The testis protein ZNF165 is a SMAD3 cofactor that coordinates oncogenic TGFβ signaling in triple-negative breast cancer

Cancer/testis (CT) antigens are proteins whose expression is normally restricted to germ cells yet aberrantly activated in tumors, where their functions remain relatively cryptic. Here we report that ZNF165, a CT antigen frequently expressed in triple-negative breast cancer (TNBC), associates with SMAD3 to modulate transcription of transforming growth factor β (TGFβ)-dependent genes and thereby promote growth and survival of human TNBC cells. In addition, we identify the KRAB zinc finger protein, ZNF446, and its associated tripartite motif protein, TRIM27, as obligate components of the ZNF165-SMAD3 complex that also support tumor cell viability. Importantly, we find that TRIM27 alone is necessary for ZNF165 transcriptional activity and is required for TNBC tumor growth in vivo using an orthotopic xenograft model in immunocompromised mice. Our findings indicate that aberrant expression of a testis-specific transcription factor is sufficient to co-opt somatic transcriptional machinery to drive a pro-tumorigenic gene expression program in TNBC.


28
Tumors frequently activate genes whose expression is normally restricted to the testes 29 (Sahin et

54
Unfortunately, ZNF165 lacks conservation in mice, thereby precluding genetic studies that 55 could illuminate its normal function within the testes (Tirosvoutis et al. 1995). However, 56 investigation of its function in TNBC revealed that ZNF165 associates with chromatin to 57 regulate expression of its target genes, which includes approximately 25% of all TGFβ-58 responsive genes in TNBC cells (Maxfield et al. 2015).

59
The cellular response to TGFβ is highly context selective, often resulting in dramatically 60 different phenotypes. For example, TGFβ is well known to facilitate growth inhibition in 61 breast epithelia and most other somatic tissues, primarily through transcriptional activation of 62 cyclin-dependent kinase inhibitors that result in cell cycle arrest (Hannon and Beach 1994; 63 Polyak et al. 1994; Morikawa et al. 2016). Conversely, TGFβ has also been shown to

76
Here, we investigate the mechanisms by which ZNF165 modulates TGFβ-dependent 77 transcription in TNBC. We find that ZNF165 co-occupies a subset of SMAD3 genomic 78 binding sites and physically interacts with SMAD3 on chromatin, enhancing SMAD3 79 recruitment to specific loci. We also identify ZNF446 and TRIM27, two proteins never . Intersecting SMAD3 and ZNF165-98 associated genes revealed that >90% of ZNF165 target genes are also bound by SMAD3, 99 an overlap representing a statistically significant enrichment (p = 5.50e-120, hypergeometric 100 distribution) ( Figure 1A). This is in agreement with our previous findings that ZNF165 targets 101 a significant fraction of the TGFβ-responsive transcriptome (Maxfield et al. 2015). Mapping 102 the distance between binding sites further revealed that 90% of shared targets were bound 103 by both transcription factors at a distance of less than 100 kb from one another ( Figure 1B).

104
Furthermore, 36% of shared target genes had ZNF165 and SMAD3 binding sites within 1 kb, 105 a proximity at which both factors exhibited overlap on chromatin due to the broad enrichment 106 profile of SMAD3 as observed within our dataset ( Figure 1B-E). We next established 107 ZNF165 and SMAD3 binding profiles in SUM159 cells, which are also classified within the 108 claudin-low molecular subtype and require ZNF165 for viability (Prat et

114
We next applied a motif discovery strategy to the 118 regions where both transcription 115 factors were bound within 1 kb of each other. More than 50% of these co-bound sites were 116 present in gene promoters (-1 kb or +100 base pairs from TSS) while SMAD3 binding sites 117 were relatively evenly distributed across genetic regions ( Figure 1F). This analysis revealed 118 that GC-rich elements are significantly enriched at co-bound regions (n=118) compared to 119 those bound by ZNF165 alone (n=204) ( Figure 1G and 1H  Given the enrichment of SMAD3 and ZNF165 at shared sites and their general proximity to 136 one another on chromatin, we next asked whether these transcription factors cooperate to 137 regulate expression of co-bound genes. We leveraged a previously generated gene 138 expression data set of TGFβ responsiveness in WHIM12 cells (Maxfield et al. 2015). Here, 139 we assembled a set of 65 TGFβ-responsive genes that are co-bound by SMAD3 and 140 ZNF165 (Figure 2-figure supplement 1A). We then compared the impacts of ZNF165 and 141 SMAD3 depletion on a subset of these genes. This analysis revealed a significant correlation 142 (r = 0.52) between expression fold-changes in response to depletion of either transcription 143 factor ( Figure 2A). In particular, ZNF165 and SMAD3 were required to either activate or 144 repress expression of the majority of shared target genes (quadrant I and III, Figure 2A). We 145 also observed opposing effects of SMAD3 and ZNF165 on a smaller subset of genes, 146 suggesting that in some cases, the activities of these two proteins may oppose one another 147 (quadrant II and quadrant IV, Figure 2A).

148
The observed expression changes in Figure 2A suggest two modes of regulation for 149 ZNF165-SMAD3 targets. First, ZNF165 collaborates with SMAD3 to activate or oppose 150 TGFβ-mediated gene expression changes (quadrant I and III, Figure 2A). Alternatively, 151 ZNF165 can antagonize SMAD3-mediated activation or repression of shared target genes in 152 response to TGFβ (quadrant II and IV, Figure 2A). The majority of genes fell into the former 153 category, with SMURF2 and RRAD being the two most responsive and were thus selected 154 as archetypes for follow-up analysis. Importantly, we found that RRAD and SMURF2 protein 155 accumulation corresponded to mRNA alterations upon ZNF165 or SMAD3 depletion ( Figure   156 2B and 2C).

157
Given the observed collaborative function of ZNF165 and SMAD3 on target gene 158 expression, we next asked whether their association with chromatin was cooperative.

159
ZNF165 knockdown resulted in a ~50% reduction of SMAD3 enrichment at the ZNF165 160 binding sites proximal to both SMURF2 and RRAD ( Figure 2D). Importantly, SMAD3 binding 161 at the promoter of LBH, a site not co-bound by ZNF165, was unaffected by ZNF165 162 depletion. SMAD3 commonly recruits the histone acetyltransferase, p300, to activate gene 163 expression of its targets (Feng et al. 1998;Janknecht et al. 1998). Consistent with this, we 164 observed a reduction of H3K27ac at the ZNF165-SMAD3 binding site proximal to RRAD in 165 response to ZNF165 knockdown ( Figure 2E). Conversely, an increase in H3K27ac was 166 detected upon ZNF165 overexpression ( Figure 2F). These findings suggest that ZNF165 is 167 essential for recruitment of SMAD3 to shared target gene promoters and transcriptional 168 activation. We reasoned that this activity could promote expression of genes required for 169 tumorigenic phenotypes. Indeed, a report from 2001 indicated that RRAD is sufficient to

191
As ZNF165 appears to facilitate SMAD3 recruitment to chromatin, we asked whether these 192 proteins physically associate. Interaction studies using co-expression/co-193 immunoprecipitation indicated an association between ZNF165 and SMAD3 ( Figure 3A). In 194 addition, we found that ZNF165 interacts with the active, phosphorylated form of SMAD3 on 195 chromatin in SUM159 cells stably expressing ZNF165-V5 ( Figure 3B). Activation of SMAD3 196 and its resulting nuclear translocation is dependent on the phosphorylation of its C-terminus 197 by TGFβRI in response to TGFβ (Massague 2012 Figure 3D). Furthermore, we found that endogenous ZNF165 and SMAD4 also 204 interact in a TGFβ-dependent manner ( Figure 3E). Taken together, these results indicate 205 that in response to TGFβ, ZNF165 physically associates with phosphorylated SMAD3.  Figure 4D). We then employed the SMAD3/4 PLA assay described above and 231 found that ZNF446 interacts with both SMAD3 and SMAD4 in a TGFβ-dependent fashion, 232 similar to ZNF165 ( Figure 4E and 4F).

233
To determine if ZNF446 also associates with ZNF165-SMAD3 target genes, we

247
With respect to target genes, we found that over half of those shared by ZNF165 and 248 SMAD3 (251/446) are also targeted by ZNF446 ( Figure 5I). The impact of ZNF446 depletion 249 on expression of these genes significantly correlated with both ZNF165 (r = 0.49) and 250 SMAD3 (r = 0.86) ( Figure 5J). Collectively, these findings implicate ZNF446 as a member of 251 the ZNF165-SMAD3 complex that can influence TGFβ-dependent transcription in TNBC.  264 Ma et al. 2016). However, its function as a transcriptional regulator in TNBC has not been 265 characterized. Co-expression/co-immunoprecipitation experiments revealed that both 266 ZNF446 and ZNF165 interact with TRIM27 and that all three proteins immunoprecipitate as 267 a complex ( Figure 6B). We subsequently confirmed a nuclear interaction between ZNF165-268 V5 and endogenous TRIM27 by proximity ligation in SUM159 cells and found that TRIM27 269 depletion also phenocopied changes in ZNF165-SMAD3 target gene expression (r = 0.81) 270 ( Figure 6C and 6D). Notably, TRIM27 appears to be bi-functional, either positively or 271 negatively regulating expression of the ZNF165-SMAD3 targets ( Figure 6D). Indicative of a 272 co-activator function, we found that TRIM27 is required for the transcriptional activation of 273 RRAD following ZNF165 or ZNF446 overexpression ( Figure 6E; Figure

289
Given that TRIM27 was critical for ZNF165-dependent gene regulation, we next asked 290 whether TRIM27 was essential for malignant behaviors. While depletion of TRIM27 does not 291 seem to strongly affect TNBC tumor cell growth in 2D, we observed a significant defect in 292 anchorage-independent growth using soft agar assays (    were transfected with siRNA for 32h and subjected to SMAD3 ChIP followed by qPCR using primers targeting the indicated binding sites. Fold enrichment was determined by dividing the percent input values for SMAD3 by those for IgG. Error bars represent mean ± SEM. P-values were calculated using an unpaired, two-tailed Student's t-test. Data are representative of four independent experiments. (E) SUM159 cells were transfected with siRNA for 32h and subjected to H3K27ac ChIP followed by qPCR using primers targeted to the RRAD promoter or a negative control region. Error bars represent mean + SEM. P-value was calculated using an unpaired, two-tailed Mann-Whitney test. Data are representative of four independent experiments. (F) SUM159 cells stably expressing ZNF165-V5 or control cDNA were subjected to H3K27ac ChIP followed by qPCR using primers targeted to the RRAD promoter or a negative control region. Error bars represent mean + SEM. P-value was calculated using an unpaired, two-tailed Mann-Whitney test. Data are representative of four independent experiments.  TFAP2C  DUSP4  FOXQ1  CDKN1B  PER1  TOB1  MIDN  BDKRB2  FAM84B  IRF1  KCNE4  RASSF3  RASGRF1  EML6  MAP3K1  PPP4R1L  CYP1B1  PHF15  CIRBP  IL18  TSPAN8  BCL2L1  PEX2  ZNF32  TSHZ2  PCID2  TBC1D2  EXOC6  GLUD2  MYC  PICK1  SLC30A1  SUV39H2  PKIA  POU2F3  SPSB2  IGLL1  ZNF827  MEX3B  PFKFB3  MAFK  CXCL12  POU5F1B  SYNE1  HIC1  TCF7L2  RDH10  SMOX  SMURF2  ST3GAL1  ZMIZ1  MDGA1  PTPRE  CXXC5  EPHB2  RRAD  GEM  DLX1  HAS2  APCDD1L  TRIB1  SMAD7  DLX2  Proximity ligation assays (PLAs) performed using antibodies against endogenous ZNF165 and SMAD3 in SUM159 cells, where either antibody alone was used as a negative control. Cells were pre-treated with 20 μM SB-431542 or DMSO for 15 minutes, followed by stimulation with 5 ng mL -1 TGFβ for 30 minutes. Scale bars, 10 μm. The mean PLA signal (number of foci per nucleus) is quantified (right), where each data point represents the mean signal calculated within one image. P-value was calculated using an unpaired, two-tailed Student's t-test. Ten images were used per condition and data are representative of three independent assays. (E) As in (D) except antibodies against endogenous ZNF165 and SMAD4 were used.  interactors that contain a SCAN domain as identified from a systematic yeast two-hybrid screening approach (interactome.dfci.harvard.edu). A domain map of ZNF165 is shown above for reference. (B) Forty-eight hours after transfection with indicated siRNAs, RNA was extracted from WHIM12 or SUM159 cells and qPCR was used to determine the relative expression of SMURF2. Log 2 fold-change values normalized to control knockdown samples are displayed. Error bars represent mean + SEM. P-values were calculated for each sample using a one-tailed Mann-Whitney test. Data are representative of three independent experiments. (C) Forty-eight hours after transfection with indicated cDNA, HEK293T lysates were subjected to immunoprecipitation with V5 antibody. Immunoblotting was performed with indicated antibodies. Data are representative of three independent experiments. (D) SUM159 nuclear lysates were immunoprecipated with antibodies against endogenous ZNF165 or IgG. Immunoblotting was performed with indicated antibodies. (E) Proximity ligation assays (PLAs) performed using antibodies against endogenous ZNF446 and SMAD3 in SUM159 cells, where either antibody alone was used as a negative control. Cells were pre-treated with 20 μM SB-431542 or DMSO for 15 minutes, followed by stimulation with 5 ng mL -1 TGFβ for 30 minutes. Scale bars, 10 μm. The mean PLA signal (number of foci per nucleus) is quantified (right), where each data point represents the mean signal calculated within one image. P-value was calculated using an unpaired, two-tailed Student's t-test. Ten images were used per condition and data are representative of three independent assays. (F) As in (E) except antibodies against endogenous ZNF446 and SMAD4 were used.      Figure 6. TRIM27 is essential for ZNF165 transcriptional activity and tumor growth in vivo. (A) Cartoon maps showing the domains of ZNF446 and TRIM27. The KRAB-binding domain of TRIM27 is highlighted for reference. (B) Forty-eight hours after transfection with indicated cDNA, HEK293T lysates were subjected to immunoprecipitation with HA antibody. Immunoblotting was performed using indicated antibodies. Data are representative of three independent experiments. (C) Proximity ligation assays (PLAs) performed using V5 and TRIM27 antibodies in SUM159 cells stably expressing ZNF165-V5. Parental SUM159 cells were used as a negative control with the V5/TRIM27 antibody combination. Scale bar, 10 μm. The mean PLA signal (number of foci per nucleus) is quantified (right), where each data point represents the mean signal calculated within one image. P-values were calculated using an unpaired, two-tailed Student's t-test. Five images were used per condition and data are representative of four independent assays. (D) WHIM12 cells were transfected with siRNA for 48h and qPCR was used to quantify relative expression (log 2 fold change) of shared ZNF165-SMAD3 target genes upon depletion of ZNF165 (x-axis) or TRIM27 (y-axis). The Pearson correlation coefficient is indicated by r. Data are representative of three independent experiments. (E) SUM159 cells stably expressing indicated cDNA were transfected with siRNA targeting TRIM27 for 48h. Relative RRAD expression was measured using qPCR and the data were normalized to the CTRL sample (grey). Error bars represent mean + SEM. P-values were calculated using an unpaired, two-tailed Mann-Whitney test. Significance is indicated by asterisks, where * = p < 0.05. Data are representative of four independent experiments. (F) PLAs performed using antibodies against endogenous TRIM27 and SMAD3 in SUM159 cells, where either antibody alone was used as a negative control. Cells were pre-treated with 20 μM SB-431542 or DMSO for 15 minutes, followed by stimulation with 5 ng mL -1 TGFβ for 30 minutes. Scale bar, 10 μm. The mean PLA signal (number of foci per nucleus) is quantified (right), where each data point represents the mean signal calculated within one image. P-value was calculated using an unpaired, two-tailed Student's t-test. Ten images were used per condition and data are representative of three independent assays. (G) As in (F) except antibodies against endogenous TRIM27 and SMAD4 were used. (H) Tumor volumes from mice orthotopically injected with SUM159T-Luciferase cells stably expressing shRNAs against TRIM27 (n=10) or a non-targeting control (n=9). P-value was calculated using an unpaired, two-tailed Student's t-test. (I) Representative images of bioluminescence (BLI) measurements taken for mice from each group. BLI data was quantified (right) and the p-value was calculated using an unpaired, two-tailed Student's t-test. Model for assembly of a neomorphic ZNF165 transcriptional complex in TNBC. Aberrant expression of testis-specific ZNF165 in TNBC co-opts somatic transcriptional machinery to modulate TGFβ-dependent transcription. The ZNF165 SCAN domain recruits ZNF446, and together these factors associate with SMAD3 at shared binding sites throughout the genome and alter gene expression to promote neoplastic behaviors. Via its KRAB domain, ZNF446 recruits the co-activator/co-repressor TRIM27, which is essential for ZNF165 transcriptional activity and tumor growth in vivo. Given the testis/tumor-restricted expression of ZNF165, disruption of this complex holds potential as a therapeutic strategy to specifically inhibit pro-tumorigenic TGFβ signaling in TNBC.
Tumor-specific transcriptional complex