Induction of PARP7 Creates a Vulnerability for Growth Inhibition by RBN2397 in Prostate Cancer Cells

The ADP-ribosyltransferase PARP7 modulates protein function by conjugating ADP-ribose to the side chains of acceptor amino acids. PARP7 has been shown to affect gene expression in prostate cancer cells and certain other cell types by mechanisms that include transcription factor ADP-ribosylation. Here, we use a recently developed catalytic inhibitor to PARP7, RBN2397, to study the effects of PARP7 inhibition in androgen receptor (AR)-positive and AR-negative prostate cancer cells. We find that RBN2397 has nanomolar potency for inhibiting androgen-induced ADP-ribosylation of the AR. RBN2397 inhibits the growth of prostate cancer cells in culture when cells are treated with ligands that activate the AR, or the aryl hydrocarbon receptor, and induce PARP7 expression. We show that the growth-inhibitory effects of RBN2397 are distinct from its enhancement of IFN signaling recently shown to promote tumor immunogenicity. RBN2397 treatment also induces trapping of PARP7 in a detergent-resistant fraction within the nucleus, which is reminiscent of how inhibitors such as talazoparib affect PARP1 compartmentalization. Because PARP7 is expressed in AR-negative metastatic tumors and RBN2397 can affect cancer cells through multiple mechanisms, PARP7 may be an actionable target in advanced prostate cancer. Significance: RBN2397 is a potent and selective inhibitor of PARP7 that reduces the growth of prostate cancer cells, including a model for treatment-emergent neuroendocrine prostate cancer. RBN2397 induces PARP7 trapping on chromatin, suggesting its mechanism of action might be similar to clinically used PARP1 inhibitors.


INTRODUCTION
The PARP family of enzymes contributes to a variety of cellular pathways, many of which occur in the nucleus and involve events associated with the regulation of chromatin structure, transcription, and DNA damage signaling and DNA repair [1,2]. PARP enzymes contain a conserved catalytic domain that uses NAD + as a cofactor for ADP-ribose conjugation to the side chains of acceptor amino acids, and in the case of enzymes that generate polymers, conjugation to ADP-ribose itself [3]. The founding member of the PARP family, poly-ADP-ribosyltransferase 1 (PARP1) has been a major focus in the field since its discovery. Detailed, mechanistic analysis of PARP1 has been complemented by the development of highly specific compounds that compete for NAD + binding, and as a consequence, inhibit PARP1 enzymatic function [4]. Clinical trials with PARP1 inhibitors such as Olaparib have shown patient benefit in ovarian, breast, and prostate cancer, particularly in tumors that harbor mutations in DNA repair genes, notably BRCA1/2 and ATM [5].
Most PARP family members mediate a single round of ADP-ribose attachment and are categorized functionally as mono-ADPribosyltransferases [3].
Mono-ADPribosylation provides a post-translational mechanism for modulating protein function, likely in a reversible manner since cells encode hydrolases that can remove ADPribose from amino acids [6]. TIPARP/PARP7, a mono-ADPribosyltransferase expressed in multiple cell and tissue types, was characterized as a key effector of signaling and gene expression mediated by the aryl hydrocarbon receptor (AHR) in the context of detoxification pathways in liver [7]. Our group identified PARP7 in a prostate cancer cell signaling pathway that controls assembly of a multi-subunit complex containing the androgen receptor (AR) [8]. In this nuclear pathway, PARP7 "writes" ADP-ribose onto cysteine (Cys) residues in AR. The ADP-ribosyl-Cys sites are subsequently "read" by macrodomains in PARP9; this provides a highly selective mechanism for assembling a complex that contains AR, PARP9, and the ubiquitin E3 ligase DTX3L. Depletion of the DTX3L affects the expression of a subset of ARregulated genes [8], suggesting PARP7 and assembly of the complex has a regulatory role in transcription. PARP7 has been functionally linked to other transcription factors including the estrogen receptor and liver X receptors [9,10]. It seems plausible that PARP7 contributes to a wide variety of gene expression pathways through mechanisms that include, but are not limited to, transcription factor ADP-ribosylation.
PARP7 also exerts an effect on transcription through a signaling-based mechanism involving the kinase, TBK1. PARP7 negatively regulates TBK1 kinase activity, which restrains phosphorylation and activation of the transcription factor IRF3 [11]. In this pathway, PARP7 was proposed to serve as a brake for sensing cytosolic nucleic acids that trigger Type I IFN signaling [12]. Ribon Therapeutics developed RBN2397, a first-in-class mono-ADP-ribosyltransferase inhibitor, and showed it blocks PARP7 negative regulation of TBK1 [12]. Treating cells with RBN2397 promotes TBK1 phosphorylation of IRF3, restoration of Type I IFN signaling and effects on gene expression, and enhances tumor immunogenicity in lung cancer models [12]. These data provide the rationale for targeting PARP7 with RBN2397, which is orally bioavailable and under evaluation in a Phase 1 clinical trial (NCT04053673).
Here, we explore the use of RBN2397 as a PARP7 inhibitor in prostate cancer cell lines. Consistent with a previous report [12], RBN2397 has selectivity for PARP7 versus other PARP family members in ADPribosylation when assayed using core histones as substrates. In cells, we show that RBN2397 has nanomolar potency for inhibiting PARP7 ADP-ribosylation of AR. We determined that RBN2397 inhibits the growth of prostate cancer cells dependent on transcriptional induction of PARP7, that can be achieved by treating cells with androgen to activate AR, or with ligands that activate AHR. RBN2397 exerts a nuclear trapping effect on PARP7, analogous to the effects of clinically-used inhibitors to PARP1 [13,14]. Finally, we used chemical inhibitors to TBK1 and JAK1/2 to show that RBN2397 inhibition of prostate cancer cell growth can be distinguished from the RBN2397 effect on PARP7 regulation of Type I IFN signaling. The available data indicate that PARP7 inhibition with RBN2397 exerts effects on cancer cells through multiple mechanisms.

RESULTS
RBN2397 reduces the growth of lung cancer models by inhibiting PARP7, restoring IFN signaling, and increasing tumor immunogenicity [12]. We evaluated whether RBN2397 has potential utility in prostate cancer using biochemical approaches and cell growth assays in ARpositive (AR+) and AR-negative (AR-) prostate cancer lines. In previous work we reported that PARP7 directly ADPribosylates eleven Cys residues in AR [8]. AR ADP-ribosylation is induced by androgen in prostate cancer cells because the PARP7 gene is a direct target of AR, and because PARP7 selectively ADPribosylates the agonist conformation of AR [8,15]. Androgen-induced ADPribosylation of AR in cells and detection with fluorescently labeled Af1521 (Fl-Af1521) on blots [16] provides a sensitive assay to assess the effect of chemical inhibition of PARP7 in cells. To this end, we treated PC3-AR cells with synthetic androgen (R1881) to induce PARP7 transcription and AR-ADP-ribosylation, in the absence and presence of the PARP7 inhibitor RBN2397. We found that AR ADPribosylation was reduced by 1 nM RBN2397 and essentially eliminated by 3 nM RBN2397 (Fig. 1A). RBN2397 also inhibited androgen-induced ADPribosylation of AR in VCaP cells (Fig. 1B). The selectivity of RBN2397 for PARP7 versus the PARP family was characterized previously[12], primarily with biophysical assays that measure competitive binding with NAD + , which is the basis of how RBN2397 inhibits PARP7 enzyme function [12]. We measured the effect of RBN2397 on substrate ADP-ribosylation using recombinant PARP7 and PARP1 and a mixture of Histone H2A and H2B to determine EC50 values. In vitro, the EC50 of RBN2397 for PARP7 is approximately 7.6 nM, whereas the EC50 of RBN2397 for PARP1 is 110 nM (Fig. 1C). In the presence of 7.6 nM RBN2397, there is relatively little inhibition of other PARP enzymes, including family members with expression levels in prostate tumors that exceed the level of PARP7 (Fig. 1D).
RBN2397 treatment of PC3-AR cells also increases the level of PARP7 protein detected by immunoblotting, suggesting the inhibitor might stabilize PARP7 protein (Fig. 1A). We performed a time course of cycloheximide treatment and found RBN2397 increases the protein half-life of PARP7 approximately 4-fold (Fig. 1E). Stabilization of PARP7 protein provides additional evidence for the on-target effects of RBN2397 in prostate cancer cells. Lastly, we examined the effect of RBN2397 on PARP7 auto-ADP-ribosylation, as automodification is a common biochemical property of PARP enzymes. RBN2397 treatment of cells eliminated PARP7 auto-ADP-ribosylation in cells, detected by immunoprecipitation of PARP7 and using Fl-Af1521 to probe for mono-ADP-ribose (Fig. 1F). Treating cells with the proteasome inhibitor MG132 blocks PARP7 degradation and increases the level of auto-ADP-ribosylated protein (Fig.  1F).
We examined the effects of RBN2397 on growth with cell lines used to model prostate cancer and therapy resistance. VCaP cells encode WT AR and the V7 variant, and are used to model castrateresistance owing to amplification of AR. CWR22Rv1 is an informative model for therapy resistance mediated by AR splice variants that lack the ligand binding domain. PC3-AR is a PC3 derivative engineered to express WT AR, which drives a gene expression profile that overlaps other prostate lines [17]. PC3 cells are AR negative, resistant to therapies based on androgen deprivation, highly motile, and used to study prostate cancer metastasis. Overall, treating the four prostate lines with RBN2397 had little or no effect on growth. RBN2397 slightly increased the growth of VCaP, it reduced the growth of CWR22Rv1 and PC3 cells, and had no significant effect on PC3-AR cells ( Fig. 2A). While the effects of RBN2397 treatment were statistically significant in some cases, the effects were quantitatively small. Reasoning that low PARP7 expression levels might limit the effect of RBN2397 on cell growth, we treated cells with androgen (R1881) to activate AR and induce PARP7 expression. We found that RBN2397 was growth inhibitory in androgen-treated VCaP, CWR22Rv1, and PC3-AR cells ( Fig. 2A). It is well established that androgen treatment is growth-repressive in some prostate cancer lines. R1881 was growth inhibitory in VCaP and PC3-AR cells; RBN2397 further reduced the growth of VCaP and PC3-AR cells, and in CWR22Rv1 cells had greater effects on growth in the presence of R1881 ( Fig. 2A). Dose response curves for cell growth in the presence of androgen revealed RBN2397 had half-maximal effects in the low nanomolar range ( Fig.  2B; Supplementary Table 1), which is consistent with its potency for inhibiting PARP7 ADP-ribosylation of AR in cells (Fig. 1A). Growth inhibition by RBN2397 does not involve accumulation of cells in a specific phase of the cell cycle, though within the data a G1 effect of androgen is apparent in VCaP and PC3-AR cells (Supplemental Fig. 1A). By phase contrast microscopy, RBN2397 promotes the appearance of rounded cells (Fig. 2C) and RBN2397 can increase the number of dead cells detected by Trypan blue staining in CWR22Rv1 and PC3-AR cells (Fig. 2D). Overall, the data suggest that the growth inhibition by RBN2397 occurs through several mechanisms, but the effects do not seem to involve blocking the cell cycle at a specific phase to the extent it can be measured by DNA content.
We next tested if the androgendependence of the RBN2397 growth inhibition is associated with PARP7 expression. Using an siRNA that targets PARP7 (siPARP7), we determined the cell growth reduction caused by RBN2397 in androgen-treated PC3-AR cells was reduced by PARP7 knockdown (Fig. 3A). PARP7 protein levels revealed by immunoblotting after R1881 treatment are very low, but the additional treatment with RBN2397 stabilizes PARP7 protein and facilitates its detection by immunoblotting ( Fig. 3A-C). Reducing PARP7 levels with shPARP7 (stably-expressed) gave similar results. Though the efficiency of PARP7 depletion, and the growth differences +RBN2397 in control and shPARP7 cell lines were quantitively small, the effect of PARP7 knockdown was statistically significant in both PC3-AR and CWR22Rv1 cells (Fig. 3B, C).
One mechanism by which androgens repress the growth of prostate cancer lines through AR is by induction of the CYCLINdependent kinase inhibitor p21 [18]. Given the growth repressive effects of RBN2397 were associated with androgen treatment, we tested whether the RBN2397 effect on growth involves p21 expression. By immunoblotting, p21 protein expression in VCaP and PC3-AR cells is positively regulated by androgen, and p21 levels were increased further by RBN2397 treatment (Fig. 3D). The levels of p21 were not affected by androgen in AR-positive CWR22Rv1 cells, and as expected, there was no androgen effect in AR-negative PC3 cells (Fig. 3D). Using siRNA to reduce p21 expression, we found that p21 depletion resulted in a partial rescue of the growth inhibitory effect of RBN2397 (Fig.  3E). Similarly, eliminating p21 expression by making a deletion in the CDKN1A gene in PC3-AR cells (CRISPR-CAS9) partially rescued the effects of RBN2397 (Fig. 3F). These data show that in the context of androgen signaling, the growth inhibitory effects of RBN2397 can be partly mediated through p21, but this mechanism is not operational in all prostate cancer cell lines.
Growth inhibition by RBN2397 in androgen-treated VCaP, CWR22Rv1, and PC3-AR cells raises the intriguing possibility that PARP7 might be an actionable target in prostate cancer. As an alternative approach for inducing PARP7 in prostate cancer cells, we turned to the transcription factor AHR, a member of the basic helix-loop-helix /Per-AHR nuclear translocator-Sim protein family. AHR is known to control PARP7 expression in other settings, and has been studied in the context of detoxification mechanisms in liver [7,19].
We tested two ligands with potent AHR agonist activity for PARP7 induction in prostate cancer. 6-Formylindolo[3,2b]carbazole (FICZ) is a naturally occurring tryptophan photoproduct, and 10-Chloro-7H-benzimidazo[2,1a]benz [de]isoquinolin-7-one (10-CL-BBQ, or BBQ in short) is a synthetic ligand. FICZ and BBQ induce AHR target genes through a chaperone-mediated mechanism[20]. FICZ and BBQ (10-100 nM) were both effective for sensitizing cells to growth inhibition by RBN2397 (Supplementary Table 1), and the growth inhibitory effect was associated with PARP7 protein stabilization detected by immunoblotting (Fig. 4A). These data show that PARP7 can be induced by AHR in AR-positive (CWR22Rv1 and VCaP) and AR-negative (PC3 and DU145) prostate cancer cells for the purpose of sensitizing cells to RBN2397. Combining PARP7 induction with BBQ and RBN2397 treatment reduced the protein levels of CYCLIN A and CYCLIN B in PC3, DU145, and CWR22Rv1, though the effect was small and not observed in VCaP cells (Fig. 4B). BBQ and RBN2397 treatment increased the G1 fraction in PC3 and DU145 cells, and slightly reduced the S-phase fraction in CWR22Rv1 cells (Supplementary Fig.  1B). By phase contrast microscopy there was a reduction in the number of adherent cells in PC3, DU145, and CWR22Rv1 cells treated with BBQ and RBN2397 (Fig. 4C), which is consistent with our finding that combining BBQ and RBN2397 increased the number of Trypan blue-positive cells in PC3, DU145, and CWR22Rv1 cells (Fig.  4D, E).
The cell growth and immunoblotting data for PARP7, CYCLIN A, CYCLIN B, and the Trypan blue detection of dead cells, imply the BBQ and RBN2397 combination is more effective in PC3, DU145, and CWR22Rv1 than in VCaP cells (Supplementary Table 1). The RBN2397 dose response curves for cell growth were also consistent with this notion, where PC3 and DU145 cells were inhibited by nanomolar concentrations of RBN2397 and the overall response was greater than VCaP cells (Fig. 4F-H). To determine if the RBN2397 effect is dependent on PARP7 expression, we compared the RBN2397 sensitivity of control and PARP7 knockdown cells treated with BBQ. Although the PARP7 knockdown was partial, the RBN2397 effect was reduced significantly in shPARP7 compared with shGFP cells (Fig. 4I, J). Taken together, the data show that RBN2397 can be used to inhibit growth in prostate cancer cells where PARP7 is induced by AHR signaling.
The cellular effects of RBN2397 in preclinical models of lung cancer were attributed to reversal of a PARP7-TBK1 inhibitory mechanism, resulting in IRF3 phosphorylation, restored IFN signaling, and increased tumor immunogenicity [12]. Because IRF3 can also regulate the cell cycle [21], we queried whether TBK1 activity and IFN signaling are involved in the growth inhibitory effects of RBN2397 in prostate cancer cells. To this end, we employed specific inhibitors to TBK1 (GSK8612; GSK) and JAK1/2 (Ruxolitnib; Ruxo) in prostate cancer cell lines and induced PARP7 via AHR (PC3, DU145) and AR (PC3-AR) (Fig. 5). In PC3 cells, RBN2397 treatment increased the basal level of phospho-STAT1, an effect that was eliminated by JAK1/2 (Ruxo) and TBK1 (GSK) inhibitors (Fig. 5A). These data are consistent with TBK1 acting upstream of JAK1/2 in PC3 prostate cancer cells, as shown in other cell types [12]. Induction of PARP7 by the AHR ligand BBQ does not further increase STAT1 phosphorylation, suggesting the relatively low, basal level of PARP7 expression is sufficient for the RBN2397 effect on STAT1 phosphorylation in PC3 cells (Fig. 5A, left panel). This contrasts with the growth inhibitory effect of RBN2397, which is responsive to PARP7 expression induced by treating cells with ligands to AHR and AR (Figs. 2, 4).
The growth inhibitory effect of RBN2397 in PC3 cells is not reversed by JAK1/2 and TBK1 kinase inhibition with Ruxo and GSK, respectively (Fig. 5A, (Fig. 5B, left panel). Thus, the growth reduction in DU145 cells caused by RBN2397 is not associated with an effect on STAT1 phosphorylation, and blocking JAK1/2 and TBK1 kinase function in this line slightly increases the growth inhibitory effect of RBN2397 (Fig. 5B, right panel). In PC3-AR cells, the signaling and growth data are very similar to PC3 cells; RBN2397 increases STAT1 phosphorylation, and blockade of JAK1/2 and TBK1 inhibitors largely eliminates the RBN2397 effect on STAT1 phosphorylation without affecting the growth inhibitory effect of RBN2397 (Fig. 5C). In VCaP and CWR22Rv1 cells, RBN2397 also has no obvious effect on basal STAT1 phosphorylation (Supplementary Fig. 2). From these data, we conclude the growth inhibitory effect of RBN2397 in prostate cancer cells is separable from its effects on IFN signaling that are linked to the kinase activities of JAK1/2 and TBK1.
Chemical inhibitors to PARP1 exert effects on cells by blocking enzyme function, but also via cytotoxic effects attributed to stabilizing PARP-chromatin interactions in a process termed trapping [13]. Druginduced trapping of PARP1 can be detected biochemically by immunoblotting the detergent-resistant nuclear fraction [13]. To test if RBN2397 induces PARP7 trapping, we induced its expression in PC3-AR cells with R1881 +RBN2397 and performed cell fractionation and immunoblotting. Antibodies for TUBULIN and LAMIN A were used to confirm the release of soluble proteins and recovery of Triton X-100 resistant, nuclear fraction. As shown in previous figures, R1881 treatment induces PARP7 which is released into the soluble fraction and virtually absent from the insoluble fraction (Fig. 6A, lanes 3 and 7). By contrast, the combination of R1881 and RBN2397 results in a pool of PARP7 that partitions to the insoluble fraction (Fig. 6A, lanes 4 and  8). In PC3 cells, inducing PARP7 using BBQ to activate AHR also resulted in PARP7 trapping (Fig. 6B, lanes 4 and 8).
The proportion of PARP7 trapped by RBN2397 treatment is similar to that observed when PARP1 is trapped by Olaparib and Niraparib [13]. RBN2397mediated trapping of PARP7 on chromatin could contribute to the growth inhibition observed in prostate cancer cells.
It has been noted that for PARP1 and PARP7, genetic knockout is not equivalent to drug-induced inhibition of the enzymes[12, 13]. A logical interpretation of these findings is that drug binding to PARP1 and PARP7 exerts dominant negative effects that are in addition to the cellular changes associated specifically with preventing substrate ADPribosylation. We therefore considered the possibility that RBN2397 inhibition of prostate cancer cell growth involves a dominant negative effect. An alternative explanation, at least in prostate cancer cells, is that gene expression induced by AR and AHR creates a PARP7 dependency for cell growth. We chose the prostate cancer line C4-2b for this analysis. The basal level of PARP7 expression is extremely low in C4-2b cells, RBN2397 treatment alone has no effect on growth of these cells, and activation of AHR signaling sensitizes the cells to RBN2397 inhibition (Fig. 6C).
To test for a dominant negative effect of RBN2397-inhibited PARP7, ectopic PARP7 was expressed with an N-terminal Avi-tag in C4-2b cells (Avi-PARP7; Fig.  6D). Avi-PARP7 is enzymatically active in these cells since it ADP-ribosylates AR and the modification is blocked by RBN2397 (Fig. 6D). We found that expression of Avi-PARP7 is sufficient to confer a growth inhibition response to RBN2397 (Fig. 6E). Growth inhibition by RBN2397 in these cells is enhanced by AHR activation with BBQ, which increases the cellular level of PARP7 by induction of endogenous PARP7. The possibility that additional factors regulated through AHR contribute to the RBN2397 vulnerability cannot be excluded.
As a first step towards evaluating whether PARP7 levels in human prostate cancer are potentially actionable with RBN2397, we analyzed PARP7 gene expression data from primary prostate tumors and metastatic AR+ and AR-prostate tumors [22]. To assess PARP7 mRNA levels, we used data from the online resource recount3, which uniformly reprocesses publicly available RNA-seq datasets using a Monorail analysis pipeline [23]. This allowed us to compare transcriptomic data from different studies, including cell line and patient samples. We observe there are no changes in PARP7 mRNA levels in primary tumors compared to normal prostate cells, but expression is reduced in metastatic tumors, and the difference is greater for metastatic tumors that are AR+ (Fig. 7A). Since growth inhibition by RBN2397 in cell culture is dependent on PARP7 expression, we used our RNA-seq data [17] from VCaP cells treated +R1881 and processed through recount3 to define basal and induced levels of PARP7 (Fig. 7A). We then used the induced level to assign a cut-off for RBN2397 vulnerability in prostate tumors. By this criterion, 50% of primary tumors, 41% of metastatic AR-and 11% of AR+ tumors are predicted to have PARP7 expression levels that are sufficient for a response to RBN2397. Approximately 7% of metastatic prostate tumors (34/444) show genomic copy number alternations in PARP7 (Supplementary Fig. 3), which could contribute to protein expression levels.
To query how PARP7 expression is potentially regulated in tumors, we computed Spearman's rank correlation between PARP7, AR, and AHR gene expression. We found a moderate correlation between PARP7-AR expression (rs = 0.30, p < 0.001, n = 497) and a strong positive correlation between PARP7 and AHR expression (rs = 0.56, p < 0.001, n = 497) (Fig. 7B left) in primary tumors. In both AR+ and AR-metastatic tumors the expression of PARP7 and AR is not significantly correlated. However, the PARP7-AHR correlation is significant and is especially striking in AR-tumors (rs = 0.62, p < 0.001, n = 27) (Fig. 7B middle,  left). These results suggest that PARP7 expression in primary tumors could be influenced by AR and AHR, but in metastatic tumors the association with AR is lost, whereas the AHR influence is retained.
PARP7 has been proposed to restrain type I IFN signaling in lung cancer models, an effect relieved by RBN2397 [12]. We analyzed the PARP7 influence on type I IFN signaling in primary prostate tumors, using gene set variation analysis (GSVA) [24]. We found that lung cancer (LUAD) samples exhibit a much higher enrichment score and pathway activity for genes that are up-regulated in response to IFNa proteins (Type I IFN response) compared to prostate cancer (PRAD) samples (Fig.  7C). Moreover, the enrichment scores in LUAD shows an association with PARP7 expression level, but this is not the case in PRAD (Fig. 7C). This suggests that either PARP7 does not influence Type I IFN signaling in primary prostate cancer, or that low levels of PARP7 are sufficient for the effect.

RBN2397 was developed by Ribon
Therapeutics through optimization of an unselective mono-ADP-ribosyltransferase inhibitor [12]. Prior biochemical characterization of RBN2397 included using a probe displacement assay to demonstrate potent PARP7 inhibition (IC50 < 3 nM) and PARP7 selectivity (>50-fold) within the PARP family [12]. The same group assessed the effect of RBN2397 in PARP7-overexpressing cells by staining for MARylation [12,25]. These and other data demonstrated the potency and selectivity of RBN2397 for PARP7. We characterized the effect of RBN2397 in prostate cancer cells, first by examining its ability to inhibit PARP7 ADP-ribosylation of AR.
In prostate cancer cells, AR is a well-defined PARP7 substrate in which the ADPribosylation sites have been defined and characterized by mutagenesis [8]. In PC3-AR cells grown without androgen, virtually no ADP-ribose is detected on AR using Fl-Af1521 as a probe. Androgen treatment stimulates AR induction of endogenous PARP7, which then ADP-ribosylates AR on Cys sites primarily within its transactivation domain [8]. Biochemical and RNA-seq data are both consistent with PARP7 operating as the androgen-regulated mono-ADPribosyltransferase that modifies nuclear AR in prostate cancer cells [8]. We found that low nanomolar concentrations of RBN2397 were sufficient to eliminate PARP7 ADPribosylation of AR in PC3-AR cells. RBN2397 treatment also prevented ADPribosylation of AR in VCaP cells, which demonstrates the effectiveness of RBN2397 in a setting where both PARP7 and its inducer/substrate (AR) are expressed from endogenous genes. Since RBN2397 can also inhibit PARP7 ADPribosylation of the estrogen receptor [9], RBN2397 might be useful to target signaling and growth in breast cancer.
A second finding in our study was that RBN2397 can inhibit the growth of prostate cancer cells, and that this occurs under conditions where PARP7 undergoes induced expression. In prostate cancer cell lines, PARP7 can be induced by treating cells with agonists for AR and AHR. The fact that there is little or no growth inhibition by RBN2397 prior to PARP7 induction suggests that the contribution of basal levels of PARP7 to cell growth of these cells in culture is minimal. Consistent with this view, DepMap analysis of VCaP and CWR22Rv1 cells show essentially no dependency on basal PARP7 expression. The growth inhibition by RBN2397 under conditions of PARP7 induction is suggestive of a dominant-negative effect of RBN2397-bound PARP7. Support for a dominant-negative mechanism derives from showing that ectopic PARP7 expression is sufficient to partially sensitize C4-2b cells to the growth inhibition by RBN2397. The growth inhibition of RBN2397 mediated through ectopic PARP7 was not as penetrant as the effect of RBN2397 in androgen-treated cells. The reduced penetrance might be due to an insufficient level of ectopic PARP7, or that androgen signaling through AR affects the expression of genes (in addition to PARP7) that promotes RBN2397 sensitivity. The basal level of PARP7 expression in some cell types such as NCI-H1373 is sufficiently high to confer a sensitivity to nanomolar RBN2397 [23]. In DepMap, NCI-H1373 cells display a PARP7-dependency.
The available data indicate that RBN2397 treatment of PARP7-expressing cancer cells has at least two effects, modulation of IFN signaling and reduction of cell growth. Whether these effects are linked or separate might be context-dependent, but our data favors the conclusion the effects are separate. RBN2397 enhancement of basal IFN signaling reported by Ribon was shown to occur because PARP7 can restrain TBK1 activity [11,12]. We found that RBN2397 increases basal IFN signaling in PC3 and PC3-AR cells, as RBN2397 treatment was sufficient to increase phospho-STAT1 levels. The effect on phospho-STAT1 was abrogated by inhibiting either TBK1 or JAK1/2, suggesting the PARP7 effect on TBK1 and IFN signaling in PC3 cells is comparable to some other cell types [12]. And although IFN signaling can be growth repressive, the inhibitory effect of RBN2397 on cell growth in our experiments is not linked to the RBN2397 effect on IFN signaling. Thus, in the aforementioned experiment with TBK1 and JAK1/2 inhibitors that eliminated the RBN2397 enhancement of IFN signaling, blocking TBK1 and JAK1/2 did not alter the RBN2397 inhibition of cell growth. Moreover, RBN2397 had no obvious effect on basal IFN signaling in VCaP, CWR22Rv1, and DU145 cells, despite the fact RBN2397 inhibited growth of these cells. It was noted that RBN2397 can restore basal IFN signaling in CT26 cells without affecting cell proliferation in culture [12]. Whether PARP7 and RBN2397 affect IFN signaling, cell growth, or both, is expected to be dependent on PARP7 expression levels, and also the context. There is an association between the enrichment score for Type 1 IFN signaling in LUAD but not in PRAD. The contribution of PARP7 to IFN signaling might be dependent on the tumor type, but it is also possible that low levels of PARP7 are sufficient for an effect on TBK1.
Potential therapeutic benefits of RBN2397 could depend on PARP7 expression levels in tumors, particularly if the growth inhibitory mechanism in tumors involves a dominant negative effect as suggested by our cell culture data. Using recount3, we were able to compare RNA-seq data from cultured cell experiments to data from human prostate tumors. In VCaP cells, growth inhibition by RBN2397 required the addition of agonists for AR or AHR to induce PARP7 expression. Thus, PARP7 levels prior to, and after, androgen treatment of VCaP cells allowed us to define a threshold for PARP7 expression that results in a growth inhibitory response to RBN2397. Using this value as a cutoff, we estimate that about half of all prostate cancers should express sufficient PARP7 to be vulnerable to RBN2397. In ARmetastatic tumors, PARP7 RNA levels are higher on average than AR+ tumors; this suggests AR status might help predict the utility of RBN2397 in a subgroup of patients.
In summary, we have shown that PARP7 expression confers sensitivity to growth inhibition by RBN2397. PARP7 expression levels can be regulated in prostate cancer cells using ligands for AR and AHR. PARP7 expression in primary prostate cancer tumors is correlated with both AHR and AR, which is logical given the TIPARP gene is a direct target of both transcription factors.
Interestingly, the positive correlation between AR and PARP7 expression appears to be lost in metastatic prostate tumors, suggesting that PARP7 becomes highly dependent on AHR signaling for its expression in advanced disease. Combining AHR agonists with RBN2397 might provide PARP7-based strategy for inhibiting advanced prostate cancer.

Lentivirus Production and Stable Cell Lines
PC3-AR, PC3-AR(HA-PARP7) and PC3-AR(shGFP or shPARP7) have been previously described [8]. Lentiviral plasmid shPARP7 (or shGFP control, or pLH3/AviTag-PARP7) and two accessory plasmids pMD2g and psPAX2 (half as much as the lentiviral plasmid) were transfected into HEK293T cells with transfection reagent ViaFect (Promega PRE4981). After 16 hrs of transfection, growth medium was changed to DMEM + 35% FBS, and cells grew another 24 hrs. The growth medium which contained lentiviruses was transferred to a conical tube. Cell debris was removed with centrifugation at 1,000 RPM. The virus solution was passed through 0.4 µm filter, concentrated with Lenti-X Concentrator (Takara 631231). CWR22Rv1 or C4-2b cells were infected with virus for 24 hrs in the presence of 8 µg/ml of polybrene in the growth medium. After that, cells were changed into fresh growth medium, further grew for 2-3 doubling time, and then selected with 2 µg/ml of puromycin.
The sgRNA-containing pX330 plasmids were verified by Sanger sequencing (Eurofins) using U6-specific primers. To generate PC3-AR cells with p21 deletion, the cells were grown to sub-confluency, and transiently transfected with the two CAS9expressing plasmids (sg-p21-1 and sg-p21-2; 2.0 µg each), along with 0.5 µg pMSCVpuro vector (Clontech) containing a puromycin selectable marker. Transfections were performed using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol. After 24 hours of transfection, cells were selected in the presence of puromycin (2 µg/ml) for 48 hours, after which single clones were isolated via serial dilutions of the transfected pool. To identify clones with p21 deletion, we performed genotyping on DNA isolated from individual clones. Briefly, candidate clones were expanded in culture, lysed overnight at 55°C in lysis buffer (100 mM NaCl, 10 mM Tris-HCl pH 8, 25 mM EDTA, 0.5% SDS) supplemented with 20 µg of proteinase K. DNA was isolated from the lysed cells using phenol chloroform/isoamyl alcohol extraction. Genotyping was performed via PCR amplification of the targeted p21 locus with primers flanking the two predicted CAS9 cleavage sites followed by Sanger sequencing (Eurofins). The following primers were used to amplify a 339 bp sequence spanning the two sgRNA target sites: p21-F: 5'-TCACCTGAGGTGACACAGCAAAGC-3', p21-R: 5'-GGCCCCGTGGGAAGGTAGAGCTT-3'. Immunoblotting of the individual p21 clonal knockout cells using an anti-p21 antibody was further used to confirm p21 deletion.

Protein Half-Life Measurement
PC3-AR(HA-PARP7) cells, with or without 1 hour of 10 nM RBN2397 pre-treatment, were incubated with 0.1 mg/ml of Cycloheximide for indicated time periods. After medium aspiration, cells were lysed in 1xSDS-PAGE loading buffer, followed with heating, sonication, SDS-PAGE, HA-PARP7 Western Blot (with TUBULIN as control), and quantification on Odyssey CLx (LI-COR).

In Vitro PARP Assays
The assays were conducted by BPS Bioscience (San Diego, CA) in duplicates (PARP7) or triplicates (PARP1 etc) at Room Temperature for 2 hrs with reaction mixture containing PARP, inhibitors, b-NAD + , and Biotin-b-NAD + in 96-well plates pre-coated with histone substrates. After enzymatic reactions, 50 μl of Streptavidinhorseradish peroxidase (prepared with Blocking Buffer) was added to each well and the plate was incubated at room temperature for an additional 30 min. The wells were washed again and 100 µl ELISA ECL substrate was added to each well. Luminescence was measured using a BioTek SynergyTM 2 microplate reader.

Cell Growth Assays
Cells (100 µl/well, most times 1:40 seeding in growth medium combined with drug R1881, RBN2397, BBQ, FICZ, Ruxo, GSK, IFNaA were grown at 37°C in 96-well plates in eight biological replicates, with medium changes every 2 days, till the no drug control cells reached relatively high cell density. The cells were washed once with PBS, incubated at 37°C for 2 hours with 100 µl of phenol-red free DMEM (for VCaP) or RPMI (for others) + 2% FBS + 0.25 mg/ml MTT, followed with medium removal, extraction with 100 µl of DMSO, and measurements on Synergy HT plate reader (BioTek). The numbers were subtracted of the background DMSO alone reading and quantified with Prism (GraphPad). Error bars represent standard deviation. P value **** <0.0001; *** <0.001; ** <0.01; * <0.05; ns, not significant.

Cell Cycle Analyses
Cells were drug-treated for 48 hrs (slower growing VCaP and CWR22Rv1) or 24 hrs (faster growing PC3-AR and PC3), and then dispersed into single cells after trypsin treatments and pipetting. Cells were collected after centrifugation at 2,000 RPM at 4°C for 5 minutes. The cell pellets were PBS washed once, resuspended in 0.5 ml of cold PBS, mixed to 4.5 ml of cold 70% ethanol, and put on ice for over 2 hours. The cells were re-harvested by centrifugation and washed with PBS. Cells (~3.5 x 10 5 ) were resuspended in 0.1 ml of staining solution (PBS + 0.1% Triton X-100 + 80 µg/ml of Propidium Iodide + 0.2 mg/ml of RNase A), incubated at 37°C for 40 minutes, diluted with 0.4 ml of PBS + 0.1% Triton, measured on Attune NxT Cytometer (Thermo Fisher), and quantified with FCS Express 7 (De Novo Software).

Cell Morphology Analyses and Dead Cell Measurements
Cells were drug-treated for 6 days (slower growing VCaP and CWR22Rv1) or 3 days (faster growing PC3-AR and PC3). Pictures were taken under the EVOS Cell Imaging System (Thermo Fisher). Cells were also dispersed into single cells after trypsin treatments and pipetting. Cells were collected after centrifugation at 2,000 RPM for 3 minutes. Cells were re-suspended in 1:1 of PBS and Trypan Blue Stain 0.4% (Invitrogen T10282), and counted on Countess II FL (Life Technologies).
RNA-seq data analysis. All of the data used in this study except GSVA analysis was obtained from recount3. Primary tumor data for prostate (PRAD) and lung (LUAD) used for GSVA analysis was generated by the TCGA