Spatial control of mRNA translation underlies MKRN3-mediated delay of puberty

Summary

Loss-of-function mutations in MKRN3 cause central precocious puberty, but the underlying mechanisms remain unclear, and findings from knockout mice are inconsistent. MKRN3 contains RNA-binding and ubiquitin-ligase domains, so we hypothesized that it regulates GnRH function through RNA and protein interactions. We performed transcriptomic, proteomic, and RNA/protein interactome analyses in GnRH neuronal cells overexpressing Mkrn3. Results indicated predominantly post-transcriptional effects, and the Mkrn3 interactome included the RNA-binding protein Igf2bp2, which is also linked to pubertal timing. Mkrn3 precipitated numerous mRNAs, notably Gnrh1 mRNA. Igf2bp2 also binds this mRNA but in an Mkrn3-dependent manner. Mkrn3 and Igf2bp2 share many mRNA targets, especially nuclear-encoded mitochondrial transcripts, suggesting roles in RNA transport and localized translation. Subcellular fractionation and reporter assays revealed that Mkrn3 alters the localization of proGnRH and Gnrh1 3′UTR-dependent translation and together support a model in which Mkrn3 spatially restricts Gnrh1 mRNA translation, limiting precursor processing and subsequent secretion of the mature peptide.

Graphical abstract

Highlights

• •Mkrn3 binds the Gnrh1 3′UTR and enables the binding of Igf2bp2
• •Mkrn3 alters Gnrh1 mRNA localization
• •The altered mRNA localization influences the site of proGnRH translation
• •This spatial control of proGnRH links Mkrn3 to peptide processing

Neuroscience; Endocrine regulation; Proteomics; Transcriptomics

Introduction

Loss-of-function mutations in Makorin 3 (MKRN3), first reported 12 years ago to be strongly associated with central precocious puberty (CPP),^1,2 are now recognized as the most common genetic cause^3,4,5 and account for 19% of familial cases.⁶ These findings, together with the observation that MKRN3 levels drop markedly leading up to puberty, both in the circulation of children^7,8 and in the rodent hypothalamus,^1,9,10 support its role as a pubertal “brake.” Still, the precise mechanisms through which MKRN3 inhibits pubertal onset have remained unclear.

Like other members of the Makorin family, MKRN3 contains a ubiquitin ligase RING domain and three C3H zinc fingers predicted to bind RNA.^11,12,13 Missense mutations in either of these domains or in the intervening hinge region are associated with CPP,^5,6 implicating both molecular functions in its repression of pubertal onset. Both functions have been demonstrated for MKRN1^{14,15,16,17,18,19} and MKRN2,^{20,21,22,23,24} and MKRN3 is considered an intron less retrocopy of MKRN1.²⁵ However, at least with respect to regulating pubertal timing, loss-of-function mutations in MKRN3 are not compensated for by the ubiquitously expressed MKRN1, suggesting a distinct role for MKRN3 in the hypothalamic arcuate nucleus (ARC), where it suppresses activation of the reproductive axis.²⁶

Mkrn3 has been reported to ubiquitinate various proteins, and Mkrn3-mediated ubiquitination of Mbd3 was found to inhibit its recruitment to the Gnrh1 promoter, resulting in reduced Gnrh1 transcription.^27,28,29 Mkrn3 was also reported to ubiquitinate Pabpc1, reducing its binding to the Gnrh1 mRNA 3′UTR and suppressing Gnrh1 translation.³⁰ Accordingly, elevated Gnrh1 mRNA and protein levels were seen in the Mkrn3 knockout (KO) mouse model of this group.³⁰ Notably, however, another Mkrn3 KO model showed no changes in levels of either Gnrh1 mRNA or protein despite also exhibiting CPP.³¹

Aside from these findings indicating Mkrn3 actions in GnRH neurons, a role for Mkrn3 in regulating transcription in the upstream Kiss1 neurons was also proposed. Mkrn3 mRNA was detected in Kiss1-expressing cells in the mouse ARC, and Mkrn3 overexpression (OE) repressed promoter activity of human KISS1 and TAC3 (encodes neurokinin B: NKB) in reporter assays.⁹ However, Kiss1 and Tac3 mRNA and protein levels were unchanged in both of the Mkrn3 KO mouse models,^28,31 as was the downstream Gnrh1 in this second KO model.^26,32,33 Thus, while there is a consensus that MKRN3 represses pubertal onset, its molecular actions remain unresolved, with current evidence pointing to a role most likely in the downstream GnRH neurons.

Here, we hypothesized that Mkrn3 interacts with specific RNAs and proteins to modulate activity of GnRH neurons in a manner that suppresses the activation of the reproductive axis. Mkrn3 OE in a mouse GnRH neuronal cell line revealed primarily post-transcriptional effects and interaction with proteins involved in translation, RNA-binding and transport, including Igf2bp2, whose connection to pubertal timing was recently reported.³⁴ Importantly, we found that Mkrn3 binds Gnrh1 mRNA, as does Igf2bp2, but only in the presence of Mkrn3. We then discovered that Mkrn3 alters the subcellular distribution of the GnRH precursor (proGnRH), an effect recapitulated by a reporter construct containing the Gnrh1 3′UTR. These findings define a post-transcriptional mode of regulation that inhibits the processing of the GnRH precursor to the active hormone, thereby providing a mechanistic explanation for the inhibitory role of Mkrn3 in the initiation of puberty.

Results

Mkrn3 affects protein levels mostly without change in level of the encoding mRNA

We examined first the effects of Mkrn3 OE on the proteome and transcriptome of GnRH neuronal GT1-7 cells stably expressing GFP-Mkrn3 (Figure S1A). The protein profiles of these cells, determined by mass spectrometry-based proteomics (MS/MS), differed markedly from those of wild-type (WT) GT1-7 cells (Figure S1B), and the level of 623 proteins was significantly altered (log2FC < -1 or >1 and q-value <0.05: Figure 1A). GO analysis of the significantly upregulated proteins indicated enrichment in processes such as nucleotide binding, translation, and apoptosis, while the significantly downregulated proteins were enriched for processes such as transport, cellular localization, vesicle organization, and cytoskeletal protein binding (Figure S1C). Interestingly, levels of the Gnrh1 precursor protein (proGnRH: pro-pyroGnrh and GnRH-associated peptide, GAP) were elevated (Figure S1D).

Figure 1.

Mkrn3 affects protein levels mostly without change in level of the encoding mRNA

(A–C) Volcano plots showing changes in (A) protein levels, (B) total RNA, or (C) cytosolic RNA levels in GFP-Mkrn3-expressing (OE) cells compared to wild-type (WT) GT1-7 cells. Purple/red dots represent proteins/RNAs significantly downregulated (q-value <0.05 and log2FC < -1 [protein] or Padj <0.05 and log2FC < −1 [RNA]), and green dots those significantly upregulated (q-value <0.05 and log2FC > 1 [protein], or Padj <0.05 and log2FC > 1 [RNA]).

(D and E) Log2FC of protein level (top), RNA from whole lysate (middle), and RNA from cytosolic fractions (bottom) for the (D) upregulated proteins or (E) downregulated proteins in Mkrn3 OE compared to WT cells. Up or down regulated proteins are those with log2FC > 1 and <-1, respectively, and q-value <0.05. Gray represents RNAs whose levels were not significantly changed. Pie charts show the corresponding proportions among the affected proteins. The p values assessing the relationship between the changes in RNA expression and protein expression levels were calculated using mHG.

(F) Bar graph summarizes RNA-level changes among proteins significantly altered by Mkrn3. See also Figure S1.

RNA from the total lysates and cytosolic fractions of these cells was sequenced in parallel. Here too, we observed a clear difference in the transcriptomes of Mkrn3 OE and WT cells (Figure S1E). Analysis revealed 352 differentially expressed genes (DEGs: Padj <0.05 and log2FC < -1 or >1) in the total lysate extract, and 223 in the cytosolic extract (Figures 1B and 1C); Gnrh1 mRNA levels were not significantly altered (37.3 ± 2.46 normalized counts in WT and 43.4 ± 12.13 in Mkrn3 OE cells, n = 3; p > 0.05).

Integration of the RNA-seq and proteomics datasets revealed 2178 proteins for which information on the mRNA was also available. Pearson’s correlation analysis indicated a weak correlation between the changes in protein and RNA levels (r of 0.436 for total RNA and 0.41 for cytosolic RNA: Figures S1F and S1G). However, of the 137 upregulated proteins represented also in the RNA-seq data, levels of most of their mRNAs were unaltered (113 in whole-cell lysate, 112 in cytosolic fraction: Figure 1D). Moreover, the majority of the 180 downregulated proteins also did not exhibit changes at the mRNA level (145 in whole-cell lysate and 138 in cytosolic fractions: Figure 1E). Interestingly, using the mHG test, which detects relationships in a ranked manner, we found a correlation between the protein and RNA levels only among those proteins that were significantly downregulated (p values of 3.9e−6 for whole-cell lysate and 0.001 for cytosolic RNA levels) and not among the upregulated proteins (Figures 1D and 1E). These findings, together with the presence of a predicted RNA-binding domain and reported ubiquitin ligase activity, suggest that Mkrn3 effects are predominantly via post-transcriptional regulation, including translation and possibly co- or post-translational processes.

Mkrn3 preferentially binds short transcripts and those encoding mitochondrial proteins, as well as Gnrh1 mRNA

We next looked to identify the RNA targets of Mkrn3 through sequencing after RNA immunoprecipitation (RIP-seq) using a GFP tag. In order to capture the broadest possible range of transcripts, these experiments were performed in transiently transfected cells, resulting in higher Mkrn3 protein levels than in the stable clones. The transiently transfected Mkrn3 localized primarily to the cytosol and in specific foci (Figures S2A and S2B). RIP samples from the GFP-Mkrn3 and control GFP-only GT1-7 cells were clearly distinguished (Figure S2C). Further analysis identified 603 enriched RNA targets (RIP enrichment score >2, padj <0.05: Figure 2A). The vast majority of these (97.35%) were protein-coding mRNAs, with just 2% non-coding RNAs and 0.65% pseudogenes. GO analysis revealed significant enrichment for mRNAs encoding proteins involved in cytosolic and mitochondrial translation, as well as oxidative phosphorylation, and for proteins localized to ribosomes and various mitochondrial compartments (Figures 2B and S2D). Further examination of the RIP-seq data revealed that 24% of the Mkrn3 RIP targets are nuclear mRNAs that encode mitochondrial proteins, while this population of RNAs makes up only 11.7% of the 10,600 RNAs that are not Mkrn3 targets (p value 1.64e−18; Chi-square test; Figure 2C).

Figure 2.

Mkrn3 preferentially binds short transcripts and those encoding mitochondrial proteins, as well as Gnrh1mRNA

(A) Scatterplot shows Mkrn3 RNA targets based on RIP-seq: Red dots are RNA targets with RIP enrichment >2 (Mkrn3 IP vs. Mkrn3 input divided by GFP IP vs. GFP input) and padj <0.05.

(B) GO analysis of Mkrn3 RNA targets describing the enriched cellular components.

(C) Pie charts with the distribution of genes encoding mitochondrial proteins (from MitoCarta3.0) in the RIP-seq Mkrn3 targets and non-targets: p value: 1.64e−18; Chi-square.

(D) Length of the 5′UTR, CDS, and 3′UTR of Mkrn3 RNA targets, shown in violin plots with length as log10 + 1, and p values from the Wilcoxon test.

(E) The top twenty Mkrn3 RNA targets, sorted by Padj values, with those encoding mitochondrial proteins highlighted.

(F) FISH shows the localization of Gnrh1 mRNA in GT1-7 cells relative to GFP-Mkrn3; nuclei stained with DAPI. Scale bars, indicate 10 μm.

(G) RIP-qPCR of FLAG-Mkrn3 association with either (gray bars) three distinct transcripts (shown schematically) comprising dsRed CDS, with the entire or first 31 nt of the Gnrh1 3′UTR and polyA, or (blue bars) H2b mRNA as negative control. Levels shown are IP/input; mean ± SEM and individual values. See also Figure S2.

Interestingly, the median lengths of the coding sequence (CDS) and 3′UTR of Mkrn3 target mRNAs were significantly shorter than of non-targets (CDS: 640 bp vs. 1925 bp, p value 7.67e−43; 3′UTR: 488 bp vs. 1551 bp, p value 2.16e−15) while the median length of the 5′UTR did not differ significantly (116 bp vs. 162 bp, p value 0.254; Wilcoxon test). The preference for short transcripts was evident in both the RNAs encoding mitochondrial proteins and in the other Mkrn3 RNA targets (Figure 2D) and likely reflects distinct classes of genes and/or modes of translation. For example, shorter RNAs are more associated with monosomes,³⁵ which are found in neuronal dendrites, axons, and synapses where they direct active translation for localized protein synthesis.³⁶

Most striking was that Gnrh1 mRNA was among the top 20 most significant Mkrn3 RIP targets, half of which encode mitochondrial proteins (Figure 2E). Fluorescence in situ hybridization (FISH) confirmed the colocalization of Mkrn3 with Gnrh1 mRNA, revealing its significantly higher intensity (∼3-fold: p value 9.6e−10) at Mkrn3 foci as compared to other regions of the cell (Figures 2F, S2E, and S2F).

This interaction was further verified by RIP-qPCR in GT1-7 cells stably transfected with a construct containing the dsRed CDS and SV40 polyA tail with the full Gnrh1 3′UTR, a truncated form lacking the first 31 nt, or without any of this sequence. Although the control (without Gnrh1 3′UTR) transcript was modestly enriched following Mkrn3 OE, enrichment was markedly greater for the transcript containing the full Gnrh1 3′UTR (Figure 2G). Notably, the deletion of the first 31 bp of the 3′UTR abolished this effect completely (Figure 2G), indicating a role for the proximal region of this 3′UTR in Mkrn3 binding of the Gnrh1 transcript.

The Mkrn3 protein interactome is enriched with translational factors, including Igf2bp2, which, in the presence of Mkrn3, also binds Gnrh1 mRNA

For further indication on the possible actions of Mkrn3, we performed affinity purification (AP) in GFP-Mkrn3-expressing cells, followed by mass spectrometry (AP-MS/MS) to identify the Mkrn3 protein interactome. We identified 33 interacting proteins (log2FC > 0.5 and q-value <0.05, compared to control GT1-7 cells expressing only GFP: Figure 3A). These included Psma4, a component of the proteasome seen previously to interact with Mkrn3.^29,30 GO analysis (of 165 proteins with log2FC ≥ 1.5 and p value <0.05) indicated enrichment for proteins involved in cytoplasmic and synaptic translation, mRNA or 3′ UTR binding, RNA localization and stability (Figures 3B and S3A).

Figure 3.

The Mkrn3 protein interactome is enriched with translational factors, including Igf2bp2, which, in the presence of Mkrn3, also binds Gnrh1 mRNA

(A) Volcano plot shows proteins precipitated by GFP in cells expressing GFP-Mkrn3 compared to only GFP: Those in red are significantly enriched (q-value<0.05, log2FC > 0.5).

(B) GO analysis of the Mkrn3 protein interactors (with log2FC > 1.5) indicating the enriched biological processes.

(C) Intersection of Mkrn3 interacting proteins, with genes whose expression correlates with Mkrn3 across development and those associated with puberty (RNA data from¹⁰).

(D) Correlation (Pearson’s coefficient 0.87) of Igf2bp2 and Mkrn3 expression across development in rat mediobasal hypothalamus; levels are mean ± SEM, and individual points shown (data are from GSE94080).

(E) CoIP in GT1-7 cells expressing GFP-Mkrn3 and/or FLAG-Igf2bp2; immunoprecipitation with FLAG-beads was followed by western blotting and detection using anti-FLAG and anti-GFP. As marked, some of the samples were treated with RNase or benzonase before the IP.

(F) Confocal microscopy of GT1-7 cells expressing GFP-Mkrn3 and dsRed-Igf2bp2; nuclei are stained with Hoechst 33342. Scale bars, indicate 10 μm.

(G) Venn diagram shows overlap between the Mkrn3 RNA targets (in Figure 2A) and RNAs more occupied (p value: 2.99e−106; Chi-square) with monosomes on day 35 compared to day 45 in the ARC (data from³⁴).

(H and I) RIP-qPCR for: (H) FLAG-Igf2bp2 in GT1-7 cells with OE of just GFP or GFP-Mkrn3; or (I) FLAG-Mkrn3 in cells with OE of just dsRed or dsRed-Igf2bp2. Values (mean and individual are shown) are IP/input, n = 2; significant p values from Student’s t test are shown; otherwise, p > 0.05. H2b is a negative control and Ndufa4 a positive control for Igf2bp2. See also Figure S3.

Amongst the 33 most significantly interacting proteins, we identified just two (Figure 3C), Igf2bp2 and Actb (whose mRNA is a downstream target of Igf2bp2³⁷), that have reported connections with puberty and whose expression corelates with that of Mkrn3 during development.¹⁰ Igf2bp2 was particularly notable, not only because of its drop in expression leading up to puberty (Figure 3D; Pearson’s correlation coefficient 0.87), but also given recent reports that it acts in the ARC to delay pubertal onset in mice,³⁴ and its preference also for binding 3′UTRs.^{24,38,39,40,41} Co-immunoprecipitation confirmed the direct interaction of Mkrn3 and Igf2bp2 in transiently expressing GT1-7 cells, and the lack of effect of RNaseA or benzonase confirmed that this interaction is not dependent on RNA, DNA, or RNA-DNA hybrids (Figure 3E). Moreover, the proteins were seen to be colocalized in the cell (Pearson’s correlation coefficient 0.685; Figures 3F and S3B–S3E).

Interestingly, Igf2bp2 also binds numerous nuclear mRNAs that encode mitochondrial proteins⁴² and promotes the recruitment of monosomes to RNA in the ARC.³⁴ We therefore compared the Mkrn3 RNA-targets in the GT1-7 cells with the monosome-enrichment of Mkrn3 RNA targets in the rat ARC, through the analysis of the published data.³⁴ This revealed that 60% of the Mkrn3-targeted RNAs were more enriched with monosomes on postnatal day (PND) 35 as compared to PND 45 (p value 2.99e−106; Figure 3G), correlating with the changes in Mkrn3 levels. We then looked for common RNA targets of these two proteins, drawing on our Mkrn3 RIP data, and enhanced cross-linking and immunoprecipitation (eCLIP) data for Igf2bp2 in human embryonic stem cells (hESCs).⁴³ Despite these very different cell types and source organisms, 143 (23%) of the 603 Mkrn3 RNA targets in GT1-7 cells were also found to be targets of Igf2bp2 in hESCs (p value 0.0001; Chi-square test). Taken together, these findings reveal that Mkrn3 and Igf2bp2 interact physically and bind to common and likely unique RNA targets.

To determine whether Igf2bp2 also binds Gnrh1 mRNA, we performed RIP-qPCR (with FLAG beads) in GT1-7 cells transiently expressing FLAG-Igf2bp2. In WT cells, Igf2bp2 appeared not to bind Gnrh1 mRNA, and levels were similar to those at the negative control H2b (Figure 3H). However, in cells over-expressing Mkrn3, binding was 6-fold higher, and similar to levels to the positive control Igf2bp2 RNA target, Ndufa4. Igf2bp2 binding was not increased (p > 0.05) by Mkrn3 OE at the Ndufa1 or H2b transcripts (Figure 3H). We also conducted the complementary experiment involving RIP-qPCR of FLAG-tagged Mkrn3, with or without Igf2bp2 OE. Igf2bp2 OE increased two-fold the association of Mkrn3 with Gnrh1 mRNA, while its binding to Ndufa4 was not altered (Figure 3I). Together, these observations indicate a functional interaction between these proteins in GnRH neurons, in which Mkrn3 specifically promotes the binding of Igf2bp2 to Gnrh1 mRNA.

Mkrn3 affects the cellular localization of the Gnrh1 precursor protein

The association of these proteins not only with Gnrh1 mRNA, but also with transcripts encoding mitochondrial proteins, suggested to us that Mkrn3 might play a role in RNA transport. While the latter mRNAs are typically translated near their site of function, the mitochondria,^44,45,46 the mature GnRH peptide must reach secretory granules at the axon termini, following the proteolytic cleavage of its precursor protein as it passes through the ER and Golgi apparatus.⁴⁷ We thus considered that Mkrn3 might direct Gnrh1 mRNA translation to a specific site, perhaps hindering proGnRH from entering this pathway.

To address this possibility, we first fractionated cells to separate the various subcellular compartments (Figure S4A). Centrifugation at 3 K and 12 K is expected to enrich for “heavy” and “light” mitochondria, respectively, while other smaller organelles precipitate mostly in the 12 K fraction or resulting supernatant. Proteins from each of these fractions were collected and identified by MS. Various markers confirmed the enrichment, and ER proteins were found mostly in the final supernatant (Figures 4A and S4B). In Mkrn3 OE cells, Mkrn3 was detected only in the 3 K and 12 K fractions (Figure 4B), and proGnRH was also predominantly present in these fractions (Figure 4C). Consistent with our earlier findings (Figure S1D), the proGnRH was found at higher levels in these cells than in the controls. However, its subcellular distribution differed markedly: In control cells, it was enriched in the 12 K supernatant, whereas in Mkrn3-OE cells, the protein failed to reach this fraction (Figures 4D, 4E, and S4C). Further analysis of this and the RIP-seq data revealed that proGnRH was virtually the only protein of those encoded by Mkrn3 RNA targets whose location was altered by Mkrn3 (Figures 4F and 4G). The relative localization of other secreted proteins did not appear similarly affected (Figure S4D).

Figure 4.

Mkrn3 affects the cellular localization of the Gnrh1 precursor protein

(A–E) Proteins in GT1-7 WT (solid bars) and Mkrn3 OE (hatched bars) cell lysates, fractionated at 3 K, 12 K and the final supernatant (Sup), were identified by mass spectrometry. Levels (normalized to the median in the same fraction) of (A) selected organelle markers, (B) Mkrn3 and (C) proGnrh are shown for each fraction in two replicates. (D, E) Also shown are the relative proGnrh levels across fractions in the (D) control and (E) Mkrn3-expressing cells. See also Figures S4A–S4D.

(F and G) Volcano plots show relative levels of (F) the proteins encoded by Mkrn3 RIP-targets (red dots) and non-targets (gray dots; data from Figure 2A) in each of the fractions, and (G) those whose levels across the fractions changed (purple dots) in the Mkrn3 OE cells. The proGnrh1 is marked in all.

(H–K) Fluorescence microscopy of GT1-7 cells transiently expressing a dsRED reporter plasmid without (H and I) or with (J and K) the full-length Gnrh1 3′UTR; some of the cells (I and K) also expressed FLAG-Mkrn3; nuclei are stained with Hoechst 33342, and scale bars indicate 10 μm. See also Figure S4E.

As our findings pointed to a highly specific role for Mkrn3 in regulating the distribution of proGnRH in GnRH neurons, via binding to the Gnrh1 3′UTR, we next examined the localization of a dsRed reporter expressed from the CDS with or without this 3′UTR. Some of the cells were co-transfected with Mkrn3, allowing us to determine the effect of Mkrn3 on the reporter protein localization and the contribution of the 3′UTR. The dsRed was distributed diffusely in all of the cells except those in which Mkrn3 was co-expressed and the dsRed CDS had been fused to the Gnrh1 3′UTR (Figures 4H–4K and S4E). In these cells, the reporter protein was highly localized to distinct foci (Figures 4K and S4E), revealing a role for the Gnrh1 3′UTR in mediating the Mkrn3-directed localization of the translated protein.

Discussion

Following strong clinical evidence implicating MKRN3 in pubertal timing,^1,2 prior studies proposed that it represses the reproductive axis by downregulating gene expression in various hypothalamic neurons.^9,28,30 However, KO models, despite exhibiting early onset of puberty, did not support the proposed mechanisms, leaving unclear the role of Mkrn3 in this context.^28,31 We found that Mkrn3 OE had a more pronounced effect on the proteome than on the transcriptome, with only a weak correlation between the changes in protein levels and their encoding mRNAs. The association between Mkrn3 and post-transcriptional processes was evident in all of our experiments, and it was seen to bind a large number of mRNAs, including, most notably, the Gnrh1 transcript. This identification of Mkrn3 as a functional mRNA-binding protein is in accordance with its domain composition and consistent with demonstrated roles of other MKRN proteins.^{13,24,38,48,49} We report here a highly specific function in which, via the Gnrh1 3′UTR, it alters the subcellular distribution of proGnRH. We propose that this change in location prevents access to the processing machinery required for the conversion of this precursor protein to the mature peptide that activates the reproductive axis.

The enrichment of transcripts encoding mitochondrial proteins among Mkrn3 RNA targets, together with the fact that Mkrn3 was seen to interact with Igf2bp2, which also preferentially binds these transcripts,⁴² pointed to a possible role in transport of RNA to specific subcellular sites. The trafficking of mRNA for localized translation is common in highly polarized cells such as neurons,^45,50,51 and many mRNAs encoding mitochondrial proteins are translated near the outer mitochondrial membrane.^44,52 Indeed, Mkrn3 co-fractionated with mitochondrial proteins, although these fractions also contain markers of other organelles, including early endosomes (markers: Rab5 and EEA1), which are associated with mitochondria in neuronal dendrites and axons,⁵³ and their close interactions and co-precipitation hinder precise identification.⁵⁰ Still, regardless of the exact location, the subcellular distribution of proGnRH was very clearly affected by Mkrn3.

The regulatory effect of Mkrn3 on Gnrh1 mRNA, and perhaps its specific outcome, could relate to its interaction with Igf2bp2. The fact that Igf2bp2 requires Mkrn3 to associate with this transcript suggests a binding mechanism distinct from that of its other targets. Igf2bp2 is known to bind m6A-methylated RNA,⁵⁴ and this modification has roles in the subcellular localization of RNAs in neurons,⁵⁵ presumably mediating Igf2bp2 association with many of its targets. We are not aware of any reports of m6A in the Gnrh1 3′UTR but, even if present, it is clearly not sufficient for Igf2bp2 recruitment to this transcript. It is perhaps feasible that the addition of m6A might contribute to Igf2bp2 binding to its other targets, while its association with Gnrh1 mRNA is directed specifically by Mkrn3.

A prior study reported that Mkrn3 interacts with an Igf2bp2 paralog, Igf2bp1, in CoIP studies performed in human embryonic kidney HEK293 and neuronal SH-SY5Y cells.³¹ A negative correlation was observed between Igf2bp1 and Mkrn3 levels in the mouse ARC, and the authors concluded that the interaction between these proteins contributes to the Mkrn3-mediated inhibition of puberty onset, although a mechanistic explanation was not provided.³¹ This contrasts with the positive correlation between Mkrn3 and Igf2bp2 expression in the brain during rodent development, along with our finding that Mkrn3 enables the binding of Igf2bp2 to Gnrh1 mRNA. In our analysis of various postnatal developmental datasets, Igf2bp1 was found to be expressed at much lower levels than Igf2bp2, and it was not among the Mkrn3-interacting proteins identified in GT1-7 cells, despite being expressed. Notably, Igf2bp2 was also identified as a “high-confidence interacting protein” in both cell types in the previous study,³¹ and in a screen of MKRN3 interacting proteins in HEK cells.⁵⁶ Taken together, these findings suggest that MKRN3 likely interacts with both proteins in a developmental and possibly context-specific manner, dependent also on the association of other RNA-binding proteins, to impart distinct effects.

Although we identified a role for the proximal 31 nt of the Gnrh1 3′UTR for Mkrn3 binding, the predicted motifs in this sequence correspond to broadly expressed RNA-binding proteins (e.g., Hnrnpl and Rbm45) that were not enriched in our AP-MS/MS dataset, and do not explain the selective effect on Gnrh1. Mkrn3 likely exerts transcript-specific effects across its numerous target mRNAs, reflecting not only variation in 3′UTR sequence and interactions with other RNA-binding proteins, but also distinct pathways governing the mRNA transport, localization, and processing. For example, translation of the many transcripts encoding mitochondrial proteins bound by Mkrn3 is unlikely to be affected by its sharp post-pubertal decline, indicating diverse consequences of this association. Thus, although our data support a role for the Gnrh1 3′UTR in Mkrn3 interaction and function, the precise mechanisms underlying this specificity remain to be defined, and we hypothesize that the Gnrh1-specific effects reflect the coupling of Mkrn3-dependent localization to downstream processing.

While the mechanisms through which Mkrn3 regulates its other RNA targets are not yet known, its capacity to direct proGnRH, via the Gnrh1 3′UTR, to a specific subcellular location is striking. Like other neuropeptides, this protein needs to be cleaved by various specific peptidases as it passes through the ER-Golgi pathway before storage of the mature peptide in secretory vesicles.^47,57 The altered localization might affect this processing and trafficking, leading to the accumulation of proGnRH in Mkrn3-expressing cells. This increase in the GnRH precursor protein was unexpected given the repressive effect of Mkrn3 on pubertal onset, but it does not appear to reflect levels of the active hormone.

Numerous studies have focused on transcriptional regulation in the central control of reproduction,²⁶ and here we shed light on a post-transcriptional regulatory mechanism with implications for translation and post-translational processes. Although we discovered the effects of Mkrn3 on Gnrh1 mRNA translation in GnRH neurons, it may well exert similar regulation on mRNAs in the other reproductive control cells. However, any effects in the upstream Kiss1 neurons would be expected to affect also Gnrh1 transcription, which was not consistently seen in the Mkrn3 KO mice. In fact, although we have yet to demonstrate the downstream effects on GnRH secretion and LH pulsatility in an animal model, our findings do resolve the observations of earlier puberty and elevated LH concentrations in the KO mouse models without requiring a change in Gnrh1 mRNA levels. Nevertheless, it is important to note that the main conclusions of this study are based on observations in cells over-expressing Mkrn3, which may differ from WT cells.

More broadly, by identifying Mkrn3 as an RNA-binding protein that regulates the spatial translation of Gnrh1, this study advances our understanding of how localized mRNA translation can impact downstream protein function in polarized cells such as neurons. The specificity of its effect on proGnRH localization, despite its interaction with numerous transcripts, suggests that Mkrn3 acts through a context-dependent mechanism, involving cooperation with factors such as Igf2bp2 that control subcellular RNA fate. Beyond the reproductive axis, MKRN3 may act similarly to play broader roles in directing the intracellular localization of protein synthesis, with possible implications for the biosynthesis and activity of additional neuropeptides.

Limitations of the study

A limitation of this study is that our mechanistic conclusions are based on Mkrn3 OE in GT1-7 cells, a female-derived cell line, and were not tested in male-derived cells. In addition, the downstream effects on GnRH and LH secretion have yet to be validated in vivo.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Philippa Melamed (philippa@technion.ac.il).

Materials availability

Plasmids generated in this study will be distributed subject to restrictions of the original sources of their components.

Data and code availability

• •Data: Accession codes for the data reported in this paper are noted in the key resources table later in discussion, and comprise: RNA datasets at GEO (GSE301395) and Proteomic datasets at ProteomeXchange (PDB: PXD065897, PXD065908 and PXD065952).
• •Code: This paper does not report original code.
• •Other items: Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

We thank staff at the Technion Azrieli Genome Center, Smoler Proteomics Center, and the Life Science and Engineering Infrastructure Center for help with the genomic, proteomic, and microscopy studies, respectively. We thank Rawad Hanna and Oded Kleifeld for advice on the proteomics analysis, and Tamar Lahav for assistance with data handling. Thanks also to Yoav Arava and Ayala Shiber for many fruitful discussions. Funding support was from the Israel Science Foundation.

Author contributions

Conceptualization: D.S., L.P., P.M., and Y.M.G.; data curation: D.S.; formal analysis: D.S. and L.P.; funding acquisition: P.M. and Y.M.G.; investigation: D.S. and L.P.; methodology: D.S., L.P., G.G., A.A., T.B.d.L., G.P., and T.H.; project administration: P.M. and Y.M.G.; resources: D.S., L.P., G.G., and T.B.d.L.; supervision: P.M. and Y.M.G.; visualization: D.S.; writing – original draft: D.S. and P.M.; writing – review and editing: P.M. and Y.M.G.

Declaration of interests

The authors declare that they have no conflict of interest.

Declaration of generative AI and AI-assisted technologies in the writing process

During the preparation of this work, the authors used ChatGPT to improve language, after which the authors reviewed and edited the content, and take full responsibility for the publication.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-GFP antibody	Abcam	Cat#ab290; RRID:AB_303395
Mouse anti-FLAG M2 antibody	Sigma-Aldrich	Cat#F1804: RRID:AB_262044
Goat anti-rabbit HRP-conjugated secondary antibody	Santa Cruz Biotechnology	Cat#sc-2004: RRID:AB_631746
Goat anti-mouse HRP-conjugated secondary antibody	Santa Cruz Biotechnology	Cat#sc-2005; RRID:AB_3717730
Chemicals, peptides, and recombinant proteins
G418	Sigma-Aldrich	Cat#A1720
3x FLAG peptide	Sigma-Aldrich	Cat#4799
Critical commercial assays
Lipofectamine 3000 Transfection Reagent	Invitrogen	Cat#L3000008
RNA Clean & Concentrator-5 kit	Zymo Research	Cat#R1014
qScript cDNA Synthesis Kit	Quanta Biosciences	Cat#95047
2x qPCRBIO Fast qPCR SyGreen Blue Mix	PCR Biosystems	Cat#PB20.16–51
NEBNext library prep kit with ribosomal RNA depletion	New England Biolabs	Cat#E7760
GFP-Trap beads	ChromoTek	Cat#GTA-10
ANTI-FLAG M2 agarose beads	Sigma-Aldrich	Cat#A2220
FLAG magnetic beads	Millipore	Cat#M8823
Mitochondria Isolation Kit for Cultured Cells	Thermo Fisher Scientific	Cat#89874
TRIzol reagent	Ambion	Cat#15596-018
Deposited data
RNA datasets	This paper	GEO: GSE301395
Proteomics datasets	This paper	ProteomeXchange: PXD065897; PXD065908; PXD065952
Experimental models: Cell lines
Mouse: GT1-7 hypothalamic GnRH neuronal cell line	Pamela Mellon (UCSD) available in the lab	N/A
Oligonucleotides
Primers and FISH probes used in this study	This paper	Table S1
Recombinant DNA
Plasmid: pCMV-EGFP (pEGFP-C1)	Clontech; available in the lab	N/A
Plasmid: p3xFLAG-CMV-10	Sigma-Aldrich; available in the lab	N/A
Plasmid: CMV-script	Stratagene; available in the lab	N/A
Plasmid: pLV-CMV-LoxP-dsRED-LoxP-eGFP	Addgene; available in the lab	N/A
Plasmid: FLAG-ATF3 K42R in pXJ40	Available in the lab	N/A
Plasmid: GFP-Mkrn3	This paper	N/A
Plasmid: FLAG-Mkrn3	This paper	N/A
Plasmid: FLAG-Igf2bp2	This paper	N/A
Plasmid: dsRED-Igf2bp2	This paper	N/A
Plasmid: CMV-dsRED-SV40	This paper	N/A
Plasmid: CMV-dsRED-Gnrh1 3′UTR-SV40	This paper	N/A
Plasmid: CMV-dsRED-truncated Gnrh1 3′UTR-SV40	This paper	N/A
Software and algorithms
Galaxy	usegalaxy.org	https://usegalaxy.org/
cutadapt	Martin58	https://cutadapt.readthedocs.io/
RNA-STAR	Dobin et al.59	https://github.com/alexdobin/STAR
HTSeq	Anders et al.60	https://htseq.readthedocs.io/
DESeq2	Love et al.61	https://bioconductor.org/packages/DESeq2/
g:Profiler	Kolberg et al.62	https://biit.cs.ut.ee/gprofiler/
MSFragger/FragPipe	Kong et al.63; Yu et al.64	https://fragpipe.nesvilab.org/
Perseus	Tyanova et al.65	https://maxquant.net/perseus/
R	The R Foundation	https://www.r-project.org/
ImageJ	Fiji (https://imagej.net/ij/)	RRID:SCR_002285
ZEN Microscopy Software	Zeiss (https://www.zeiss.com/microscopy/en/products/software/zeiss-zen.html)	RRID:SCR_013672
Other
Illumina HiSeq 2500 platform	Illumina	N/A
Exploris 480 mass spectrometer	Thermo Fisher Scientific	N/A
Q Exactive Plus mass spectrometer	Thermo Fisher Scientific	N/A
LSM710 confocal microscope	Zeiss	N/A
LSM 980 spectral inverted confocal microscope	Zeiss	N/A
FUSION imaging system	Vilber	N/A
ImageQuant LAS4000 imaging system	GE Healthcare/Cytiva	N/A

Experimental model and study participant details

The GT1-7 hypothalamic GnRH neuronal cell line is derived from adult female C57Bl/6 J x BALB/cJ mice and contains a transgene comprised of the SV40 T-antigen coding sequence driven by the Gnrh1 promoter. They were cultured as previously⁶⁶ in high glucose Dulbecco’s Modified Eagle Medium containing 10% FBS, 1% penicillin-streptomycin and sodium pyruvate (all from Biological Industries, Beit Haemek), maintained at 37 °C with 5% CO2 at 50–90% confluency, passaging 1–2 times a week, and tested routinely for mycoplasma and authentication via marker gene expression. Transient transfections were in cells at 50% confluency, using Lipofectamine 3000 reagent (L3000008, Invitrogen), also as previously.⁶⁷ Stable transfections were performed similarly with AseI- or Pci1-linearized plasmids, and fresh medium replaced after 24 h, followed by selection using G418 (400 μg/mL, Sigma).

Method details

Plasmid construction

The Mkrn3, Igf2bp2 and Gnrh1 sequences were amplified from GT1-7 cDNA. For GFP-Mkrn3, Mkrn3 was amplified (primers mMKRN3_ORF_R_BamH GGATCCCTACAGATTCAAACTGAAATATTC and Mkrn3 ORF_F_HindIII AAGCTTTTATGGAAGAGTCTACAGCTCCCATTG) and cloned into pCMV-EGFP (pEGFP-C1, Clontech) cut with BamHI and HindIII. For FLAG-Mkrn3, Mkrn3 was amplified using the same primers, and cloned into pC3xFLAG-CMV-10 (Sigma) cut with BamHI and HindIII. For FLAG-Igf2bp2, Igf2bp2 was amplified (primers mIGF2bp2 1 F aagcttATGATGAACAAGCTGTACATTG and mIGF2bp2 1779 R ggtaccTCACTTGCTGCGCTGTGGG) and cloned into pXJ40 cut with HindIII and KpnI. For dsRED-Igf2bp2, dsRed was amplified (primers dsRed 1 F GCGGCCGCATGGATAGCACTGAGAACGTC and dsRed 675 R GGATCCCTGGAACAGGTGGTGGCG) from pLV-CMV-LoxP-dsRED-LoxP-eGFP plasmid (Addgene #65726) as template, with Igf2bp2 amplified as above, and these sequences were cloned into CMV-script (Stratagene) cut with Not1 and BamHI for dsRed, and HindIII and KpnI for Igf2bp2.

For dsRED-SV40 polyA, dsRed was amplified (primers dsRed 1 F GCGGCCGCATGGATAGCACTGAGAACGTC and dsRed 674 R catgtcgactaCTGGAACAGGTGGTGGCGG) from the same pLV-CMV-LoxP-dsRED-LoxP-eGFP plasmid, and cloned into CMV-script (Stratagene) cut with Not1 and SalI. For dsRED-Gnrh1 3UTR-SV40, Gnrh1 3UTR was amplified (primers mGnRH1 197 F gctagcATGCACTGGCCCAGGTGGAT and mGnRH1 345 R tctagatTGATCCACAACAAACTACTTTA), cloned into pGEM-T easy, and then cut with Apa1 and SalI to be cloned into CMV-script dsRed (as above). For dsRED-Gnrh1 truncated 3UTR-SV40, Gnrh1 3UTR was amplified (primers mGnRH1 228 F gagctcGAGTATAACATTGACCCTGAG and mGnRH1 345 R tctagatTGATCCACAACAAACTACTTTA), cloned into pGEM-T easy, and then cut with Apa1 and SalI to be cloned similarly into CMV-script dsRed.

RNA extraction and real-time quantitative PCR (qPCR)

Total RNA was extracted using TRIzol (Ambion), treated with DNase and purified using RNA Clean & Concentrator-5 kit (R1014, Zymo Research). The cDNA was synthesized using qScript cDNA Synthesis Kit (Quanta Biosciences) according to the manufacturer’s instructions, and qPCR was carried out using 2x qPCRBIO Fast qPCR SyGreen Blue Mix PB20.16–51(PCR Biosystems) and primers listed in Table S1. Amplicon levels were quantified using standard curves and normalized to the Rplp0 housekeeping gene. Enrichment of the cytosolic RNA was as described.⁶⁸

RNA sequencing

Total RNA and cytosolic RNA were extracted from WT GT1-7 cells and those stably expressing GFP-Mkrn3 as above, for RNA sequencing which was performed using the CEL-seq2 protocol⁶⁹ on an Illumina HiSeq 2500 platform. Sequencing reads were uploaded to the Galaxy web server (https://usegalaxy.org/), processed using the cutadapt software,⁵⁸ aligned to the reference genome (mm10) using RNA-STAR,⁵⁹ and then HTSEQ count⁶⁰ was used to generate raw count data. DESeq2 R package⁶¹ was used for differential gene expression analyses. GO enrichment was analyzed using gprofiler website.⁶²

RIP-seq and RIP-qPCR

GT1-7 cells at 70% confluency (100 mm plates), were transfected with GFP or GFP-Mkrn3 plasmids (6 μg) for 24 h, before lysis with 1 mL of RNase-free RIP lysis buffer. The lysate was passed through a 27-gauge needle 10 times and, after 10 min on ice, cleared by centrifugation (10,000 RPM, 10 min at 4°C). From each sample lysate, 5% was taken as input, and the remainder used for IP with GFP-traps (GTA-10 Chromotek), with incubation overnight (ON) at 4°C. The beads were washed ×4 with RIP lysis buffer before RNA extraction with Trizol, including DNase treatment. Libraries were prepared with the NEB kit (NEBNext #E7760) using standard protocol with ribosomal RNA depletion. The libraries were sequenced on an Illumina HiSeq 2500 platform. The raw FASTA files were analyzed by cutadapt, with alignment to the mouse reference genome (mm39) using RNA-STAR, and then HTSEQ count. Alternatively, cells were prepared similarly, and RNA was precipitated using FLAG magnetic beads (M8823, Millipore), or FLAG antibody (F1804 Sigma) with proteinG magnetic beads (10004D Invitrogen), after FLAG-Mkrn3 or FLAG-Igf2bp2 OE. The cDNA was synthesized and diluted 1:4; 1 μL was then used in qPCR reactions, which included positive and negative transcripts.

Tandem mass spectrometry (MS/MS) and proteomic analysis

WT GT1-7 cells and those stably expressing GFP-Mkrn3 (3 × 100 mm plates) were lysed with NP-40 lysis buffer, cleared and diluted in the same lysis buffer without NP-40, before tryptic digestion and analysis by liquid chromatography–mass spectrometry (LC-MS/MS) on Exploris 480 mass spectrometer (Thermo) at the Smoler proteomics center (Technion).

The raw files were processed with MSFragger (fragpipe.nesvilab.org/docs/tutorial_fragpipe. html) using the label-free quantification with match-between-runs (LFQ MBR) workflow and default parameters.^63,64 Partial cleavage, loss of ammonia at peptide N-termini (nQnC) and ubiquitination were considered, and data aligned to the Mus musculus proteome (Uniprot ID: UP000000589), with a decoy database to determine the false discovery rate (FDR). Identified peptides were filtered with high confidence, top rank, mass accuracy, and high confidence peptides passed the 1% FDR threshold. Downstream analysis was performed with Perseus (maxquant.net/perseus/),⁶⁵ MaxLFQ values were log2 transformed and filtered to include proteins detected in at least two replicates. Missing values were imputated from a normal distribution followed by a two-sample t test. A permutation-based FDR with S0 value of 0.1 was applied. Gnrh1 peptide levels were determined by summing the peak areas of peptide-spectrum matches (PSMs), and each peptide intensity was normalized to that of a reference Gapdh peptide.

Affinity purification mass spectrometry (AP-MS/MS)

GT1-7 cells stably expressing GFP or GFP-Mkrn3 were lysed in RIPA lysis buffer, and immunoprecipitation performed using GFP traps (GTA-10 Chromotek) ON, according to the manufacturer’s protocol. The immunoprecipitated proteins were washed 3 times and eluted with Lammeli sample buffer before separation on a 12.5% gel. After staining, 2.5 cm gel fragments were excised for each sample and taken for tryptic digestion and analysis in a Q Exactive plus mass spectrometer (Thermo), also at the Smoler proteomics center. The raw files were analyzed using MSfragger, as above, run separately for the IP and control samples, ensuring that the match between runs (MBR) option did not compare between these samples which are supposed to be different. The outputs were then combined using a custom R script, and downstream analysis was performed on Perseus, as described above.

Co-immunoprecipitation

After (24 h) transient OE of GFP-Mkrn3 and FLAG-Igf2bp2, the GT1-7 cells were lysed in FA-lysis buffer (with PMSF+ and protease inhibitor cocktail), and IP performed using anti FLAG M2 agarose beads (Sigma, A2220). The proteins were eluted with 3X FLAG peptide (Sigma, 4799) and separated on 10% SDS-PAGE before transfer to nitrocellulose membranes. After blocking (TBST with 5% BSA), the membranes were incubated with antibody for GFP (Abcam, ab290); or FLAG (Sigma,F1804), for 1 h at RT. The blot was washed three times with TBST, and then incubated with secondary HRP conjugated antibody (goat anti-rabbit HRP, Santa Cruz sc-2004, goat anti-mouse HRP, Santa Cruz sc-2005) at 1:10000 for 1 h at RT. Membranes were washed with TBST and proteins were detected using SuperSignal West Pico PLUS (ECL) kit (Thermo Fisher Scientific) and analyzed with the FUSION or Imagequant LAS4000 imaging system. For RNAse/benzonase treatment: before IP, the lysate was treated for 45 min with benzonase (1:100 with cat#70664 Sigma) or 1 h with RNAseA (0.5 μg/μL final, cat#R4642 Sigma) on a roller at 4 °C.

Microscopy and fluorescence in situ hybridization (FISH)

For live cell imaging, cells were plated on a 35 mm cell culture imaging dish with coverslip bottom (ibidi, 81156), and nuclei stained with Hoechst 33342 (10 min at RT). Cells were visualized using an LSM710 confocal microscope (Zeiss), and images analyzed using Fiji and Zen software. Alternatively, cells expressing tagged proteins and plated on coverslips in 24 well plates, were fixed (4% formaldehyde for 15 min at RT) and incubated with cold 70% ethanol at 4°C for at least 24 h. After staining of the nuclei with Hoechst (10 min at RT), they were mounted, examined as above and analyzed using Fiji software.

For FISH, cells prepared as above were washed with PBS, and incubated in wash buffer (15% formamide, 1X SSC) for 15 min in RT. FISH probes recognizing Gnrh1 mRNA (Table S1) were hybridized to probes labeled with Cy5 (Cy5Cy5/AATGCATGTCGACGAGGTCCGAGTGTAA/Cy5Cy5). The labeled probes were diluted in DDW to 2.5 μM, and 5 μL added to 45 μL of hybridization buffer, for incubation with each sample ON at 37°C. The next day, samples were washed twice with PBS, and mounted together with 1:200 of DAPI (10 μg/mL) on a glass slide. Samples were visualized on a spectral inverted confocal LSM 980, analyzed as above.

Cell fractionation for mitochondrial enrichment

Mitochondria separation was performed using the Mitochondria Isolation Kit for Cultured Cells (Thermo, 89874), according to the manufacturer’s instructions. After the initial 3,000 g precipitation for the major “heavy” mitochondrial fraction, an additional precipitation at 12,000 g was performed on the supernatant to enrich for “light” mitochondria and other organelles such as peroxisomes and lysosomes. The proteins in these two pellets, as well as in the final supernatant, were digested by trypsin and analyzed on Exploris 480 mass spectrometer (Thermo), as described above. The intensity values were normalized to the median of each sample, and Student’s t test (two-tailed) performed to establish significance.

Quantification and statistical analysis

The experimental data are from multiple biological repeats (n-value, stated in figure legends) which were assayed individually. DESeq2 R package⁶¹ was used for differential gene expression and RIP-seq analyses, and all proteomic analysis was performed with Perseus, both as detailed above. F-test determined similarity of variance and data analyzed accordingly by Student’s t test (two-tailed) to establish significance, defined as p < 0.05. Chi-square was used to determine significance of the overlaps between Mkrn3 RIP targets and RNAs encoding mitochondrial proteins, Igf2bp2 eCLIP targets, and RNAs whose monosome occupancy changes along rat development. The analysis was done using the “chisq.test” function in R. Minimal hypergeometric test (mHG) determined whether the total or cytosolic RNA differed significantly among the up or downregulated proteins, performed using the mhg package in R. Wilcoxon rank test, to assess the significance of differences in the lengths of Mkrn3 RNA targets compared to non-targets was also done in R. Correlation of developmental gene expression levels, and of colocalization of proteins was performed using Pearson correlation coefficient (in R): gene expression based on the set of average values at each timepoints (n = 4) normalized and transformed (log2(CPM+4)) counts. For protein colocalization in cells, fluorescence intensity along an arbitrarily drawn line was measured for each channel and Pearson correlation coefficients calculated between each intensity profile along the line.