PRDM4 mediates YAP‐induced cell invasion by activating leukocyte‐specific integrin β2 expression

Abstract

Yes‐associated protein (YAP) is a transcriptional co‐activator and a major effector of the Hippo pathway that promotes cell proliferation and stemness, while inhibiting apoptosis. YAP plays a central role in organ size control, and its deregulation strongly promotes cancer initiation and progression. However, the mechanisms by which YAP promotes cell invasion and metastasis are not fully understood. Here, we report that YAP induces leukocyte‐specific integrin β2 (ITGB2) expression in cancer cells, thereby promoting cell invasion through the endothelium in a manner mimicking leukocytes. Through independent biochemical purification and a functional screen, we further identified PR/SET domain 4 (PRDM4) as a transcription factor interacting with the WW domains of YAP to mediate ITGB2 expression and cell invasion. Consistently, ITGB2 and PRDM4 mRNA levels are significantly increased in metastatic prostate cancer. In addition, PRDM4 contributes to YAP‐induced tumorigenesis possibly via mediating the expression of other YAP target genes. Our results demonstrate that YAP promotes cell invasion by inducing leukocyte‐specific integrin expression, and identify PRDM4 as a novel transcription factor for YAP targets.

Introduction

The Hippo pathway plays an evolutionarily conserved role in organ size control 1, 2. Mutations in this pathway lead to characteristic enlargement of organ size due to enhanced cell proliferation, reduced apoptosis, and expansion of tissue‐specific progenitor cells. In this pathway, the MST1/2 kinases (Hippo in Drosophila) in association with an adaptor protein SAV1 (Sav in Drosophila) phosphorylate LATS1/2 kinases (Wts in Drosophila). LATS1/2 associate with another adaptor protein MOB1A or MOB1B (collectively referred to as MOB below, Drosophila Mats homolog) to phosphorylate and inactivate a transcription co‐activator YAP and its paralog transcriptional co‐activator with PDZ‐binding motif (TAZ, YAP, and TAZ are Drosophila Yki homologs), thus regulate gene expression 3, 4, 5, 6, 7, 8.

Dysfunction of the Hippo pathway plays a key role in tumorigenesis. First, Hippo pathway genes such as YAP, NF2, and GNAQ/GNA11 are amplified or mutated in cancers 9, 10, 11, 12. Second, other Hippo pathway genes, such as SAV1, RASSF1A, MST1/2, and LATS1/2, are deregulated on epigenetic and gene expression levels 13. Third, genetic engineered mice unequivocally demonstrated a tumor suppressive role of the Hippo pathway and an oncogenic function of YAP 14, 15, 16, 17, 18, 19, 20. Metastasis is the major reason for cancer‐related mortality. The Drosophila Yki promotes malignant growth and invasion in vivo via activating a transcription factor network 21. YAP has also been shown to promote metastasis through several mechanisms, such as triggering epithelial mesenchymal transition (EMT) 22, 23, altering F‐actin/G‐actin turnover 24, and inhibiting anoikis 25. However, it is unclear whether and how YAP could promote invasion of the endothelium, a key step during metastasis.

ITGB2 is an integrin subunit normally expressed only in leukocytes, in which it heterodimerizes with integrin alpha subunits to promote leukocyte adhesion to endothelium and the ensuing extravasation 26. Here, we report that robust induction of ITGB2 expression by YAP promotes cancer cell invasion through the endothelium, thus may help breach a key barrier in cancer metastasis. Furthermore, through two independent approaches, we identified PRDM4 as a new target transcription factor of YAP mediating ITGB2 induction. These findings reveal a new mode of transcriptional regulation by the Hippo pathway in the context of cancer metastasis.

Results and Discussion

YAP induces ITGB2 expression in cancer cells

ITGB2 is one of the most strongly induced YAP target genes in MCF10A human mammary epithelial cells 27, 28. This is surprising because the expression of ITGB2 is restricted to leukocytes in physiological conditions. We confirmed that in several epithelial or cancer cell lines, ITGB2 mRNA and protein levels could be induced by the expression of YAP, or its constitutively active 5SA mutant (serine to alanine mutation of all five Hippo pathway target sites; Figs 1A and B, and EV1A). Noteworthy, the fold induction of ITGB2 is even stronger than that of CTGF, a well‐established YAP target gene 22, in all three cell lines examined (Fig EV1B). We further examined ITGB2 expression in cancer cells. We found that both ITGB2 and CTGF expressions are relatively low in MCF10A, HCT116, A549, and HTB182 cells (Fig 1C). However, in ACHN and MDA‐MB‐231 cells, in which YAP is activated due to mutation of Sav and NF2 29, respectively, both ITGB2 and CTGF expression levels are elevated. Glioblastoma cell line SF268 with YAP gene locus amplification also exhibits higher expression of YAP and ITGB2. ITGB2 and CTGF levels are highest in an ovarian cancer cell line ES‐2, in which YAP mRNA level is also relatively high (Fig 1C). Therefore, ITGB2 is expressed in many cancer cell lines, and its expression correlates with CTGF expression and YAP activity. Furthermore, expression of small interfering RNAs (siRNAs) against YAP and TAZ efficiently decreased the mRNA and protein levels of both CTGF and ITGB2 in ACHN cells (Fig 1D and E). Similar effect has been confirmed in ES‐2 cells (Fig EV1C and D). However, knockdown of Yap in primary mouse bone marrow stromal stem cells (BMSSC) repressed Ctgf but not Itgb2 expression (Fig EV1E), suggesting a species difference in ITGB2 induction by YAP. Indeed, YAP overexpression did not induce Itgb2 in two different mouse cell lines B16 and 3T3L1 (Fig EV1F and G). Taking together, ITGB2 is a YAP target gene that could be induced by YAP activation in human cancer cells.

Figure 1. YAP induces ITGB2 expression in cancer cells.

• A, B
  YAP induces ITGB2 expression. mRNA levels were determined by quantitative RT–PCR (qPCR) (A), and protein levels were examined by Western blotting (B). Molecular weight markers in kDa. Empty vector (Vec) was used as control.
• C
  ITGB2 expression correlates with YAP activity in different cell lines. The mRNA levels were determined by qPCR (A549 as unit one).
• D, E
  Endogenous YAP and TAZ regulate ITGB2 expression. ACHN cells were transfected with siRNAs. mRNA levels (D) and protein levels (E) of ITGB2 were determined.

Data information: Results are representatives of three independent experiments. Values represent mean ± s.d. from three technical repeats. *P < 0.05; **P < 0.01; ***P < 0.001.Source data are available online for this figure.

Figure EV1. Data related to Fig 1 .

1. YAP induces ITGB2 protein expression in PC3 cells. Cell lysates were examined by Western blotting.
2. The expression level of YAP and CTGF in cells the same as these in Fig 1A. mRNA levels of YAP and CTGF were determined by qPCR.
3. Endogenous YAP regulates ITGB2 mRNA level. ES‐2 cells were transfected with siRNAs. ITGB2 and YAP mRNA levels were determined by qPCR.
4. Endogenous YAP/TAZ regulates ITGB2 protein level. ES‐2 cells were transfected with siRNAs. ITGB2 protein levels were determined by Western blotting.
5. Knockdown of Yap did not affect Itgb2 mRNA level in BMSSC. BMSSC were transfected with siRNAs targeting Yap or Taz, and mRNA levels were determined by qPCR.
6. Overexpression of YAP in B16 mouse melanoma cell line decreased Itgb2 mRNA level. The mRNA levels of Itgb2 in B16 stable cells were determined by qPCR.
7. Overexpression of YAP in 3T3L1 cells did not significantly induce Itgb2 mRNA level. The mRNA levels of Itgb2 in 3T3L1 stable cells were determined by qPCR.

Data information: Results are representatives of three independent experiments. Values represent mean ± s.d. from three technical repeats. *P < 0.05; **P < 0.01; ***P < 0.001.Source data are available online for this figure.

ITGB2 mediates YAP‐induced cancer cell transendothelial invasion

Physiologically, ITGB2‐containing integrins mediate transendothelial migration of leukocyte 30, 31. An intriguing possibility is that through YAP activation, cancer cells hijack this mechanism for metastasis. Flow cytometry analysis of A375 melanoma cells confirmed the effect of YAP in inducing ITGB2 protein, which partially localizes to the plasma membrane (Fig 2A). We also examined the expression levels of four alpha subunits that could dimerize with ITGB2. Among ITGAL, ITGAM, and ITGAX, which were detectable in A375 cells, ITGAM and ITGAX mRNA levels were dramatically decreased by YAP activation through an unknown mechanism (Fig EV2A). Furthermore, the expression of active YAP mildly increased plasma membrane‐localized ITGAL, ITGAM, and ITGAX (Fig EV2B), possibly due to co‐translocation with ITGB2. Therefore, YAP‐induced ITGB2 should mainly exist as LFA‐1, a heterodimer with ITGAL. Consistently, expression of active YAP strongly promoted the adhesion of A375 cells to ICAM‐1, an LFA‐1 ligand, especially in the presence of Mn²⁺, which increases the affinity of integrins 32 (Fig 2B). In contrast, YAP did not promote the adhesion of A375 cells to VCAM‐1, an integrin α4β1 ligand (Fig 2B). By immunofluorescence, YAP‐induced ITGB2 partially localize to actin‐rich lamellipodia but not focal adhesion‐like structures when cells were plated on ICAM‐1‐coated surfaces (Fig EV2C). Thus, YAP promotes cancer cell adhesion to LFA‐1 ligand, possibly through the induction of ITGB2.

Figure 2. ITGB2 mediates YAP‐induced cancer cell transendothelial invasion.

• A
  YAP‐induced ITGB2 partially localizes to plasma membrane. Flow cytometry analysis of A375 stable cells for total and membrane ITGB2 expression. Empty vector (Vec) was used as control.
• B
  YAP promotes cell adhesion to ICAM1‐coated surface. A375 stable cells were subjected to flow chamber assay with or without 1 mM Mn²⁺.
• C
  Knockout of ITGB2 in A375 cells expressing active YAP‐127A mutant. Lysates were examined by Western blotting.
• D, E
  ITGB2 knockout suppresses YAP‐induced cell migration. Cells were subjected to transwell migration assay. A parallel set of cells was seeded into regular culture plates as a control for cell number. Relative migration was calculated by normalization of migrated cells with the number of total seeded cells (E).
• F, G
  ITGB2 knockout suppresses YAP‐induced transendothelial invasion. A375 stable cells were labeled with CFSE and then seeded onto a layer of TNF‐α‐pretreated HUVEC endothelium formed in transwells. A parallel set of cells was seeded into regular culture plates as a control for cell number. Invasion was quantified in (G).
• H
  ITGB2 facilitates YAP‐induced cancer cell metastasis in vivo. 1.5 × 10⁶ of each stable cell was injected into nude mice via the tail vein. Mice were sacrificed after 30 days. Two representative lungs out of five were presented. Arrowheads indicate tumors.

Data information: Results are representatives of three independent experiments. Values represent mean ± s.d. from three technical repeats. ***P < 0.001. Scale bars 100 μm.Source data are available online for this figure.

Figure EV2. Data related to Fig 2 .

1. The mRNA levels of ITGAL, ITGAM, ITGAD, and ITGAX in A375 cells stably expressing YAP‐5SA as determined by qPCR.
2. YAP increases plasma membrane localization of ITGAL. A375 stable cells were analyzed for membrane localization of ITGAL, ITGAM, and ITGAX by flow cytometry without cell permeabilization.
3. YAP‐induced ITGB2 partially localized to lamellipodia when cells were grown on ICAM‐1 coated surfaces. A375 stable cells were plated onto fibronectin or ICAM‐1 coated cover glasses. Cells were stained for ITGB2 and F‐actin. Scale bars 10 μm.
4. ICAM1 expression was induced by inflammatory conditions in HUVEC cells. HUVEC cells were treated with 10 ng/ml TNF‐α or IL‐1β, and the mRNA level of ICAM1 was determined by qPCR.

Data information: Results are representatives of three independent experiments. Values represent mean ± s.d. from three technical repeats. **P < 0.01.

We further generated ITGB2 knockout A375 cells by CRISPR/Cas9‐mediated genome editing using two independent single guide RNAs (sgRNAs) (Fig 2C). The ability of these cells to migrate through transwell inserts was examined. While YAP strongly promoted cell migration toward ICAM‐1‐coated surfaces, knockout of ITGB2 largely reduced the effect (Fig 2D and E). Invasion through the endothelium is a major barrier for cancer metastasis 33. We thus investigated whether YAP‐induced ITGB2 could further facilitate transendothelial invasion. Pretreatment of human umbilical vein endothelial cells (HUVEC) with IL‐1β or TNF‐α induced ICAM‐1 expression to a similar level (Fig EV2D). Remarkably, YAP strongly induced A375 cells to invade through TNF‐α pretreated HUVEC cell layer in a manner largely dependent on ITGB2 (Fig 2F and G). Furthermore, while YAP promoted inoculated cancer cells to form lung lesions in nude mice, knockout of ITGB2 blocked the effect (Fig 2H). These data indicate that YAP‐induced ITGB2 expression facilitates cancer cells to extravasate through the endothelium.

PRDM4 is a transcription factor interacting with YAP WW domains

We next investigated how YAP induces ITGB2 expression. TEAD family transcription factors interact with YAP in a serine 94 (S94)‐dependent manner 22. Mutation of S94 dramatically repressed ITGB2 induction by YAP (Fig 3A and B). Interestingly, mutation of YAP WW domains also largely abolished ITGB2 induction (Fig 3A and B). In contrast, the induction of CTGF was inhibited by mutation of S94 but not the WW domains (Fig EV3A). However, ITGB2 expression was not reduced by knockdown of YAP WW domain‐interacting transcription factors, such as RUNX (data not shown). Thus, other transcription factors mediating YAP‐induced gene expression may exist.

Figure 3. Identification of PRDM4 as a transcription factor interacting with YAP WW domains.

• A, B
  YAP induces ITGB2 expression in a TEAD‐ and WW domain‐dependent manner. ITGB2 mRNA (A) and protein (B) levels in MCF10A stable cells were determined. Empty vector (Vec) was used as control.
• C
  TAP‐mass spectrometry identified PRDM4 as a YAP‐interacting protein. The number of peptide hits for PRDM4 and several known YAP‐interacting proteins is shown.
• D
  Illustration of PRDM4 domain organization. ZNF, zinc finger motif.
• E–G
  YAP co‐immunoprecipitates with PRDM4. Nuclear extracts (NE) were made from transfected HEK293 cells and then immunoprecipitated with indicated antibodies. Samples were examined by Western blotting.
• H
  YAP activated PRDM4 in a transcription factor library screen. The screen was done as described in Materials and Methods. Fold activation of TEAD3, TEAD4, and PRDM4 by YAP are shown.
• I–K
  The WW domains of YAP (I), and the PPXY motifs of PRDM4 (J) are required for the activation of PRDM4, which is inhibited by the Hippo pathway (K). The reporter assay is similar to that in (H).

Data information: Results are representatives of three independent experiments. Values represent mean ± s.d. from three technical repeats. *P < 0.05; **P < 0.01; ***P < 0.001.Source data are available online for this figure.

Figure EV3. Data related to Fig 3 .

1. The induction of CTGF expression by YAP depends only on TEAD binding but not the WW domains. Cells were the same as these in Fig 3A, and mRNA levels were determined by qPCR.
2. Endogenous YAP interacts with PRDM4 in A375 cells. A375 cells stably expressing PRDM4‐WT or mutant were established by lentiviral infection. PRDM4 was immunoprecipitated by anti‐Flag antibody, and samples were examined by Western blotting.
3. WW domains of YAP are required for interaction with PRDM4. Experiments are similar to these in Fig 3G.

Data information: Results are representatives of three independent experiments. Values represent mean ± s.d. from three technical repeats. **P < 0.01; ***P < 0.001.Source data are available online for this figure.

We next investigated YAP‐interacting proteins via tandem affinity purification (TAP). Mass spectrometry identified known YAP‐interacting PPXY‐containing proteins, including LATS1 and AMOTL2 (Fig 3C). Interestingly, we also recovered a potential transcription regulator, PRDM4, which has two PPXY motifs, six zinc finger motifs, and a PR domain (20–30% identical to the SET domain of histone lysine methyltransferase; Fig 3C and D) 34. By co‐immunoprecipitation, we confirmed that YAP specifically interacts with PRDM4 (Fig 3E). Furthermore, although we did not find an antibody suitable for detection of endogenous PRDM4, we confirmed that endogenous YAP could interact with ectopically expressed PRDM4 (Figs 3F and EV3B). In addition, mutation of each WW domain of YAP or PPXY motif of PRDM4 strongly inhibited their interaction (Figs 3F and G, and EV3B and C). Thus, PRDM4 is a transcription factor interacting with YAP WW domains.

Remarkably, PRDM4 has also been recovered through an independent approach. We have previously reported the identification of TEADs as YAP target transcription factors in a luciferase reporter‐based screen of a plasmid library encoding Gal4‐fusion human transcription factors 22. Interestingly, PRDM4 turned out to be activated by YAP to a similar level as TEADs in this screen (Fig 3H). Using the same reporter assay, we demonstrated that mutation of either YAP WW domains or PRDM4 PPXY motifs abolished reporter activation by YAP and PRDM4 (Fig 3I and J). In addition, the transcriptional activity of the YAP‐PRDM4 complex is inhibited by the Hippo pathway kinases (Fig 3K). Taken together, via two independent unbiased approaches, we identified PRDM4 as a YAP WW domain‐interacting transcription factor.

PRDM4 cooperates with TEAD to mediate YAP‐induced ITGB2 expression

We next determined whether PRDM4 mediates ITGB2 induction by YAP. Expression of PRDM4 to five times the endogenous level leads to a ninefold induction of ITGB2, which was hampered by mutation of PPXY motifs (Fig EV4A). More importantly, knockdown of PRDM4 markedly reduced ITGB2 expression in ACHN and PC3 cells (Figs 4A and B, and EV4B). Furthermore, knockdown of TEAD1/3/4 also strongly repressed ITGB2 expression (Fig 4C and D). Therefore, both PRDM4 and TEAD play important roles in regulation of ITGB2 expression.

Figure EV4. Data related to Fig 4 .

1. Expression of PRDM4 induces ITGB2 mRNA. mRNA levels in ACHN stable cells were determined by qPCR.
2. Knockdown of PRDM4 inhibits ITGB2 protein expression in PC3 cells. Experiments were similar to these in Fig 4B.
3. Activation of ITGB2 reporter by PRDM4 depends on the PPXY and the zinc finger motifs. HEK293T cells were transfected, and luciferase activity was normalized to co‐transfected beta‐galactosidase activity.
4. The PR domain of PRDM4 is not required for the activation of ITGB2 promoter. The reporter assay was similar to these in Fig 4F.
5. Two potential PRDM4 binding sites are not required for transactivation of ITGB2 reporter by YAP‐PRDM4. Proximal ITGB2 reporter or mutants were co‐transfected with indicated plasmids into HEK293T cells. Luciferase activity was measured and normalized to co‐transfected beta‐galactosidase activity.
6. Three potential TEAD binding sites are not required for transactivation of ITGB2 reporter by YAP‐TEAD. Experiments were similar to these in (E) except that TEAD bindings site mutants were used.
7. GGCATC repeats are important for transactivation of ITGB2 reporter by YAP‐TEAD. Experiments were similar to these in (E) except that indicated mutants were used.
8. Mouse ITGB2 promoter is strongly transactivated by YAP. Reporters of human or mouse 1 kb ITGB2 promoters were co‐transfected with indicated plasmids, and luciferase activity was measured and normalized to co‐transfected beta‐galactosidase activity. Indicated P‐values are in comparison with reporter only group.

Data information: Results are representatives of three independent experiments. Values represent mean ± s.d. from three technical repeats. *P < 0.05; **P < 0.01; ***P < 0.001.Source data are available online for this figure.

Figure 4. PRDM4 cooperates with TEAD to mediate YAP‐induced ITGB2 expression.

• A, B
  Knockdown of PRDM4 inhibits ITGB2 expression. ITGB2 mRNA (A) and protein (B) levels in shRNA‐infected ACHN cells were determined. Relative protein levels of ITGB2 were quantified.
• C, D
  Knockdown of TEAD inhibits ITGB2 expression. ACHN cells infected with scramble or shRNA targeting TEAD1/3/4.
• E
  Illustration of ITGB2 reporters in scale. Potential TEAD and PRDM4 binding sites as well as a GGCATC repeats were indicated.
• F
  YAP, PRDM4, and TEAD2 coordinately activate ITGB2 reporter. HEK293T cells were transfected and luciferase activity was normalized to co‐transfected β‐galactosidase activity. Indicated P‐values are in comparison with the reporter only sample.
• G, H
  YAP‐TEAD2 activates the proximal promoter, and YAP‐PRDM4 activates both the distal and proximal promoters of ITGB2. Experiments were similar to these in (F).
• I
  YAP, TEAD, and PRDM4 bind to a proximal 0.5 kb region of ITGB2 promoter. Control or Flag‐PRDM4‐expressing HepG2 cells were processed for ChIP followed by qPCR detection.
• J
  YAP, TEAD, and PRDM4 could form a heterotrimeric complex. NE of transfected HEK293T cells were immunoprecipitated with indicated antibodies.

Data information: Results are representatives of three independent experiments. Values represent mean ± s.d. from three technical repeats. *P < 0.05; **P < 0.01; ***P < 0.001.Source data are available online for this figure.

To understand how TEAD and PRDM4 regulate ITGB2 expression, we constructed a luciferase reporter driven by a 2 kb ITGB2 promoter (Fig 4E). Co‐expression of YAP with either PRDM4 or TEAD2 potently induced the reporter, which was further enhanced by combination of all three proteins (Fig 4F). We further truncated the promoter into two complementary 1 kb reporters, distal and proximal (Fig 4E). While YAP‐TEAD activates only the proximal reporter (Fig 4G), YAP‐PRDM4 could activate both (Fig 4H). However, chromatin immunoprecipitation (ChIP) of endogenous YAP, TEAD1, or ectopically expressed PRDM4 indicated that they all bind around the −0.5 kb region of the ITGB2 promoter (Fig 4I). Furthermore, deletion of PRDM4 zinc finger motifs, which mediate DNA‐binding, abolished cooperation with YAP to activate ITGB2 reporter (Fig EV4C). The PR domain of PRDM4 is catalytically inactive 35. Indeed, even when we further mutated Y528, a residue conserved in the active site of PRDM9, and G444, a residue important for PR domain structure 35, it still cooperated with YAP to activate the ITGB2 reporter (Fig EV4D). The proximal 1 kb promoter of ITGB2 contains three perfect TEAD binding sites and a stretch of eight GGCATC repeats that also fit TEAD binding consensus (Fig 4E), as well as two GAAAC motifs that match the core sequence of a PRDM4 binding consensus (Fig 4E). However, mutation of sites other than the GGCATC repeats did not affect reporter activation by YAP, TEAD, or PRDM4 (Fig EV4E and F). Nevertheless, deletion of the GGCATC repeats largely eliminated induction by YAP/TEAD (Fig EV4G), suggesting their functional importance. The exact PRDM4 binding sites remains elusive. While mouse ITGB2 could not be induced by YAP, its promoter contains potential TEAD and PRDM4 binding sites, and could be strongly transactivated (Fig EV4H). Therefore, the inability of mouse ITGB2 to respond to YAP is possibly due to lack of other conditions such as additional transcription factor or chromatin condition. Since YAP interacts with TEAD and PRDM4 through distinct domains, we examined the possibility of a heterotrimeric complex. Indeed, when YAP, TEAD4, and PRDM4 were co‐expressed, immunoprecipitation of any of them would pull down the other two (Fig 4J). However, when YAP was omitted, the interaction between TEAD4 and PRDM4 was much weaker (Fig 4J). Taken together, YAP could concomitantly interact with both PRDM4 and TEAD within a proximal region of ITGB2 promoter to induce its expression.

PRDM4 mediates YAP‐induced cell invasion

To further determine the functions of PRDM4, we knocked out this gene in A375 cells, which was confirmed by genomic DNA sequencing (Fig EV5A). YAP‐induced ITGB2 expression was largely suppressed in these cells (Fig 5A and B). The incompleteness of this suppression suggests a basal activity of YAP to induce ITGB2 expression without PRDM4, likely via TEAD. Importantly, knockout of PRDM4 largely blocked the transactivation of ITGB2 reporter by YAP (Fig EV5B), indicating a critical role of PRDM4 in the interaction of YAP with ITGB2 promoter. Furthermore, we observed that knockout of PRDM4 clearly inhibited YAP‐expressing A375 cells to invade the endothelium (Fig 5C and D). Taken together, PRDM4 is a transcription factor partner of YAP mediating YAP‐induced cell invasion.

Figure EV5. Data related to Fig 5 .

1. Confirmation of PRDM4 knockout by genomic DNA sequencing. Fragments of the PRDM4 gene targeted by two independent sgRNAs were PCR amplified from genomic DNA of established A375 cell clones and were sequenced. Sequences of both alleles are shown. Indels are labeled in red. Red arrow lines denote predicted Cas9 cutting sites.
2. PRDM4 knockout blocks transactivation of ITGB2 reporter by YAP. Control and PRDM4 knockout A375 cells were transfected as indicated, and luciferase activity was measured and normalized to co‐transfected beta‐galactosidase activity.
3. YAP‐induced gene signature in MDA‐MB‐231 cells is activated in metastatic prostate cancers. The 25 top YAP‐activated genes in MDA‐MB‐231 cells were analyzed in metastatic prostate cancers. Values are FPKM (fragments per kilobase of transcript per million mapped reads) of dbGaP RNA‐seq data.
4. Quantification of weights of tumors in Fig 5G.
5. Venn diagram of PRDM4‐ and YAP/TAZ‐regulated genes. The overlaps of genes regulated by PRDM4 in murine ES cells and genes regulated by YAP in murine liver, human MCF10A cells, and MDA‐MB‐231 cells were shown by Venn diagram.

Data information: Results are representatives of three independent experiments. Values represent mean ± s.d. from three technical repeats. **P < 0.01; ***P < 0.001.

Figure 5. PRDM4 mediates YAP‐induced cell invasion and deregulation of ITGB2 and PRDM4 expression in metastatic prostate cancer.

• A, B
  Knockout of PRDM4 inhibits ITGB2 induction by YAP. ITGB2 mRNA (A), and protein (B) levels were determined.
• C, D
  Knockout of PRDM4 inhibits YAP‐induced transendothelial migration. Experiments were similar to these in Fig 2F. Scale bars 100 μm. Invasion was quantified in (D).
• E
  The mRNA levels of ITGB2 and PRDM4 are significantly increased in metastatic prostate cancer. N, normal; PCa, localized prostate cancer; Met, metastatic prostate cancer. The boxes are 25^th to 75^th percentiles, and the whiskers are minimum and maximum values. Points beyond the whiskers are outliners.
• F
  Correlation of PRDM4 and ITGB2 mRNA levels in prostate cancer. FPKM values of PRDM4 and ITGB2 in prostate cancer samples were extracted from the TCGA and the dbGaP databases, and their correlation was determined by R ggplot2 package.
• G
  PRDM4 plays an important role in YAP‐induced tumorigenesis. A375 stable cells were injected subcutaneously into nude mice. Tumors were dissected after 17 days.

Data information: Results are representatives of three independent experiments. Values represent mean ± s.d. from three technical repeats. ***P < 0.001.Source data are available online for this figure.

Deregulation of ITGB2 and PRDM4 expression in metastatic prostate cancer

We have demonstrated that LATS1/2 were specifically down‐regulated in metastatic prostate cancer, thus may contribute to metastasis due to YAP activation 25. Therefore, we investigated the expression of ITGB2 and PRDM4 in prostate cancer clinical samples. We compiled RNA‐seq data from the Cancer Genome Atlas (TCGA) dataset 36, which contains only normal and localized prostate cancer, with several datasets of the database of Genotypes and Phenotypes (dbGaP) containing late‐stage prostate cancer specimens 37, 38, 39, 40. This led to a total of 87 cases of normal prostate tissue, 577 cases of localized prostate cancer, and 147 cases of metastatic prostate cancer. Interestingly, the mRNA levels of ITGB2 and PRDM4 are significantly higher in metastatic prostate cancer when compared to normal tissue or clinically localized cancer (Fig 5E). Although the mRNA level of YAP is lower in metastatic cancer samples (Fig 5E), YAP activity is higher as indicated by the activation of a YAP‐induced gene signature 41 (Fig EV5C), likely due to deregulated Hippo pathway activity. Furthermore, there is a significant correlation between ITGB2 and PRDM4 expression in prostate cancer samples (Fig 5F). These data suggest that activation of YAP and PRDM4 may contribute to the elevated ITGB2 expression in metastatic prostate cancer.

PRDM4 contributes to YAP‐induced tumorigenesis

We further examined whether PRDM4 also plays a role in YAP‐induced tumorigenesis. Expression of active YAP clearly promoted tumorigenesis of A375 cells in nude mice as xenografts. However, this effect was abolished by PRDM4 knockout (Figs 5G and EV5D). Nevertheless, the knockout of ITGB2 did not consistently affect YAP‐induced proliferation and tumorigenesis (data not shown). Therefore, YAP‐PRDM4 promotes tumorigenesis likely through other target genes. We thus compared genes regulated by YAP and PRDM4 in profiling datasets. PRDM4‐regulated genes have been studied in PRDM4 knockout murine embryonic stem (ES) cells 42. Due to the lack of a comparable dataset for YAP, we compared it with YAP‐regulated genes in murine liver, as well as human MCF10A and MDA‐MB‐231 cell lines 16, 22, 41. This analysis revealed overlap of genes regulated by PRDM4 and genes regulated by YAP/TAZ (Fig EV5E). Interestingly, gene ontology analysis indicated an enrichment of genes involved in biological processes such as mitosis, cell cycle, DNA replication, cell differentiation, and liver regeneration (Appendix Table S1). The above results indicate that PRDM4 contributes to YAP‐induced tumorigenesis possibly by mediating other YAP target gene expression.

In this study, we demonstrate that YAP promotes cancer cell invasion of the endothelium by transcriptional activation of ITGB2. Interestingly, it has been reported that primary tumors could induce expression of chemokines in distant tissues that results in recruitment of myeloid cells expressing MAC‐1 (an integrin composed of ITGAM and ITGB2) 43. Recruited myeloid cells secrete inflammatory factors that further prepare a microenvironment facilitating metastatic lesion 44. It is therefore intriguing to speculate that YAP‐induced ITGB2 expression could promote adhesion of cancer cells to these inflammatory pre‐metastatic niches, and promote the ensuing extravasation.

The WW domains of YAP/Yki mediate interaction with specific proteins, including regulatory proteins such as LATS1/2 and the AMOT family proteins 7, 45, 46; DNA‐binding transcription factors, such as Smad1, RUNX1/2, p73, Oct4, KLF5, and ErbB4 cytoplasmic domain 47, 48, 49, 50, 51, 52, 53; and cofactors, such as WBP2 and NcoA6 (for Yki) 54, 55, 56. NcoA6 is especially interesting because it plays a critical role in the expression of bona fide Yki target genes 55, 56. However, in mammals, most YAP target genes depend only on TEAD but not the WW domains. Intriguingly, the YAP transcription activation domain is absent in Yki, suggesting a transfer of the transactivation function away from the WW domains during evolution. Nevertheless, functional studies support that the WW domains of YAP play a role in growth control and oncogenic transformation 28, 57. Here, we demonstrated that by binding to the WW domains, PRDM4 plays a specific role in YAP‐induced cell invasion of endothelium through the induction of ITGB2. Furthermore, PRDM4 also contributes to YAP‐induced tumorigenesis. This study thus demonstrates a PRDM4‐mediated transcriptional mechanism of the Hippo pathway in cancer cell invasion and tumor growth.

Most PRDM family proteins, including PRDM4, are catalytically inactive. However, PRDM4 may regulate histone arginine methylation through recruiting protein arginine methyltransferase 5 (PRMT5) 58. In this report, we found that PRDM4 plays an important role in the induction of ITGB2 expression possibly by recruiting other cofactors or by modulating the chromatin conformation. While both YAP and PRDM4 promote embryonic stem cell proliferation and inhibit their differentiation 42, 59, 60, Prdm4 knockout mice develop normally without overt abnormalities 42, 61. Thus, PRDM4 likely plays redundant roles with other transcription factors in embryonic development. Nevertheless, it remains possible that PRDM4 plays a specific role in cancer. It was reported that high expression of PRDM4 strongly correlates with gastric cancer recurrence and worse prognosis 62. Furthermore, in the database of a pathology atlas of the human cancer transcriptome with clinical outcome, the expression of PRDM4 associates unfavorably with patient survival in liver cancer and melanoma 63. This is in consistent with that YAP‐PRDM4 promotes cancer cell invasion and tumor growth. The structure of YAP WW domains in association with a PPXY‐containing ligand has been solved and designing of specific inhibitors targeting interactions mediated by YAP WW domains is therefore possible 64. Our findings thus suggest the YAP‐PRDM4 complex could be a potential target for therapeutic intervention.

Materials and Methods

Antibodies, plasmids, and other materials

Anti‐YAP (sc‐101199), anti‐ITGB2 (sc‐51652) 65, anti‐ITGAM (sc‐1186), and anti‐ITGAX (sc‐1185) antibodies were obtained from Santa Cruz Biotechnology. Anti‐ITGAL (MA11A10) was obtained from Thermo Fisher. Anti‐TEAD1 antibody (610923) was obtained from BD Bioscience. Anti‐Flag (F3165), anti‐alpha‐tubulin (T5168), and anti‐GAPDH (G8795) antibodies were obtained from Sigma. Anti‐Myc (MMS‐150P‐200) and anti‐HA (MMS‐101R) antibodies were obtained from Covance.

pQCXIH‐YAP wild type, S94A, W1 (W199/P202A), W2 (W258/P261A), WW (W199/P202/W258/P261A), 5SA, 5SA‐S94A and 5SA‐WW plasmids, pCMV‐Flag‐YAP wild type, S94A, WW plasmids, pcDNA3‐HA‐MST2, pcDNA3‐HA‐SAV, pcDNA3‐HA‐LATS2, pcDNA3‐HA‐MOB, pRK5‐Myc‐TEAD4, pRK5‐Myc‐TEAD2, and 5× UAS‐luciferase reporter were described before 3, 22. YAP wild type, 5SA, S127A, and S127A‐WW were sub‐cloned into pLVX‐Flag lentiviral vector. PRDM4 was cloned from HeLa cDNA library into pcDNA‐HA vector. PRDM4 mutants YA1 (Y97A), YA2 (Y124A), and YA1/2 (Y97/124A) were generated by QuikChange II XL Site‐Directed Mutagenesis Kit (Agilent). PRDM4 wild type and mutants were then sub‐cloned into pPGS‐CITE‐neo‐HA vector and pLVX‐CMV‐Flag vector. ITGB2 2 kb promoter was cloned from HeLa cell genomic DNA into pGL4.21‐TATA vector. The proximal and distal 1 kb ITGB2 reporter was then made by sub‐cloning from the 2 kb promoter. Mouse Itgb2 promoter was cloned from NIH‐3T3 cell genomic DNA into the pGL4.21‐TATA vector.

Human recombinant fibronectin (F4759), TNF‐α (T0157) and IL‐1β (SRP3083) were purchased from Sigma. Human ICAM‐1 (10346‐H03H) and human VCAM‐1/CD106 protein (Fc Tag) (10113‐H02H‐100) were purchased from Sino Biological. Iodoacetamide (IB0359) was purchased from Sangon Biotech. Common chemicals were from Sigma or Sangon Biotech.

Cell culture

HMLE, ACHN, PC3, A549, HTB182, CRL1848, SF268, HCT116, A375, MDA‐MB‐231, 3T3L1, B16, HUVEC, HEK293, and HEK293T cells were gift from Dr. Kun‐Liang Guan's laboratory at the year 2012. MCF10A cells were purchased (year 2009) from ATCC, where they were characterized by DNA finger printing. Cell line authentication was not done in the laboratory. MCF10A and HMLE cells were cultured in DMEM/F12 medium (Life Technologies) supplemented with 5% horse serum (Life Technologies), 20 ng/ml EGF, 0.5 μg/ml hydrocortisone, 10 μg/ml insulin, 100 ng/ml cholera toxin, and 50 μg/ml penicillin/streptomycin (P/S). ACHN, PC3, A549, HTB182, CRL1848, and SF268 cells were cultured in RPMI1640 medium (Life Technologies) supplemented with 10% FBS (Life Technologies) and 50 μg/ml P/S. HCT116, A375, MDA‐MB‐231, 3T3L1, B16, HEK293, and HEK293T cell were cultured in DMEM medium (Invitrogen) supplemented with 10% FBS and 50 μg/ml P/S. ES‐2 cells were gift from Dr. Hengyu Fan's laboratory and were cultured in McCoy's 5a medium supplemented with 10% FBS and 50 μg/ml P/S. HUVEC cells were cultured in EGM‐2 medium (Lonza) supplemented with 50 μg/ml P/S. BMSSC cells were isolated from femurs and tibias of 7–10 weeks healthy ICR mice (male) by opening bone cavity and flushing out marrow, followed by culture in alpha‐MEM medium (Life Technologies) with 10% FBS, 1× glutamax (Life Technologies) and 50 μg/ml P/S. Mycoplasma test for cell culture was done in a yearly basis using MycoPlasma Detection Kit (biotool.com). Cells used in experiments were within 10 passages from thawing.

Transfection and viral infection

Transfection of plasmids was performed using Lipofectamine (Life Technologies) according to the manufacturer's instructions. Transfection of siRNAs was performed using Lipofectamine RNAiMAX (Life Technologies) according to the manufacturer's instructions.

To generate stable cells, lentiviral or retroviral infections were used. Briefly, HEK293T cells were co‐transfected with viral plasmids and packaging plasmids. Forty‐eight hours after transfection, culture medium was supplemented with 5 μg/ml polybrene, filtered through a 0.45 μm filter, and used to infect cells of interest. Thirty‐six hours after infection, cells were selected with respective antibiotics in culture medium.

RNA interference

siRNAs against human or mouse YAP and TAZ were synthesized by Ribobio:

siRNA h/mYAP#1	CCACCAAGCTAGATAAAGA
siRNA hYAP#2	GGTCAGAGATACTTCTTAA
siRNA h/mYAP#3	GCACCTATCACTCTCGAGA
siRNA hTAZ#1	CCAAATCTCGTGATGAATC
siRNA hTAZ#2	CCGCAGGGCTCATGAGTAT
siRNA mTAZ#1	GATGAATCCGTCCTCGGTG
siRNA mTAZ#2	TCATTGCGAGATTCGGCTG

shRNAs against YAP and TEAD1/3/4 have been previously reported 22. To generate shRNAs against PRDM4, oligonucleotides were cloned into pLKO.1 with the AgeI/EcoRI sites. Oligonucleotide sequences are listed as below:

shRNA hPRDM4 #1‐sense	CCGGGTAGGACATGGTGGTGTGATACTCGAGTATCACACCACCATGTCCTACTTTTTC
shRNA hPRDM4 #1‐antisense	AATTGAAAAAGTAGGACATGGTGGTGTGATACTCGAGTATCACACCACCATGTCCTAC
shRNA hPRDM4 #2‐sense	CCGGAAGGGAAGGAAGGGAATAATCCTCGAGGATTATTCCCTTCCTTCCCTTTTTTTC
shRNA hPRDM4 #2‐antisense	AATTGAAAAAAAGGGAAGGAAGGGAATAATCCTCGAGGATTATTCCCTTCCTTCCCTT

CRISPR‐Cas9 mediated knockout

CRISPR‐Cas9 technology was used to generate ITGB2 and PRDM4 knockout cells. Two independent sgRNAs were designed by the Zhang laboratory CRISPR Design Tool. Annealed oligonucleotides were ligated into the PEP‐KO vector digested with SapI. For knockout, plasmids were transfected into A375 cells. Forty‐eight hours later, cells were selected with 2 μg/ml puromycin for another 48 h. Live cells were collected and re‐plated at low density for colony formation. Individual clones were examined by Western blotting or genomic DNA sequencing. sgRNA hairpin sequences are listed as below:

sgRNA hITGB2#1	GGGCTGCTCTCCCTCGGGTG
sgRNA hITGB2#2	GCCGGGAATGCATCGAGTCG
sgRNA hPRDM4#1	TGAAATGAACCTGAGTCCAG
sgRNA hPRDM4#2	AGCAATGCCTTGCCAGTCTC

Luciferase assays

For luciferase reporter assays, HEK293‐T cells were seeded in 12‐well plates. Cells were then transfected with the reporter, CMV‐β‐gal, and indicated plasmids. 36 h after transfection, cells were lysed and luciferase activity was assayed using the Luciferase Assay System (Promega) following the manufacturer's instructions. All luciferase activities were normalized to β‐galactosidase activity.

The Gal4‐TF library screen has been previously reported 22. Briefly, luciferase assay was carried out as stated above in 96‐well format and the reading from the well with YAP was divided by the reading of the respective well without YAP to get the activation fold. The screen was done in duplicate.

Western blotting

Western blotting was performed using standard protocol. Briefly, cells were lysed with 1% SDS lysis buffer and protein concentration was determined using the BCA Assay Kit (Pierce). Protein samples were resolved on SDS–PAGE and then transferred to PVDF membranes. Membranes were blocked with 5% skimmed milk, incubated sequentially with primary and secondary antibodies, and then washed. Protein expression was detected by ECL Detection Reagent (Pierce).

The expression of ITGB2 protein was determined by Western blotting after resolving on non‐reducing gels. Cells were washed once with cold PBS, lysed with cold lysis buffer (20 mM Tris pH 7.5, 100 mM NaCl, 0.5 mM PMSF, 5 mM iodoacetamide, 50 mM NaF, 1 mM Na₃VO₄, 10 mM EDTA, 10% glycerol, and 1% NP‐40), and then incubated on ice for 30 min. Cell lysates were centrifuged at 15,924 g at 4°C for 15 min. Supernatants were boiled at 65°C for 5 min. Lysates were mixed with 2× SDS loading buffer (0.125M Tris PH 6.8, 4% SDS, 20% glycerol, 0.002% bromophenol blue). Samples were then processed for Western blotting as usual.

RNA isolation and real‐time PCR

Total RNA was isolated from cells using Trizol reagent (Life Technologies). cDNA was synthesized by reverse transcription using PrimeScrip RT Master Mix (Takara) and subjected to real‐time PCR with gene‐specific primers in the presence of SYBR Green (Applied Biosystems). Relative abundance of mRNA was calculated by normalization to hypoxanthine phosphoribosyltransferase 1 (HPRT) mRNA.

Nuclear extraction and immunoprecipitation

To detect YAP‐PRDM4 interaction, nuclear extracts (NE) were made from respective cells and then processed for immunoprecipitation. Cells were washed once with PBS and once with low salt buffer (10 mM HEPES pH 7.9, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA). Cells were then scraped down in low salt buffer with additive (protease inhibitor cocktail, 1 mM DTT, 0.5 mM PMSF, 2.5 mM NaF) and incubated on ice for 5 min. Samples were then passed through 25G needle for 10 times and centrifuged at 1,000 g at 4°C for 15 min. Pellets were then incubated with high salt buffer (20 mM HEPES pH 7.9, 0.4 M NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM DTT, 1 mM PMSF, 1% NP40) on ice for 1 h. Samples were then centrifuged at 20,000 g at 4°C for 10 min, and supernatants were taken as nuclear extracts. Nuclear extracts were diluted four times with dilution buffer (10 mM HEPES pH 7.4, 50 mM NaF, 1 mM EDTA, protease inhibitor cocktail, 1% NP40) and then used for immunoprecipitation. For immunoprecipitation, nuclear extracts were incubated with respective antibodies at 4°C overnight. Protein G or protein A Sepharose (GE Healthcare) was then added and further incubated at 4°C for 2 h. The resins were then washed three times and boiled in SDS sample buffer before analysis by Western blotting.

Chromatin immunoprecipitation

ChIP assays were performed using an EZ‐ChIP kit from Millipore according to the manufacturer's instructions. Briefly, cells were cross‐linked, lysed, and sonicated to generate DNA fragments with an average size of 0.5 kb. ChIP was performed using 5 μg control IgG or antibodies against YAP, TEAD1, or Flag. The immunoprecipitates were then washed and eluted. The eluates were then de‐cross‐linked, and DNA was purified for PCR analysis. PCR primers for ChIP assay:

ITGB2‐chip‐1F (−1.7 kb)	TTCCCTCACTCAGCAGCCT
ITGB2‐chip‐1R	TCAGTTCCTG AAAGGCACCT
ITGB2‐chip‐2F (−1.1 kb)	ACCATTCCCACGTGCCTGT
ITGB2‐chip‐2R	GAATGAGTAT GACTGTGTTC C
ITGB2‐chip‐3F (−0.5 kb)	TTCAGGATGGGAAATGCGTG
ITGB2‐chip‐3R	ATGGGGGCTT CCTAAGCTCA

Cell migration assay

Cell migration assay was performed using BD Falcon Cell culture inserts for 24‐well plates with 8.0 μm pore size. Bottom sides of inserts were pre‐coated with 4 μg/ml ICAM1 protein. A375 stable cells were cultured in serum‐free medium for 12 h before use. 200 μl 2.5 × 10⁵ cells/ml cell suspensions in serum‐free medium was seeded into the inserts, and 800 μl 10% FBS‐containing medium was added into the lower chambers. Cells were cultured for 10 h, and the inserts were then stained with 0.5% crystal violet. Cells on the upper side of the inserts were wiped off, and the bottom sides of the inserts were pictured.

Transendothelial cell invasion

2.5 × 10⁴ HUVEC cells were cultured for 36 h in the upper chamber of transwell inserts and then treated with 10 ng/ml TNF‐α for 12 h. 5 × 10⁴ A375 stable cells were labeled with 5 μM CFSE (Life Technologies) and seeded into transwell filters on top of the HUVEC monolayers in serum‐free medium. 10% FBS‐containing medium was added into lower chambers. Cells were cultured for 10 h, and cells invaded to the bottom side of the inserts were imaged by fluorescence microscopy.

Flow chamber cell adhesion assay

The flow chamber assay was performed as described before 66. A polystyrene petri dish to be used as the lower wall of the chamber was coated with a 5 mm diameter, 20 μg/ml purified human ICAM‐1 Fc or VCAM1 Fc in coating buffer (PBS, 10 mM NaHCO₃, pH 9.0) for 1 h at 37°C, then blocked by 2% human serum albumin in coating buffer for 1 h at 37°C. EDTA‐prewashed cells were diluted to 1 × 10⁶/ml in Ca²⁺‐and Mg²⁺‐free Hank's balanced salt solution with 10 mM HEPES and 0.5% BSA with or without Mn²⁺ and then infused in the flow chamber. Cells were allowed to accumulate for 30 s at 0.3 dyne/cm² and for 10 s at 0.4 dyne/cm². Then, shear stress was increased every 10 s from 1 dyne/cm² up to 32 dyne/cm² in 2‐fold increments. The number of cells remaining bound at the end of each 10‐s interval for 5 min was recorded by microscope.

Flow cytometry

To detect the level of plasma membrane‐localized ITGB2, cells were dissociated by cell dissociation buffer (Thermo Fisher), washed once with cold PBS, and resuspended in PBS containing 5% FBS. Cells were then counted, and 2 × 10⁵ cells were incubated in 200 μl 5% FBS/PBS containing anti‐ITGB2 antibody (Santa Cruz sc‐51652) at 4°C for 1 h. Cells were then washed three times with cold PBS and incubated with PE‐conjugated anti‐mouse IgG secondary antibody (Miltenyi Biotec) at 4°C for 1 h. Cells were then washed three times with cold PBS and resuspended in 500 μl PBS for analysis on Beckman FC 500 MCL. Data were analyzed using FlowJo software. For detection of total ITGB2, cells were fixed with 2% PFA/PBS and permeabilized with 0.2% Tween‐20/PBS. 0.2% Tween‐20 was also added during the primary antibody incubation step and the following wash step.

Tandem affinity purification

Tandem affinity purification purification of YAP has been previously reported 45. Briefly, MCF10A stable cells were lysed with buffer containing 0.3% CHAPS. Lysates were first incubated with anti‐Flag‐antibody‐conjugated resin for 2 h at 4°C and washed three times with lysis buffer. Bound proteins were then eluted with elution buffer containing 200 ng/μl 3× Flag peptide. The eluates were further incubated with streptavidin‐conjugated resin for 2 h at 4°C. Resins were washed and then eluted with buffer containing 4 mM biotin. Samples were then processed for MS/MS analysis.

Prostate cancer gene expression data analysis

Prostate cancer RNA‐seq data from TCGA were downloaded from the Genomic Data Commons 67. RNA‐seq data of late‐stage prostate cancer with accession numbers phs000443 37, 38, phs000915 39, and phs000909 40 were downloaded from dbGaP. All data were normalized to fragments per kilobase of transcript per million mapped reads (FPKM). Boxplots and correlation plots were generated in R.

Xenograft tumorigenesis model

All animal study protocols were approved by the Zhejiang University Animal Care and Use Committee. Nude mice (nu/nu, male 6‐ to 8‐week‐old) were purchased from Shanghai SLAC Laboratory Animal Company and were housed in SPF grade facility. They were injected subcutaneously into the two flanks with 5 × 10⁵ A375 stable cells. After 17 days, tumors were dissected, weighed, and photographed. For lung metastasis assay, 1.5 × 10⁶ of indicated stable A375 cells was injected in 200 μl PBS via lateral tail vein. After 30 days, lungs were dissected and pictured.

PRDM4‐ and YAP‐regulated genes analysis

To determine genes potentially co‐regulated by PRDM4 and YAP, we compared PRDM4‐ and YAP‐regulated gene datasets. Datasets of PRMD4‐dependent genes in murine embryonic stem cells, YAP‐induced genes in MCF10A cell line, and YAP/TAZ‐dependent genes in MDA‐MB‐231 cell line were from NCBI Gene Expression Omnibus Database (accession numbers GSE46308, GSE13218, and GSE59230, respectively). Datasets were analyzed with GEO2R (https://www.ncbi.nlm.nih.gov/geo/info/geo2r.html). For PRMD4‐dependent genes in murine embryonic stem cells and YAP/TAZ‐dependent genes in MDA‐MB‐231 cell line, genes with adjusted P‐value < 0.01 and logFoldChange < −0.4 were considered significant. For YAP‐induced genes in MCF10A cell line, genes with adjusted P‐value < 0.01 and logFoldChange > 1 were considered significant. Dataset of YAP‐induced genes in murine liver is from published paper 16. Enriched gene ontology biological process terms of overlap gene lists were analyzed by Gene Ontology Consortium (http://geneontology.org/) with P‐value < 0.05.

Statistical analysis

The P‐values were determined by Student's t‐tests.

Author contributions

BZ and HL conceived the project. BZ, HL, and XD designed the study. HL, XD, XC, HY, XJ, HZ, SS, YS, and HZ performed experiments. BZ, HL, XD, JCZ, and JY analyzed data. JC, LL, and X‐HF provided conceptual advices. BZ and HL wrote the manuscript with comments from all authors.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Expanded View

Review Process File

Source Data for Figure 1

Source Data for Figure 2

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

EMBO Reports (2018) 19: e45180