, Warsaw. Briefly, proteins were directly digested on the beads and separated by liquid chromatography (LC) followed by MS measurement of peptides and their fragmentation spectra (LC-MS/MS) with a Q Exactive Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Scientific).
The bioinformatic protein identification analysis was performed as described: peak lists obtained from MS/MS spectra were identified using X! Tandem version X! Tandem Vengeance (2015.12.15.2), Andromeda version 22.214.171.124 and MS-GF+ version Beta (v10282). The search was conducted using SearchGUI version 3.2.23 [29
Protein identification was conducted against a concatenated target/decoy [30
] version of the Homo sapiens OX = 9606 (20,316, 99.8%), cRAP (49, 0.2%), the complement of the UniProtKB (version of [2017_06], 20,365 (target) sequences) [31
]. The decoy sequences were created by reversing the target sequences in SearchGUI. The identification settings were as follows: trypsin, specific, with a maximum of one missed cleavage of 30.0 ppm as MS1 and 0.1 Da as MS2 tolerances; fixed modifications: carbamidomethylation of C (+57.021464 Da), variable modifications: Oxidation of M (+15.994915 Da), fixed modifications during refinement procedure: carbamidomethylation of C (+57.021464 Da), variable modifications during refinement procedure: acetylation of protein N-term (+42.010565 Da), pyrrolidone from E (−18.010565 Da), pyrrolidone from Q (−17.026549 Da), pyrrolidone from carbamidomethylated C (−17.026549 Da). All specific algorithm settings are listed in the Certificate of Analysis available in the supplementary information
Peptides and proteins were inferred from the spectrum identification results using PeptideShaker version 1.16.19 [32
]. Peptide spectrum matches (PSMs), peptides, and proteins were confirmed at a 1.0% False Discovery Rate (FDR) estimated using the decoy hit distribution. All confirmation thresholds are listed in the Certificate of Analysis available in the supplementary information
. Protein identification results for every sample are shown in Table S3
. Only proteins inferred with corresponding peptides identified in three independent biological replicates of PUM1 or PUM2 IP and not identified in IgG IP were defined as PUM interactors (Table S4
2.8. Bioinformatic Construction of the PUM RNA Regulon
Binding motifs of putative RBP cofactors of PUM1 and PUM2 were obtained from RBPDB [33
] CISBP [34
] and POSTAR2 [35
] databases (Figure S1A
). Motif enrichment analysis was performed on the identified mRNA targets of PUMs (Figure S1
, Figure 1
B and Figure 2
B) by FIMO [36
] using a greater-than-average threshold (FIMO analysis with p
-value < 0.01; mRNAs for GO analysis bigger than average motif enrichment per sequence). mRNA groups regulated by PUM1 or PUM2 with the enrichment of the binding motif putative of RBP cofactors of the respective PUM (Table S5
, Figure S1
) were determined for each PUM–RBP cofactor pair, in comparison to negative control mRNAs (not bound and not changed under PUM1 and PUM2 silencing). To avoid influence of sequence length, we selected negative sequences, whose average length of 4177 nt (in the range 3000–16,321) was similar to the length of mRNAs regulated by PUM1 (4817 nt, in the range 449–16,862) and PUM2 (5442 nt, in the range 412–16,862). Given that the average size of negative sequences was slightly shorter compared to PUM-regulated targets, we used a coefficient for length correction. GO analysis of PUM-RBP cofactor-regulated mRNAs was performed using ClueGO version 2.5.2 [37
]. GO term selection was performed by using only experimentally validated annotation as evidence code, using the selection parameters of minimum genes annotated to term: 3, minimum percentage of genes: 4%. Annotated biological processes with adjusted p
-values after Bonferroni correction ≤ 0.05 were considered significant. GO analysis results are shown in Table S6
. Visualization of the regulon was performed using Cytoscape platform version 3.6.1.
2.9. Quantification and Statistical Analysis
Statistical analyses were performed as indicated in the above Materials and Methods subsections. To perform them, we used GraphPad Prism version 8 software or R (version 3.4.4). For the rest, statistics included in pipelines and software mentioned above were used.
In the supplementary figures
, Pearson correlation of our TCam-2 and Irie and coworkers’ [38
] TCam-2 transcriptome was calculated and visualised using Galaxy platform: deepTools plotCorrelation as described in [39
]. Principal Component Analysis (PCA) plots were obtained using Galaxy platform: deepTools plotPCA as described in Jarmoskaite I. et al., 2019 [39
Considering the high structural similarity of PUM1 and PUM2, it is still unresolved whether they form separate RNA regulons in mammals. Here, for the first time, by combining RIP-Seq and RNA-Seq upon PUMs knockdown data, together with co-IP LC/MS identification of putative protein cofactors, RNA binding motif enrichment, and GO analysis (for the first time each group of data originating from the same cells—TCam-2 cells), we obtained a model of partially divergent putative PUM1 and PUM2 RNA regulons (Figure 4
). They are reminiscent of previously proposed regulons [8
]. Importantly, a global PUM-dependent gene expression regulation was not studied in germ cells, except C. elegans
We found a much higher average representation of PBE-containing mRNAs that were selected as regulated by PUM1 and PUM2 based on combined analysis of RIP-Seq and PUM RNA-Seq upon PUMs knockdown (96.8 and 99.8%, respectively), than in targets selected based on RIP-Seq (90.8 and 85.9%, respectively) or RNA-Seq alone (59.68, 57.50%, respectively), which validates our approach (Figure 2
B). It is important to note that several of our RIP-Seq identified targets overlapped with mRNAs previously identified in HeLa [10
] and HEK293 [40
] cells, confirming our results. However, it is also important to bear in mind that PUM-mediated activation or repression, or lack of PUM regulation, may be cell-type-specific [47
]. Therefore, we can expect only a partial target overlap when PUM targets from different types of cells are compared.
It is important to note that among PUM-regulated mRNAs, there are also a small number of targets with no PBE (approximately 3% PUM1- and below 1% PUM2-regulated). As mentioned above, PUM proteins may recognize motifs slightly different to the canonical UGUAHAUW [12
]. Such variant motifs were not evaluated in this study; therefore, putative mRNA targets carrying such motifs were overlooked. It is also important to emphasize that in our approach, PUM-regulated mRNAs whose level remained unchanged (did not undergo degradation or stabilization) were overlooked. PUM2-regulated and not PUM1-regulated mRNA repression with no degradation but rather storage in P-bodies was recently suggested to be quite common in human HEK293 cells [48
]. Our result showing a lower number of PUM2 compared to PUM1-regulated targets is in line with that finding.
We found that in mRNAs positively regulated (activated/stabilized) by PUM1 or PUM2, PBE motifs were significantly more frequent in the 5′UTR (14.67% for PUM1 and 16.47% for PUM2) than in mRNAs negatively (repressed) by PUM1 and PUM2 (3.75 and 4.42%, respectively). However, this was not reported in studies on HEK293 cells [40
]. Although this observation requires further research, it may suggest that activation of these mRNAs by PUM proteins requires PBE localization in the vicinity of some 5′UTR translational signals.
Interestingly, by using the RIP-Seq approach we identified 30% of PUM1/PUM2-bound common targets. However, the combination of RIP-Seq with siRNA knockdown to identify regulated targets resulted in a decrease of common targets to 10%. We propose that this difference is due to the involvement of distinct regulatory factors for each PUM paralogue. It is worth emphasizing that we identified such regulatory factors—putative PUM-interacting protein cofactors which control different aspects of RNA metabolism (stability, localization, transport, splicing and expression regulation), whose interaction was RNA-mediated as well as protein cofactors whose interaction was RNA-independent. A substantial number of protein cofactors were PUM1- or PUM2-specific in both groups. The first group of RNA-dependent protein cofactors contains only RBPs, which was expected, and confirms our experiments as well as the analysis performed. However, RBPs were also significantly enriched in the second group representing RNA-independent protein–protein interactions. Such RBPs are likely to contain protein–protein interacting domains that bind PUM, as well as RNA-interacting domains that bind RNA. Finally, interactors with no RNA-binding domains might be important for the stabilization of ribonucleoprotein complexes, which are formed upon PUM protein binding specific mRNA targets.
Among the identified PUM putative protein interactors, we found five previously reported human PUM binding proteins, which confirms our results. MATR3 and SEC16A were previously identified in a high-throughput proteomic study in HeLa cells [49
]. Another one is G3BP1, which is a stress granule assembly factor [50
]. The next one is the fragile X mental retardation protein (FMR1) and its autosomal homologous proteins, FXR1 and FXR2. FMR1 was previously shown to colocalize with PUM2 in rat neuron stress granules [51
]. More recently, Zhang and coworkers reported that FMR1 interacts with PUM in the murine brain in an RNA-dependent manner [52
]. In our study, FMR1 proteins were identified as both PUM RNA-dependent and independent interactors. Interactions with G3BP1 and FMR1 may suggest that PUM paralogues are components of stress granules not only in mammalian neurons [51
], but also in human germ cells. The presence of both PUM paralogues in stress granules suggests their involvement in RNA storage. Interestingly, PUM proteins were also found in P-bodies of HEK293 cells, which, according to a recent report [48
], store high numbers of mRNAs.
Based on our results, we propose that cooperation of such protein cofactors (mainly RBPs) with PUM1 or PUM2 might enable regulation of selected groups of RNA targets responsible for a given metabolic pathway in TCam-2 cells. Unfortunately, neither RIP-Seq nor CLIP-Seq data for PUM protein cofactors identified in this study are available for TCam-2 cells. Therefore, we compared our results to eCLIP data from a different cell line, K562 cells, available in ENCODE. We found that the overlap of RNA targets corresponding to a specific PUM–protein cofactor pair in TCam-2 cells with RNA targets of that cofactor in K562 cells was up to 40%. Overlap between TCam-2 PUM1 or PUM2 RIP-Seq targets and eCLIP protein cofactor targets was up to 25% (Table S12
). This corroborates our results.
Notably, we found that a number of mRNAs that are enriched in TCam-2 cells compared to somatic gonadal tissue or cause infertility when mutated are under the control of PUM1 or PUM2 RNA regulons, which is in line with their divergent functions. Additionally, each of them consists of subregulons (Figure 4
). We propose that identification of germ cell-associated groups of targets that are PUM1- or PUM2-specific might indicate nonredundant roles of PUM paralogues in controlling processes of human reproduction. Notably, the majority of the PUM-regulated genes enriched in TCam-2 cells are genes involved in the development of several types of cancer, mostly of the reproductive system (Table S11
). This observation is in concordance with the fact that TCam-2 cells originate from seminoma testis germ cell tumor [24
]. It is important to underline, however, that the level of several PUM1- or PUM2-regulated targets was more significantly changed upon PUM1/PUM2 double, than single PUM knockdown. This may reflect a mutual regulation of PUM proteins, as it was previously reported that PUM1 represses PUM2 mRNA and vice versa, since they both contain PBEs in 3′UTR [53
]. We have observed this phenomenon of mutual PUM1 and PUM2 repression in TCam-2 cells [54
]. Hence, we propose that knockdown of one PUM paralogue activates a feedback, resulting in upregulation of the other PUM. This feedback likely compensates for the lack of repressive activity of the silenced PUM paralogue. Such a feedback mechanism has recently been proposed in a study on PUM regulons that are formed during mouse development, and which was published during the review process of this manuscript [55
]. Such a feedback mechanism that enriches the versatility of PUM regulatory mechanisms is an interesting issue to be explored in the future.
Interestingly, PUM RNA regulons overlap at points where PUM1 and PUM2 regulate common targets and interact with common protein cofactors (Tables S4 and S10
, respectively). On the other hand, a PUM cofactor may regulate a specific pathway dependent on binding PUM1 or PUM2. For example, FXR2 may regulate endosome transport by binding PUM1 or Rho protein signal transduction by binding PUM2. Likewise, SFPQ may regulate cytosolic or endosome transport by binding PUM1 or endothelium development by binding PUM2.
The majority of the selected PUM targets enriched in TCam-2 cells, compared to gonadal tissue, have been reported to be involved in the regulation of the cell cycle, proliferation, and apoptosis (Table S11
), processes that are important for the maintenance of germ-line status and are under precise regulation to ensure fertility. This functional profile is also in line with the recent suggestion that the evolutionarily original role of PUM proteins is regulation of stem cell self-renewal, including germline stem cells renewal [40
], which the above-mentioned three processes strongly influence. It is important to keep in mind that the TCam-2 cell line represents a model of germ cells, and does not fully reflect the physiology of human male germ cells.