MESH1 knockdown triggers proliferation arrest through TAZ repression

All organisms are constantly exposed to various stresses, necessitating adaptive strategies for survival. In bacteria, the main stress-coping mechanism is the stringent response triggered by the accumulation of “alarmone” (p)ppGpp to arrest proliferation and reprogram transcriptome. While mammalian genomes encode MESH1—the homolog of the (p)ppGpp hydrolase SpoT, current knowledge about its function remains limited. We found MESH1 expression tended to be higher in tumors and associated with poor patient outcomes. Consistently, MESH1 knockdown robustly inhibited proliferation, depleted dNTPs, reduced tumor sphere formation, and retarded xenograft growth. These antitumor phenotypes associated with MESH1 knockdown were accompanied by a significantly altered transcriptome, including the repressed expression of TAZ, a HIPPO coactivator, and proliferative gene. Importantly, TAZ restoration mitigated many anti-growth phenotypes of MESH1 knockdown, including proliferation arrest, reduced sphere formation, tumor growth inhibition, dNTP depletion, and transcriptional changes. Furthermore, TAZ repression was associated with the histone hypo-acetylation at TAZ regulatory loci due to the induction of epigenetic repressors HDAC5 and AHRR. Together, MESH1 knockdown in human cells altered the genome-wide transcriptional patterns and arrested proliferation that mimicked the bacterial stringent response through the epigenetic repression of TAZ expression.


INTRODUCTION
All organisms are constantly exposed to a wide variety of environmental stresses. For example, most solid tumor cells experience hypoxia [1,2], lactic acidosis [3,4], and nutrient deprivations [5][6][7][8][9], necessitating stress responses to ensure survival and homeostasis. In bacteria, the main stress mechanism is "stringent response" mediated by elevated alarmone (p)ppGpp via synthesis by its synthetase (RelA), or inhibited degradation by its hydrolase (SpoT). The accumulated (p)ppGpp triggers transcriptional reprogramming and proliferation arrest [10] through suppression of de novo dNTP synthesis [11,12]. While a similar stress response has not been reported in metazoan, metazoan genomes encode MESH1 (Metazoan SpoT Homolog 1, also named HDDC3). The difficulties are embedded in the fact that neither (p)ppGpp nor RelA exists in metazoan cells. A previous study [13] described the conserved structures of Drosophila and human MESH1 and their in vitro enzymatic activities to degrade (p)ppGpp. Genetic depletion of Drosophila MESH1 enhanced stress survival and triggered transcriptional responses bearing significant similarities to bacterial stringent response [13]. Recently, we have shown that MESH1 is a cytosolic NADPH phosphatase and MESH1 knockdown protected cancer cells against ferroptosis [14] and triggered integrative stresses response. However, much remains unknown about the phenotypic responses to MESH1 knockdown.
Here, we report that MESH1 knockdown robustly arrests proliferation, depletes dNTP, inhibits tumor spheres, and xenograft growth. Mechanistically, these anti-growth effects are mediated by the transcriptional repression of HIPPO effector TAZ (encoded by WWTR1, WW domain containing transcription regulator 1), known to regulate proliferation and tumorigenesis. In addition, the transcriptional repression of TAZ is associated with a repressive chromatin modulation by the induction of HDAC5 (histone deacetylase 5) and AHRR (arylhydrocarbon receptor repressor). The regulation of HDAC5, AHRR, TAZ, and cell proliferation depends on the enzymatic activities of MESH1. Together, our results indicate that MESH1 knockdown in human cells triggers proliferation arrest through epigenetically repressing TAZ transcription, and may hold antitumor therapeutic potential.

RESULTS
MESH1 knockdown reduced cell number, proliferation, and tumor growth The polymorphism in the 3′UTR of MESH1 was found associated with a risk of breast cancers, implying its relevance to tumor biology [15]. Interestingly, TCGA datasets revealed that MESH1 was generally expressed at higher levels in tumor than non-tumor tissues ( Supplementary Fig. 1a). Therefore, we investigated the phenotypic response of cancer cells to MESH1 knockdown by multiple siRNAs (Supplementary Fig. 1b, c). significantly reduced the cell number of non-small cell lung cancer (H1975) (Fig. 1a) under different serum levels tested by crystal violet staining. We expanded this finding in a broader panel of cancer cell lines, including renal cell carcinoma (RCC4, 786-O), breast cancer (BT20, BT474, MCF-7), chondrosarcoma (SW-1353), and fibrosarcoma (HT-1080), and noticed a consistent cell number reduction by MESH1 knockdown (Supplementary Fig. 1d). The antitumor effects of MESH1 knockdown in H1975, RCC4, and 293 T were further confirmed using CellTiter-Glo assay ( Supplementary  Fig. 1e). In essence, the effects were specific since they were abolished by the reexpression of the wild-type MESH1 (Supplementary Fig. 1f). Interestingly, the enzymatic dead mutant MESH1 did not restore the cell number ( Supplementary Fig. 1f). And the knockdown of NAD(H) kinase (NADK), mediating the NADPH generation opposite to MESH1's activity, largely mitigated the reduced cell growth (Supplemental Fig. 1g). These findings indicated the importance of MESH1 NADPH phosphatase activity in regulating cell growth. We further validated that MESH1 knockdown by doxycycline (Doxy)-inducible shRNAs (Supplementary Fig. 1h, i) also triggered growth inhibition tested by cell counting (Fig. 1b) and the Incucyte ® S3 over time ( Fig. 1c and Video 1a-c). Importantly, upon the Doxy withdrawal and restoration of MESH1 expression, cell growth resumed (Fig. 1d). Reciprocally, wild-type MESH1 expression enhanced proliferation, while the effect of the enzymatic dead mutant MESH1 was much reduced (Fig. 1e).
Next, we tested the anti-growth effects of MESH1 removal in 3D tumor spheres and xenografts. The doxycycline-induced MESH1 knockdown significantly reduced the number and size of tumor spheres (Fig. 1f). We then dissipated primary spheres and seeded the same number of cells for the secondary assays, which further confirmed the continuous suppression of the tumor sphere formation by MESH1 knockdown (Fig. 1g). Similar results were obtained in vivo using subcutaneous H1975 xenografts. The induction of two MESH1 shRNAs significantly reduced the volume and weight of tumors (Fig. 1h). Collectively, inducible MESH1 knockdown inhibited tumor growth both in vitro and in vivo.
Mechanistically, we investigated whether MESH1 knockdown (1) inhibited cell proliferation (BrdU incorporation assay), or (2) enhanced cell death (CellTox Green assay) to reduce cell number. MESH1 knockdown significantly reduced the BrdU incorporation (proliferation) (Fig. 1i) but did not increase cell death (Supplementary Fig. 1j). Additionally, the propidium iodide (PI) stain suggested that MESH1 knockdown inhibited the G1-S phase transition (Fig. 1j) without increasing the sub-G1 population, supporting the lack of significant cell death. Collectively, MESH1 knockdown inhibited cell proliferation to achieve anti-growth effects.
The prognostic significance of MESH1 in tumor expression datasets Based on the antitumor phenotypes associated with MESH1 knockdown, we analyzed the prognostic significance of MESH1 (HDDC3) expression in the tumor datasets. Results suggested that lower MESH1 mRNA expression was associated with significantly better survivals in patients with renal cell carcinoma (RCC) ( Supplementary Fig. 2a Fig. 2f), meningioma (GSE16581) [18] (Supplementary Fig. 2g), and follicular lymphoma (GSE16131) [19] (Supplementary Fig. 2h). These in vivo associations of low MESH1 levels and better clinical outcomes suggested a functional role of MESH1 in tumor biology.
Transcriptional responses and dNTP depletion upon MESH1 knockdown One prominent feature of bacterial stringent response is transcriptional reprogramming [10,20]. MESH1 removal in Drosophila downregulated the DNA and protein synthesisrelated gene expression [13]. Therefore, we analyzed the global transcriptional responses to MESH1 knockdown (Fig. 2a) in H1975 cells. 994 genes were selected based on the filtering criteria of at least seven observations with absolute log 2 values >0.47, and arranged by clustering (Fig. 2a) and the differentially expressed genes were shown in Table 1. Gene Set Enrichment Analysis (GSEA) indicated that MESH1 knockdown inhibited multiple cell cycle progression-related pathways (Fig. 2b), including repressing the expression of CDC6 (cell division cycle 6), and CDK1 (cyclindependent kinase 1) (Fig. 2a), consistent with the proliferation arrest phenotypes (Fig. 1). MESH1 knockdown also induced the expression of HDAC5 and HDAC9 (histone deacetylases), Fig. 1 MESH1 knockdown led to cell proliferation arrests and delayed xenograft growth. a Representative image of the crystal violet staining (left) and quantification for replicates (right) of H1975 cells under different serum concentrations showed consistent cell number reduction upon MESH1 knockdown. Higher absorbance represents higher cell numbers. (mean + s.d.). b Cell number measurement by hemocytometer on Days 4, 7, and 10 revealed reduced cell number upon MESH1 knockdown by doxycycline-inducible shRNA of MESH1 in RCC4 cells. (mean ± s.d.). c Image-based quantification of cell count and representative fluorescence-labeled images (right) after 5 days revealed that MESH1 knockdown significantly reduced cell numbers. H1975 were stably transduced with histone H2B-mcherry before being transfected with the indicated siRNAs. Images and fluorescence intensity were obtained and quantified by the Incucyte S3 every 8 h for 5 days after siRNA treatment. Fluorescence intensity (quantified by the Incucyte S3) was proportional to the proliferating cell number. (mean ± s.d.). d Cell number measurement by hemocytometer on Days 4, 7, 10, 13, and 16 revealed that cell number reduction upon doxycycline-induced MESH1 knockdown in RCC4 cells was reversible when doxycycline was withdrawn and MESH1 protein was reexpressed confirmed by the Western blot. Estimated cell division frequency per 24 h were calculated based on cell count (log2(fold change)/days). (mean ± s.d.). e Imagebased quantification of fluorescence intensity by Incucyte S3 in RCC4 revealed that wild-type MESH1 expression largely stimulated cell growth. The enzymatic dead mutant MESH1 had a reduced capacity to enhance cell growth (mean ± s.d.). f Representative image (left) and quantification (right) of the primary tumor sphere formation assay revealed that MESH1 knockdown reduced both the number and the size of H1975 tumor spheres. Colors of bars from lighter to darker successively represent tumor spheres with diameters of >400 μm (V), 300-400 μm (IV), 200-300 μm (III), and 100-200 μm (II). (mean + s.d.). g Representative image (left) and quantification (right) of the secondary tumor sphere formation assay confirmed the same reduction in number and size of H1975 tumor spheres by MESH1 knockdown. The colors of bars represent the same category as (e). (mean + s.d.). h Tumor size and weight measurement showed xenograft growth inhibition upon doxycycline-induced MESH1 knockdown in the xenografted tumor model. p value was calculated by the one-way ANOVA followed by Tukey's posttest. (mean + s.d.). i Representative image and quantification for the BrdU incorporation assay before and after the doxycycline-induced MESH1 knockdown in H1975 revealed a significant reduction of the BrdU incorporation rate upon MESH1 knockdown. Scale bars: 200 μm.
(mean + s.d.). j Cell cycle distribution of H1975 upon MESH1 knockdown by PI stain. The percentage of cells in each individual stage was calculated by the FlowJo software and labeled in the histograms correspondingly. Bar graphs represent the average percentage of cells in each individual stage in each group. siMESH1 reduced the portions of both S and G2 + M phases and arrested cells in the G0 + G1 phase. (mean + s.d.). For a; f; g; i; j, p values were calculated by the two-way ANOVA followed by Tukey's posttest. *P~(0.01, 0.05); **P~(0.001, 0.01); ***P~(0.0001, 0.001); ****P < 0.0001; NS no significance.
implicating epigenetic modulations. Additionally, MESH1 knockdown repressed the expression of RRM1 (ribonucleotide reductase M1) and RRM2 (Fig. 2c), two subunits of ribonucleotide reductase (RNR) responsible for de novo synthesis of dNTPs [21,22]. As ppGpp depletes dNTPs in bacteria [12], we measured the dNTP levels in H1975 and RCC4 (Fig. 2d) and found that MESH1 knockdown largely reduced the levels of all four dNTPs. Together, these results suggested that MESH1 knockdown depleted dNTPs and repressed the expression of genes regulating dNTP synthesis and cell cycle progression. MESH1 knockdown repressed TAZ mRNA expression that contributed to the antitumor phenotypes Interestingly, we noted consistent repression of WWTR1 (WW domain containing transcription regulator, which encodes TAZ (transcriptional coactivator with the PDZ-binding motif) (Fig. 2a). TAZ and its paralog YAP (Yes-associated protein) are wellrecognized HIPPO effectors that regulate proliferation and selfrenewal of cancer cells [23,24]. While YAP/TAZ are conventionally co-regulated by protein phosphorylation, and translocation [23,24], MESH1 knockdown reduced TAZ mRNA level in all tested cancer cells (Fig. 2e, f and Supplementary Fig. 3a). Importantly, wild-type, but not mutant, MESH1 restoration mitigated TAZ repression, showing the specificity ( Supplementary Fig. 3b). The rescue by NADK knockdown further reiterated the importance of MESH1 enzymatic activity ( Supplementary Fig. 3c). While TAZ levels are affected by cell density at the posttranslational level [25][26][27][28], we found that MESH1 knockdown reduced TAZ mRNA to comparable degrees in H1975 and RCC4 grown with low, medium, or high cell density ( Supplementary Fig. 3d). Similarly, MESH1 knockdown did not alter the nuclear/cytosolic distribution of TAZ measured by fractionation/Western blots ( Supplementary Fig. 3e) and immunofluorescence ( Supplementary Fig. 3f). Consistently, GSEA analysis indicated that TAZ-, but not YAP-, regulated pathways [29] were significantly depleted upon MESH1 knockdown in H1975 ( Supplementary Fig. 3g). In the five indicated tumor datasets, tumors with low MESH1 expression levels also displayed low TAZ expression ( Supplementary Fig. 3h), consistent with the regulatory relationships established in cultured cells. However, such a relationship was not found in other tumor datasets, which may be due to the various confounding environmental factors. Collectively, these data indicated that MESH1 knockdown specifically repressed TAZ mRNA and pathway activity by a noncanonical mechanism. Next, the contribution of TAZ repression to various phenotypic alterations of MESH1 knockdown was investigated by restoring the constitutively active TAZS89A [15,30] (Supplementary Figs. 4a, 5a). TAZ restoration significantly rescued the reduced cell number (Supplementary Fig. 4b) and BrdU incorporation (Fig. 3a) upon MESH1 knockdown. In contrast, TAZ depletion completely ablated the stimulation of cell growth by MESH1 overexpression (Fig. 3b). Consistently, TAZS89A expression partially rescued the dNTP depletion (Fig. 3c) and cell cycle arrest by MESH1 knockdown (Fig. 3d, e; gating strategy shown in Supplementary Fig. 4c).
TAZ is known to regulate the self-renewal capacity of cancer cells and tumor growth [23]. Indeed, TAZS89A increased the number of spheres by~92 folds in the shMESH1 groups, whereas only~2 folds in the control (Fig. 3f, g), a trend persisted in the secondary sphere formation assay ( Supplementary Fig. 5b). Furthermore, the limiting dilution assay showed that MESH1 knockdown massively reduced the stem cell frequency from 1/63 to 1/1869 in H1975, and was partially restored by TAZS89A expression to 1/112 (Fig. 3h). Next, we tested the potential of TAZ restoration in mitigating the tumor inhibition effects of MESH1 knockdown in xenografts. We found that the in vivo expression of TAZS89A in the MESH1-removed xenografts rescued the tumor volume and weight by~3 folds ( Fig. 3i and Supplementary Fig. 5c). Collectively, these data indicated the importance of TAZ repression to the dNTP depletion, proliferation arrest, reduced sphere formation, and tumor growth inhibition phenotypes associated with MESH1 knockdown. However, TAZ restoration only partially reversed these antitumor phenotypes of MESH1 removal, indicating that TAZ repression may not be the only factor that contributed.
TAZ downregulation contributed to the transcriptional changes of MESH1 knockdown As TAZ functions as a transcription coactivator, we determined the role of TAZ repression in regulating the transcriptional responses to MESH1 knockdown (Fig. 4) by comparing the MESH1-knockdown gene signatures between the control and TAZS89Atransfected groups (Fig. 4a). While the transcriptional responses were similar, TAZ restoration mitigated the MESH1 knockdownaffected changes in~33% (258 out of 786 genes) of both the repressed (123 of 434 genes) and induced genes (135 of 352 genes) (Fig. 4b), and the rescued gene lists were included in Table  2. Interestingly, the MESH1-affected genes restored by TAZ expression included several known cell cycle-related genes, such as CDC6, CDK1, RRM1, RRM2 (Fig. 4c), CCNE2 (Cyclin E2), KLF2 (Kruppel-like factor 2), and KLF4 (Kruppel-like factor 4) [25,31,32]. ChIP-qPCR further showed the binding of TAZ to the regulatory regions of CDC6 and RRM2 (Fig. 4d) as a transcriptional effector. Overall, these results indicated that TAZ repression mediated~1/3 of the transcriptional response to MESH1 knockdown, including the downregulation of many cell cycle-related genes that contributed to the anti-growth effects (Fig. 3).
Next, we investigated the epigenetic modulations on TAZ measuring H3K27ac (canonical active enhancer mark) levels at TAZ promoter/enhancer regions annotated by the GeneHancer [33] (Supplementary Fig. 6a). MESH1 knockdown significantly reduced the abundance of H3K27Ac at two separate loci at the promoter/enhancer region 2 of TAZ (Fig. 4e), suggesting a closed structure of the TAZ chromatin. Comparatively, H3K27Ac at the YAP promoter or the heterochromatin region was not significantly affected. Furthermore, NADK knockdown reversed the histone hypo-acetylation at TAZ promoter/enhancer region 2 (Supplementary Fig. 6b), implicating a key role of MESH1 enzymatic activity in employing its epigenetic modulation on TAZ. Importantly, epigenetic activation of TAZ by the CRISPR activation (CRISPRa) ( Supplementary Fig. 6c), which recruited a transcriptional activation domain by two individual sgRNAs (designed by the SAM Cas9 activator design tool [34]) to TAZ promoter/ enhancer region 2, also rescued the repressed expression of several TAZ target genes ( Supplementary Fig. 6d, e). Finally, TAZ re-activation significantly mitigated the proliferation arrest (Supplemental Fig. 6f) upon MESH1 knockdown. Altogether, these data revealed the critical role of TAZ epigenetic repression in inhibiting TAZ downstream gene expression and proliferation phenotypes of MESH1 knockdown. Fig. 2 The transcriptional response to MESH1 knockdown. a Heatmap of the selected genes whose expressions were significantly affected by all three MESH1-targeting siRNAs in H1975 cells. Cells were treated with the nontargeting siRNA or three distinct MESH1 targeting siRNAs and triplicates for each treatment were collected. Filtering criteria resulted in 336 probesets: at least seven observations with absolute value ≥0.65. b GSEA analysis of the inhibition of cell cycle-related and Myc-targeted genesets in H1975 cells upon MESH1 knockdown. Gene signatures were downloaded from the GSEA Molecular Signatures Database. Hallmark_G2M _Checkpoint: p = 0.0076, q = 0.16; Hallmark_Myc_Targets: p = 0, q = 0.014. c qRT-PCR validation of the reduced RRM1 and RRM2 mRNA in the MESH1-knockdown H1975 cells. (mean + s.d.). d dNTP measurement by MS-based analysis shows that MESH1 knockdown reduced the levels of all four measured dNTPs in H1975 and RCC4. (mean + s.d.). e qRT-PCR validation of the repressed TAZ mRNA expression upon MESH1 knockdown in H1975 and RCC4 cells. (mean + s.d.). f Representative image of the western blot, which validated that MESH1 knockdown repressed TAZ protein levels in H1975 and RCC4 cells. For c, p values were calculated by the two-tail student's t-test. For d, p values were calculated by the two-way ANOVA followed by Tukey's posttest. For e, p values were calculated by the one-way ANOVA followed by Tukey's posttest. **P~(0.001, 0.01); ***P~(0.0001, 0.001); ****P < 0.0001.            Histone deacetylases (HDACs) are known to reduce histone acetylation and form inhibitory complexes to repress gene expression [35]. Interestingly, MESH1 knockdown induced HDAC5 and HDAC9 expression ( Supplementary Fig. 7a, b), which might contribute to the TAZ repression. Indeed, the pan-HDAC inhibitor (Trichostatin A or TSA) reversed TAZ repression upon MESH1 knockdown (Fig. 4f). To identify the specific HDAC(s) involved in TAZ regulation, we knocked down HDAC1, 2, 3, 5, and 9 individually and measured their effects on TAZ repression upon MESH1 knockdown. Among all tested HDACs, only HDAC5 knockdown rescued TAZ repression (Fig. 4g), consistent with the rescued effect of the HDAC5-specific inhibitor LMK235 (Fig. 4h). Since HDAC5 activity can be regulated by nuclear translocation [36], we fractionalized the nucleus vs. cytosol and found that MESH1 knockdown also dramatically increased the nuclear fraction (putative transcriptionally active) of HDAC5 (Fig. 4i). Together, these data indicated that the increased levels and activities of HDAC5 by MESH1 knockdown contributed to the TAZ repression.
HDAC5 and AHRR formed a repressing complex to inhibit TAZ transcription Here, we aim to identify the HDAC5-interacting transcription factor(s) that regulates TAZ expression. Toward this goal, we compared the list of 48 published HDAC5 interactors (BioGRID database: [37]) with all eight top documented TAZ-binding transcription factors (QIAGEN CLC Genomics Workbench [38]). Such a comparison identified AHRR (aryl-hydrocarbon receptor      repressor) and AHR (aryl-hydrocarbon receptor) as potential candidates. AHRR was induced by MESH1 knockdown (Supplementary Fig. 7c) and reported to be associated with HDAC5 as a repressing complex component [39] that competes with and inhibits the function of AHR [39]. AHR is predicted to target the regulatory regions of TAZ by QIAGEN CLC Genomics Workbench [38]. Therefore, we sought to determine the potential of AHRR and AHR in the regulation of TAZ by MESH1.  First, we confirmed that the restoration of wild-type, but not mutant, MESH1 abolished HDAC5 and AHRR induction (Supplementary Fig. 7d). Next, co-knockdown of NADK and MESH1 also diminished the induction of HDAC5 and AHRR ( Supplementary   Fig. 7e), suggesting the relevance of MESH1's enzymatic activity. Second, we validated the interaction between HDAC5 and AHRR by co-immunoprecipitation in H1975 cells transfected with HAtagged AHRR and Flag-tagged HDAC5 (Fig. 5a). Third, AHRR knockdown (Supplementary Fig. 7c) substantially rescued TAZ repression by MESH1 knockdown (Fig. 5b). Consistently, the knockdown of either HDAC5 or AHRR increased the expression of CDK1, CDK6, RRM1, and RRM2 (Fig. 5c) and the cell number (Fig. 5d). Overall, these data indicated that MESHI1 knockdown upregulated HDAC5 and AHRR, forming a repressing complex, to inhibit the expression of TAZ and its target genes, leading to the proliferation arrest. Since AHR has been reported to bind at TAZ regulatory regions and promote TAZ transcription [26,29], we investigated the potential that HDAC5 and AHRR could repress TAZ expression by inhibiting the binding of AHR to TAZ regulatory regions. The canonical AHR-responsive element (AHRE) contains the core sequence 5′-GCGTG-3′. Among the predicted AHRE sites in the five TAZ promoter/enhancer regions (Fig. 5e, annotated by the GeneHancer [33]), two were validated by ChIP-qPCR (Fig. 5f). More importantly, MESH1 knockdown reduced the AHR occupancy at TAZ regulatory regions (Fig. 5f), consistent with the TAZ repression. Furthermore, AHR knockdown ( Supplementary Fig. 7g), similar to MESH1 knockdown, repressed TAZ (Fig. 5g) and proliferative gene expression (Fig. 5h) and inhibited cell proliferation (Fig. 5i) and tumor growth in vivo ( Fig. 5j and Supplementary Fig. 7h, i). Importantly, tumors in the indicated four datasets with lower MESH1 expression have higher levels of HDAC5 ( Supplementary Fig.  7j) in patients. In the tumor RNA-seq dataset that contained AHRR expression levels, tumors with low MESH1 expression also displayed a high level of AHRR ( Supplementary Fig. 7k). Altogether, the abovementioned evidence supported the idea that HDAC5 and AHRR upregulation upon MESH1 knockdown mediated the proliferation arrest phenotypes by reducing AHR binding and TAZ transcription. In conclusion, we proposed a model by which MESH1 knockdown exerts antitumor effects through repressing TAZ mRNA expression via an epigenetic modulation achieved by HDAC5-AHRR upregulation (Supplementary Fig. 8).

DISCUSSION
While metazoan genomes encode MESH1, the homolog of bacterial SpoT, our understanding of its functional role remains limited. Here, we have shown that MESH1 knockdown triggered reproducible proliferation arrest and altered transcriptional patterns that mimicked the reduced proliferation phenotype of bacterial stringent response. First, the cell cycle arrest was associated with the repression of cell cycle gene expression, similar to the transcriptional feature of the stringent response [20]. Second, MESH1 knockdown depleted dNTPs, reminiscent of the bacterial ppGpp-mediated dNTP depletion [11,12]. However, MESH1 knockdown also carried some distinct phenotypes, including TAZ repression via epigenetic modulations mediated by the induction of HDAC5/AHRR repressor complex. Therefore, our studies presented an interesting example in which the same protein homologs mediated similar phenotypes in different organisms via re-wiring through distinct substrates and mechanisms.
The Hippo signaling pathway exerts profound effects on cellular proliferation, survival, cell death, and organ sizes [25-28, 40, 41]. YAP and TAZ, two Hippo effectors, are usually tightly co-regulated by the phosphorylation of the kinase cascade of MST1/2, LATS1/2, and RASSF family proteins. Hence, our findings discovered a novel and interesting features. First, only TAZ, but not YAP, was affected by MESH1 silencing. In fact, the YAP level was slightly increased, which might reflect a negative feedback loop of HIPPO activity noted in previous studies [42,43]. Second, TAZ repression occurred at the mRNA instead of the posttranslational level. Previously, ETS (E26 transformation-specific) and MRTF/SRF have been reported to be involved in the transcriptional activation of TAZ mRNA [44,45]. These results implied that MESH1 was also required to maintain the TAZ mRNA expression, the activity of the HIPPO pathway, and thus cellular proliferation. Interestingly, many components of the HIPPO pathway (including YAP and TAZ) first emerge in cnidarians, a very ancient group of metazoans [46]. All major domains of YAP and TAZ are also conserved between cnidarians and mammals. Given that MESH1 also shares the conserved domains with the bacterial hydrolase SpoT, it is tempting to speculate the functional convergence between MESH1 and the HIPPO pathway as they both appear in metazoans during evolution.
MESH1 was a cytosolic NADPH phosphatase [14]. Here, we discovered that the enzymatic activity is essential for the HDAC5-AHRR upregulation, TAZ repression, and antitumor effects. However, the underlying mechanisms are still unknown. In the future, it will be critical to investigate the connection between NADPH and transcriptional control of the epigenetic repressing complex. Furthermore, in bacteria, different environmental stresses (i.e., amino acid deprivation, heat shock, iron deficiency, etc.) trigger the ppGpp accumulation. While MESH1 depletion leads to a stringent-like stress response, it is unclear which external stimuli activate the stringent response in metazoan, which requires further investigation. Despite these knowledge gaps, supported by the extensive amount of data, we demonstrated that MESH1 inhibition potentially represented a novel anti-growth response in cancer cells with significant therapeutic potentials. We can take advantage of the evolutionarily conserved pathways, proteins, and metabolites as well as the available X-ray structures of MESH1 [13,14], to design MESH1 inhibitors and repress tumor proliferation as novel therapeutics. Fig. 4 The contribution of TAZ repression to the transcriptional responses of MESH1 knockdown in H1975 cells. a Heatmap of the selected genes in response to the MESH1 knockdown with or without TAZS89A restoration. Filtering resulted in 1007 probesets: at least four observations with absolute values ≥0.7 were included. ct represents control. b Quantification of the MESH1-silenced gene signatures that were restored by TAZS89A restoration. 33% (258 out of 786 genes) were rescued. Expression changes ≥2 folds were included as the down or upregulated genes, among which the ones that were rescued by TAZS89A restoration by at least 1.5 folds were considered as "rescued". c qRT-PCR validation of the selected genes in a. TAZS89A restoration significantly rescued the inhibition of cell cycle-related gene expression by MESH1 knockdown. (mean + s.d.). d ChIP-qPCR analysis of the abundance of TAZ protein at the enhancer region of CDC6 and RRM2 in cells transfected with the control or MESH1 siRNA. (mean + s.d.). e ChIP-qPCR analysis of the abundance of H3K27Ac mark at the promoter region of TAZ, YAP, and a heterochromatin region in cells transfected with siNT or two siMESH1s. (mean + s.d.). f qRT-PCR validation revealed that TSA treatment for 16 h rescued TAZ repression by MESH1 knockdown. (mean + s.d.). g qRT-PCR revealed that only siHDAC5 significantly rescued TAZ repression by siMESH1. (mean + s.d.). h qRT-PCR validation revealed that LMK235 treatment for 24 h rescued TAZ repression by MESH1 knockdown. (mean + s.d.). i Representative image of the western blots showed an enhanced nucleus translocation of HDAC5 by MESH1 knockdown, implying activation of HDAC5. For c, p values were calculated by the two-tail student's t-test. For d; e; g, p values were calculated by the two-way ANOVA followed by Tukey's posttest. For f; h, p values were calculated by the one-way ANOVA followed by Tukey's posttest. **P~(0.001, 0.01); ***P~(0.0001, 0.001); ****P < 0.0001; NS no significance.

Cell culture
Quantitative real-time PCR RNA was extracted with the RNeasy mini kit (Qiagen, #74104), reverse transcribed by the SuperScript II (ThermoFisher Scientific, #18064014) to generate cDNAs for qRT-PCR using primers and the Power SYBRGreen Mix (ThermoFisher Scientific, #4367659). Primers were designed across exonexon junctions and the specificity of PCR products was checked by electrophoresis. n = 3 biologically independent replicates.

Microarray, analysis, and GSEA analysis
Total RNAs were collected with RNeasy Mini Kit (Qiagen, #74104) and assessed with the Agilent BioAnalyzer. cDNAs were generated from 200 ng RNA using the Ambion MessageAmp Premier RNA Amplification (Life Technologies, Grand Island NY, USA) and were interrogated with Affymetrix U133A GeneChip. Gene expression data were deposited into NCBI GEO (GSE135358, GSE135346, and GSE147062). The microarray data were normalized by the RMA (Robust Multi-Array) algorithm. and zero transformed to the negative control (siNT), where we compared transcript levels for each gene in siMESH1 groups to the siNT group (n = 3 biologically independent replicates in each siRNA group) as previously performed [7,9,48]. Data were then filtered with Cluster 3.0 until~1000 probesets were left and clustered by the gene. Heat maps were generated with TreeView with the indicated intensity. For GSEA analysis, siMESH1 vs. siNT microarray data were compared with the TAZ and YAP-induced genesets in the Supplementary table S1 in Zhang H, et al. [29]. (gene fold change ≤12-fold were cut off) using the Gene Set Enrichment Analysis

BrdU incorporation assay
H1975 cells transduced with indicated inducible shRNAs were seeded in six-well plates and treated with doxycycline for 48 h, and then were transferred to four-well chamber slides (Sigma, #C6932-1PAK) 1 day prior to labeling. For labeling, cells were treated with 10 μM BrdU (ThermoFisher Scientific, #B23151) for 2 h at 37°C, washed with warm PBS for 2 min three times, and fixed in 3.7% formaldehyde in PBS for 15 min at RT, followed by the incubation of permeabilization buffer (0.1% Triton X-100 in PBS) for 20 min. Permeabilized cells were buffer changed to 1 N HCl for 10 min on ice, 2 N HCl for 10 min at RT, phosphate/citric acid buffer (pH 7.4) for 10 min at RT, and washed with permeabilization buffer for 2 min three times. Processed cells were blocked with antibody staining solution (0.1% Triton X-100 + 5% normal goat serum in PBS) for 1 h at room temperature on a rocker and blotted with 1:100 anti-BrdU primary antibody (Abcam, #ab115874) overnight at 4°C, followed by 1:500 anti-mouse Alexa-Flour 594 antibody (Abcam, #ab150116) for 1 h at RT, washed three times before being mounted with SlowFade Gold Antifade Mountant with DAPI (Invitrogen, #S36938), covered with the coverslips and imaged under a fluorescent microscope. n = 2 biologically independent replicates.

PI (Propidium Iodide) stain and flow cytometry
For cell cycle analysis, 2 × 10 5 H1975 cells were reverse transfected with the indicated siRNAs for 72 h, harvested using 0.05% trypsin with the media, fixed in 3 mL of ice-cold 70% ethanol, and gently vortexed until suspended. Fixed cells were centrifuged for 5 min at 1000 rpm at room temperature and washed twice with 3 mL of PBS at room temperature. For PI stain, cells were resuspended in 0.5-1 mL of PBS with 25 μg/mL PI (Sigma, #P4864) and 10 ug/mL RNAse A for 30 min at room temperature, light protected. At least 10 4 processed cells were measured on a Canto II Flow cytometer at last. Data were analyzed using the FlowJo V10 software and the gating strategy was shown in Supplementary Fig. 4c. n = 3 biologically independent replicates. dNTP measurement H1975 and RCC4 cells were transfected with the indicated siRNAs for 3 days, washed with ice-cold PBS twice, lysed by ice-cold 65% methanol, scrapped off, and extra 100 μL of 65% methanol were added to the plate to recover all the material. Lysates were vigorously vortexed for 2 min in 4°C, incubated at 95°C for 3 min, chilled on ice for 1 min, and centrifuge in 4°C, 14,000 RPM for 3 min. The supernatant was dried by speed vacuum and analyzed in Kim's lab at Emory University for dNTP quantification using HIV reverse transcriptase-based assay [49]. n = 3 biologically independent replicates.
Tumor sphere formation assay About 0.7 × 10 5 indicated H1975 cells were seeded in each well of the sixwell plate in normal media, and 24 h later were treated with or without 2 μg/mL Doxycyclin for 48 h. A total of 7000 cells were then transferred to the 24-well low-attachment plate (Corning, #CLS3473) in CCSC media in replicates (n = 3) and were cultured for 10 days before counting. Control or doxycycline-containing CCSC media were added every 4 days to maintain knockdown. Tumor spheres were counted under the microscope by diameter size: II: 100-200 μm; III: 200-300 μm; IV: 300-400 μm; V: >400 μm. Three technical replicates were averaged to represent each biological replicate. n = 3 biologically independent replicates.

Limited dilution assay
About 0.7 × 10 5 indicated H1975 cells were seeded in a six-well plate in normal media and treated with or without 2 μg/mL Doxycycline for 48 h. Cells were then serial diluted to the 96-well round-bottom low-attachment plate (Corning, #7007) with 1000, 500, 100, 50 cells/well in CCSC media with technical replicates of 48 for each cell density. Spheres were cultured in CCSC media for 8 days before counting. Doxycycline-containing or deficient (no doxy control) CCSC media were added to each well every 4 days to maintain the knockdown level. After 8 days, the number of wells that successfully formed tumor sphere out of the 48 replicates for each cell density in each group was counted and data were plugged into the Extreme Limiting Dilution Analysis (ELDA) algorithm [50] for stem cell frequency, p value, and confidence interval calculations.

ChIP-qPCR
Five million H1975 cells were seeded in 15 cm dishes and transfected with the siRNAs (siNT and siMESH1s) for 72 h before cross-link and collection of the cell lysates. Myers Lab ChIP-seq Protocol v011014-Adherent cells was then followed (https://www.encodeproject.org/documents/6ecd8240-a351-479b-9de6-f09ca3702ac3/@@download/attachment/ChIP-seq_Protocol_v011014. Fig. 5 The effect of HDAC5 and AHRR as epigenetic co-repressors on TAZ downregulation and cell proliferation arrest in H1975. a Representative image of the co-immunoprecipitation and western blots suggested that HDAC5 interacted with AHRR in H1975 cells. The flag-HDAC5 and HA-AHRR were co-transfected into cells and flag-HDAC5 was immunoprecipitated by the flag antibody and probed with antibodies recognizing flag or HA tags. b qRT-PCR revealed that AHRR knockdown significantly rescued TAZ repression. (mean + s.d.). c qRT-PCR validation of the selected cell proliferation gene (CDK1, CDC6, RRM1, and RRM2) was rescued either by AHRR or HDAC5 knockdown. (mean + s.d.). d Representative image of the crystal violet staining (top) and quantification for replicates (bottom) of H1975 cells showed a consistent resume of cell growth by AHRR and HDAC5 knockdown. (mean + s.d.). e Analysis of binding motifs of AHR on the five TAZ promoter/enhancer regions identified by GeneHencer. AHR (TF) has several binding sites on multiple TAZ promoter/enhancer regions. Each dashed line represents a binding site and the transcription start site (TSS) of TAZ is marked with a red arrow. f AHR ChIP-qPCR data in H1975 showed enrichment at the TAZ promoter/enhancer region 2 and 5, which was repressed by MESH1 knockdown. Cells were treated with ITE for 18 h to activate the AHR before ChIP assays. (mean + s.d.). g qRT-PCR validation of TAZ repression by AHR knockdown, consistent with the effect of MESH1 knockdown. (mean + s.d.). h qRT-PCR validation of downregulation of the selected cell proliferation genes by AHR knockdown. (mean + s.d.).
i Representative image of the crystal violet staining (top) and quantification for replicates (bottom) of H1975 cells showed inhibition of cell growth by AHR silencing. (mean + s.d.). j Tumor size and weight measurement showed xenograft growth inhibition upon doxycycline-induced AHR knockdown in the xenografted tumor model. p values were calculated by the one-way ANOVA followed by Tukey's posttest. For b; d; f, p values were calculated by the two-way ANOVA followed by Tukey's posttest. For c; h; i, p values were calculated by the two-tail student's ttest. *P~(0.01, 0.05); **P~(0.001, 0.01); ***P~(0.0001, 0.001); ****P < 0.0001. pdf). Sonication condition was optimized to the High mode, 30 s on/30 s off at 4°C for 45 min using the Bioruptor Twin (Diagenode) sonicator. Dynabeads Protein G (Invitrogen, #10003D) and anti-H3K27Ac (Cell Signaling Technology, #8173), anti-TAZ (Cell Signaling Technology, #70148), anti-AHR (Santa Cruz Biotechnology, #sc-133088 X) or rabbit IgG (negative control: Santa Cruz Technology, sc-66931), mouse IgG (Santa Cruz Technology, sc-2025) were used for pull-down in this study as instructed. Pulled down DNA was then mixed with primers targeting the TAZ promoter, or YAP promoter, or TAZproximal heterochromatin region together with the Power SYBRGreen Mix (ThermoFisher Scientific, #4367659) to undergo the qPCR reaction as described above. Primers were designed following the guidelines on the Michigan University Nutritional Sciences website (http://bridgeslab.sph.umich. edu/protocols/index.php/RT-PCR_primer_design_for_ChIP). n = 3 biologically independent replicates.

Co-immunoprecipitation
H1975 cells were transfected with the indicated vectors for 48 h. Cells were washed with PBS and resuspended in lysis buffer (50 mM Tris-HCl, 150 mM NaCl, 1% NP-40 and 5 mM EDTA, protease inhibitor cocktail, phosphatase inhibitor cocktail) at 4°C with shaking for 30 min. Cell debris was pelleted by centrifugation at 16,000 × g for 15 min at 4°C. Immunoprecipitation was performed using anti-FLAG (Anti-Flag M2 Affinity Gel, Sigma-Aldrich, #A2220) agarose beads at 4°C overnight, washed three times and eluted using SDS loading buffer at 95°C for 15 min, followed by the western blot using anti-HA (CST, #3724) and anti-FLAG (CST, #14793). n = 2 biologically independent replicates.

Statistical analysis and data collection
n numbers of biologically independent replicates were included in each figure legend and illustrated by the individual data points in each bar graph (mean + s.d.). The sample size was chosen based on our previous experience and literature reports for each experiment. P values were calculated by the indicated statistical analysis methods described in each figure legend and were justified as appropriate. Data collection was assigned blindly to different researchers.