Reciprocal inhibition of PIN1 and APC/CCDH1 controls timely G1/S transition and creates therapeutic vulnerability

Cyclin-dependent kinases (CDKs) mediated phosphorylation inactivates the anaphase-promoting complex (APC/CCDH1), an E3 ubiquitin ligase that contains the co-activator CDH1, to promote G1/S transition. PIN1 is a phosphorylation-directed proline isomerase and a master cancer signaling regulator. However, little are known about APC/CCDH1 regulation after phosphorylation and about PIN1 ubiquitin ligases. Here we uncover a domain-oriented reciprocal inhibition that controls the timely G1/S transition: The non-phosphorylated APC/CCDH1 E3 ligase targets PIN1 for degradation in G1 phase, restraining G1/S transition; APC/CCDH1 itself, after phosphorylation by CDKs, is inactivated by PIN1-catalyzed isomerization, promoting G1/S transition. In cancer, PIN1 overexpression and APC/CCDH1 inactivation reinforce each other to promote uncontrolled proliferation and tumorigenesis. Importantly, combined PIN1- and CDK4/6-inhibition reactivates APC/CCDH1 resulting in PIN1 degradation and an insurmountable G1 arrest that translates into synergistic anti-tumor activity against triple-negative breast cancer in vivo. Reciprocal inhibition of PIN1 and APC/CCDH1 is a novel mechanism to control timely G1/S transition that can be harnessed for synergistic anti-cancer therapy.

ligase that is activated by CDH1 (encoded in humans by FZR1), whose activity is regulated by CDK2 and/or CDK4/6 mediated phosphorylation [7][8][9] . Cyclin D1/CDK4/6 inactivates APC/C CDH1 either directly 8 or indirectly via phosphorylating RB, thereby triggering E2F-dependent upregulation of Cyclins E/CDK2 and EMI1 9, 10 , to promote cell cycle re-entry. A functional collaboration between APC/C CDH1 and RB restrains cell cycle entry, as forced pRB-E2F expression alone is insufficient to drive cell cycle entry, and an additional loss of APC/C CDH1 E3 ligase activity is required to trigger proliferation 7 . Moreover, the inactivation of APC/C CDH1 , but not activation of pRB-E2F, represents the commitment point of no return for cell-cycle entry 7 . However, whether APC/C CDH1 activity is further regulated after CDK phosphorylation is not known.
Targeting CDK proteins to block cell proliferation has been validated as an effective anticancer therapy 11 . CDK4/6 inhibitors have been approved to treat estrogen receptor-positive (ER+) breast cancer (BC) 12 , but only have limited efficacy in triple-negative breast cancer (TNBC).
TNBC, especially RB-deficient, is the most aggressive and difficult-to-treat subtype of BC with few targeted therapeutic options 13 , underscoring an urgent need for developing novel therapies.
Pro-directed phosphorylation is further regulated by PIN1-catalyzed cis-trans prolyl isomerization, which modulates protein functions, including protein stability, interaction, and activity 14,15 . PIN1 is overexpressed and correlates with poor outcomes in most human cancers 16, 17 . PIN1 activates numerous oncogenic signaling pathways to drive cancer malignancy and drug resistance [16][17][18][19][20] . As a result, PIN1 knockout, which has no overt phenotype for half of the lifespan in mice 21 , prevents tumorigenesis induced by oncogenes or tumor suppressors [22][23][24] . Furthermore, PIN1-reducing genetic polymorphisms are associated with reduced risk for cancer in humans 25 . PIN1 was originally identified as a cell cycle regulator and is essential for mitosis in yeast 26 , and many cell-cycle proteins have been identified as PIN1 binding partners 27, 28 . PIN1 modulates RB by stabilizing phosphorylated RB to promote cell cycle progression 29 . PIN1 also regulates cyclin D1 function at transcriptional and posttranslational stabilization 22,30 . As such, pharmacologic ablation of PIN1, including using the approved drugs, offers a unique and promising approach to eradicate aggressive cancer 17-19, 31-35 . Notably, most PIN1 inhibitors identified so far not only inhibit PIN1's catalytic activity but also induce PIN1 degradation 31-33, 35 .
However, it is still unknown whether and how PIN1 protein stability is physiologically regulated and how PIN1 inhibitors induce PIN1 degradation. Thus, the special prognostic significance of PIN1 protein levels in tumors and the consistent observation of PIN1 degradation upon pharmacologic inhibition prompted us to investigate a PIN1 ubiquitin ligase. Here we show that APC/C CDH1 , a cell-cycle inhibitor, and PIN1, a cell-cycle promoter, directly interact and negatively regulate each other in a domain-oriented mechanism, which can be harnessed for cancer therapy.

Cell cycle regulator APC/C CDH1 is a physiological E3 ubiquitin ligase for PIN1
PIN1 is a well-established oncoprotein 17 , whose protein, but not mRNA, levels were strikingly correlated with poor prognosis in human BC independent of grade or proliferative indices by analyzing the dataset from the study by Tang et al 36 (Fig. 1a, Extended Data Fig. 1a, b, Supplementary Data 1). Previous studies showed that most PIN1 inhibitors, including the newly developed highly selective Sulfopin 31, 37 , covalent PIN1 inhibitor  , and the approved drug combination of ATRA and ATO (AApin) 31, 33 , not only inhibit PIN1's catalytic activity but also induce PIN1 degradation. As shown 32, 33 , we found that PIN1 inhibitors-induced PIN1 degradation was rescued by proteasome inhibitors (Extended Data Fig. 1c, d), indicating that PIN1 is degraded via the ubiquitin-proteasome pathway. These data suggest that post-translational regulation of PIN1 at the protein level may offer therapeutic opportunities against cancer. We thus investigated the molecular mechanisms controlling PIN1 protein stability.
To identify the specific E3 ubiquitin ligase for PIN1, we used immunoprecipitation coupled with mass spectrometry (IP-MS) and identified potential PIN1-interacting E3 ligases. Based on functional gene similarity, PIN1-interacting E3 ligases identified were categorized into four main groups, with the APC/C E3 ligase complex being the most enriched and validated one in different purification methods (both GST-PIN1 and Flag-PIN1 pull-down) (Fig. 1b, Extended Data Fig.   1e, f, Supplementary Table 1). As APC/C activators, CDH1 (encoded by the FZR1 gene) and CDC20 regulate the activity and substrate specificity of the E3 ligase complex 38 . We found that PIN1 had a much higher affinity for interacting and co-localizing with CDH1 than its close homologue CDC20, as evidenced by immunoprecipitation and immunofluorescence ( Fig. 1 c,  in mouse embryonic fibroblasts (MEFs) (Fig. 1e, Extended Data Fig. 1h-t). These results show that CDH1 specifically interacts with PIN1 and likely affects its protein stability.
From the late M phase throughout G1, CDKs inactivation maintains CDH1 in a dephosphorylated state; hence, APC/C CDH1 is active, preventing premature entry into S phase 39 .
Examining the dynamics of PIN1 protein levels during the cell cycle of two non-transformed cells, we found that PIN1 expression was relatively low in G1 and started to accumulate at the onset of the S phase, coincident with the inactivation of APC/C CDH1 . Moreover, compared to CDH1 wildtype (WT), CDH1 KO in these cells stabilized the protein levels, but not mRNA levels, of mitotic cyclins as well as PIN1 across the cell cycle, revealing a negative correlation between PIN1 protein levels and APC/C CDH1 activity ( Fig. 1f, g, Extended Data Fig. 2a-c).
To further assess the relationship between PIN1 and APC/C CDH1 activity in individual cells at physiological conditions, we used non-transformed MCF-10A stably expressing the Fucci reporter system with mCherry-conjugated to a Geminin fragment (aa1-110) containing the APC/C CDH1 degron motif (RXXL) 7 . Notably, the promoter region of the reporter construct is unregulated, and the reporter degradation is primarily regulated by APC/C CDH1 (Extended Data Fig. 2d). In this experimental model, an increase in the reporter signal directly reflects a decrease in APC/C CDH1 E3 ligase activity and vice versa, allowing for real-time tracking of APC/C CDH1 activity at the single cell level 7 . When the cells were synchronized in G1 followed by releasing back into the cell cycle, PIN1 levels were strongly correlated with APC-degron reporter levels across the cell cycle, confirming a negative correlation between PIN1 and APC/C CDH1 (Fig. 1h). Moreover, in the CPTAC human breast cancer dataset 40, 41 , we also found that CDH1 protein levels were negatively correlated with PIN1 protein levels. Low levels of CDH1 were associated with poor prognosis of BC tumors (Fig. 1i, Extended Data Fig. 2e). Thus, CDH1 is likely the prime candidate responsible for PIN1 degradation.
APC/C CDH1 E3 ligase activity is inhibited mainly in cancer cells, likely due to hyperphosphorylation of CDH1 induced by increased CDK kinase activity, resulting in decreased binding to the APC complex 8,42,43 . We found that manipulating CDH1 expression affected PIN1 abundance only under serum-free conditions, when CDK activity and CDH1 phosphorylation are relatively low (Fig. 2a). We mutated potential CDK phosphorylation sites that are also potential sites for prolyl isomerization, all of which are located at the N-terminus of CDH1 flanking the Cbox (Fig. 2b). Ectopic expression of the phospho-deficient CDH1-7A mutants that can bind the APC core and are constitutively active 43 , induced a senescence-like state, indicative of an irreversible G0 state in several BC cell lines (Fig. 2c, d), and reduced protein levels of PIN1 and other known APC/C CDH1 substrates including PLK1, CDC20, Cyclin B1 and Geminin, which was rescued by the proteasome inhibitor ( Fig. 2e-f). Notably, the promoter of the Flag-PIN1 construct is unregulated. Thus Flag-PIN1 degradation is primarily regulated by APC/C CDH1 . As expected, CDH1-7A mutants dramatically shortened the half-lives of its substrates, including PIN1, with limited impact on their mRNA levels ( Fig. 2g-i). Furthermore, mutations of the PIN1 active site enhanced the interaction between CDH1-7A and PIN1 and also promoted CDH1-7A-mediated PIN1 degradation and ubiquitination ( Fig. 2j-l). Thus, APC/C CDH1 is likely the physiological E3 ubiquitin ligase for PIN1, and its activity is primarily inhibited by phosphorylation of CDH1 in cancer cells.

PIN1 regulates APC/C CDH1 E3 ligase activity at post-translational levels to ensure the timely G1/S transition
Given the negative relationship between PIN1 protein levels and APC/C CDH1 , to further explore their causal relationship, we knocked out PIN1 to examine whether PIN1 may reciprocally inhibit APC/C CDH1 activity. PIN1 KO dramatically reduced cell viability in long-term clonogenic assays in both RB-proficient and RB-deficient BC cell lines (Fig. 3a, Extended Data Fig. 3a Fig. 3b, c). In contrast, the proteomic analysis of PIN1 KO in the MDA-MB-231 cell line showed that PIN1 KO had noticeable effects on cell cycle progression 33 (Fig.   3b). Thus, PIN1 may affect cell cycle progression at transcriptional and post-translational levels.
Notably, PIN1 KO resulted in a prolonged G1 phase, as shown 21 , which was accompanied by de-stabilization of APC/C CDH1 substrates across the cell cycle ( Fig. 3c- we generated PIN1 KO in RB-proficient and RB-deficient cell lines. We found that PIN1 KO markedly reduced the half-lives of APC/C CDH1 substrates independent of RB and transcriptional factors (Fig. 3h), supporting that PIN1 directly regulates APC/C CDH1 E3 ligase activity at posttranscriptional levels as well to ensure the G1/S transition.

Domain-oriented reciprocal inhibition of PIN1 and APC/C CDH1 E3 ligase
We then investigated the underlying regulation mechanism between PIN1 and APC/C CDH1 . In the APC/C CDH1 complex, CDH1 can be directly phosphorylated and inactivated by Cyclin A2/CDK2, Cyclin D1/CDK4/6, and ERK, resulting in its dissociation from the APC core complex 7,8 . We found that CDK2 and CDK4, but not CDK6, strongly interacted with CDH1 ( Fig. 4a, Extended Data Fig. 4a). Moreover, CDK4 specifically phosphorylated CDH1 at S163 and significantly increased PIN1 and APC/C CDH1 substrate abundance in serum-free conditions (Fig. 4b Fig. 4c, d). Thus, these results together support CDH1 as an alternative CDK4 substrate.
PIN1 specifically recognizes pSer/Thr-Pro motifs and catalyzes sequence-specific phosphorylation-dependent proline isomerization 28, 48 . We found that the S->A mutation of CDH1-S163, abolishing phosphorylation at this location, reduced its interaction with PIN1 ( Fig. 4c). To gain further insight into the interaction between CDH1 and PIN1 in more mechanistic details, we performed in vitro GST pull-down assays of either WT or phosphorylation-deficient CDH1 with full-length GST-PIN1 or its isolated WW or PPIase domain. As expected, CDH1-WT preferentially bound to the PIN1 WW-domain, whereas, surprisingly, the phosphorylationdeficient CDH1-7A preferentially bound to the PIN1 PPIase domain (Fig. 4d), suggesting PIN1 domain-specific interaction modes with CDH1.
To directly visualize PIN1 binding and isomerization of phosphorylated CDH1, we synthesized a CDH1-pS163 peptide and mapped the interaction with PIN1 using nuclear magnetic resonance (NMR). The perturbation data indicated that CDH1-pS163 peptide binds to the WW domain with moderately strong affinity (Fig. 4e). PIN1 residue R17 showed the most substantial perturbation, and along with residues S18, Y23, W34 and E35 formed a continuous patch that interacts with phosphoserine, pS163, and the adjoining proline, P164. Our experiment-guided model suggests that the phosphate group from pS163 has a charge: charge interaction with R17, while P164 stacks in the pocket formed by Y23 and W34 (Fig. 4f, g), which was confirmed by GST pull-down assays using PIN1 point mutations (Extended Data Fig. 4e). Interestingly, weak perturbation was observed in the PPIase domain active site, which may mediate PIN1-catalyzed isomerization of CDH1-pS163 peptide. To confirm such isomerization, we used specific 13C enrichment of the P164 and a 2D-13C HSQC spectrum to directly measure the P164 isomerization states. Our results showed that 7% cis-P164 isomer was present in the free peptide, but PIN1-catalyzed isomerization doubled this population to 14.2%, indicative of the trans to cis isomerization. This might lead to an increase in phosphorylated CDH1 because CDH1-specific phosphatase does not engage with cis proline 49 . Indeed, our mutational analysis showed that the S->A phospho-deficient mutation of CDH1-S163 abolished the binding to PIN1 WW domain and became unstable, whereas the S ->E phosphomimic mutation of CDH1-S163 restored the binding to PIN1 and became more stable (Fig.   4h, Extended Data Fig. 4f-h). Thus, PIN1 binds to and catalyzes the trans to cis prolylisomerization of the pS163-P motif in CDH1, thereby stabilizing phosphorylated CDH1 and rendering APC/C CDH1 inactive (Fig. 4i).
Distinct from CDH1-WT, phosphorylation-deficient CDH1-7A mutant preferentially bound to the PIN1 PPIase domain, which may mediate PIN1 degradation. To explore this possibility, we examined whether PIN1 contains a destruction box (D-box) since most APC/C CDH1 substrates include a D-box with the conserved consensus RXXL sequence (X presents any amino acid) 50 .
Indeed, PIN1 has a putative D-box within its PPIase domain (Extended Data Fig. 4i), recognizable by the CDH1 WD40 domain. To confirm that, we performed co-IPs using a point mutation W34A in the PIN1 WW domain and a D-box mutation, RLAA, in the PIN1 PPIase domain and a dual mutation. Indeed, the W34A mutation that is unable to bind to a pSer/Thr-Pro motif 17 prevented the PIN1 interaction with CDH1-WT. In contrast, the RLAA mutation within the PPIase domain did not interfere with the interaction with CDH1-WT (Extended Data Fig. 4j).
In keeping with these findings, the RLAA mutation in PIN1, but not the W34 mutation, conferred resistance to CDH1-7A-mediated PIN1 degradation ( Fig. 4j (Fig. 4k). These modeling results further support that the D-box in the PIN1 PPIase domain is critical for CDH1 to interact with PIN1 and target PIN1 for degradation. Notably, K117 is one of the very few residues with significantly different conformation between free PIN1 and PIN1-Sulfopin complex (Extended Data Fig. 4n), which may enhance PIN1 interaction with CDH1 to promote PIN1 degradation.
Collectively, the above data show two distinct modes of the PIN1-CDH1 interaction in a domain-oriented manner. On the one hand, when phosphorylated in cells at the G1/S transition, CDH1 binds to the WW domain of PIN1, which catalyzes trans to cis isomerization of the pS163-P motif in CDH1 to prevent CDH1 dephosphorylation, thereby rendering APC/C CDH1 inactive to promote S-phase entry. On the other hand, when unphosphorylated in cells in the G1 phase, CDH1 is active and recognizes the D-box motif in the Pin1 PPIase domain, targeting Pin1 for degradation to prevent S-phase entry. Thus, CDH1 is either a downstream substrate of PIN1 or its upstream regulator, depending on the domain-oriented binding modes to control the timely G1/S transition.

Pharmacologic inhibition of PIN1 and CDK4 restores APC/C CDH1 E3 ligase activity inducing an insurmountable G1 arrest
The above results not only identify a novel reciprocal inhibitory mechanism of PIN1 and APC/C CDH1 to control the timing of the G1/S transition but also suggest a potential novel anticancer therapy targeting PIN1 and CDKs to reactivate APC/C CDH1 synergistically. We will use CDK4/6 inhibitors instead of CDK2 inhibitors to reactivate APC/C CDH1 in our following experiments, as CDK4/6 inhibitors are highly selective and approved by FDA. To test the possibility, we first examined the effects of PIN1 KO on the fate of phosphorylated CDH1. PIN1 KO dramatically reduced CDH1 phosphorylation and promoted the binding of CDH1 to APC complex, which was enhanced by Palbociclib in both RB-proficient and RB-deficient cells (Fig.   5a, Extended Data Fig. 5a). Similarly, PIN1 inhibition, i.e., prevention of trans to cisisomerization, led to dephosphorylation of CDH1 (Extended Data Fig. 5b), presumably by the trans-selective CDH1 phosphatase 49 , to initiate the domain-oriented binding mode, in which CDH1 reduced its binding to the PIN1 WW domain, but increased its binding to the PIN1 PPIase domain (Fig. 5b, c, Extended Data Fig. 5c). This change of the binding mode might switch PIN1 from being an upstream regulator to a downstream substrate of CDH1. To support this possibility, we first assessed the impact of PIN1-and CDK4-inhibition on APC/C CDH1 activity. Indeed, overexpression of PIN1 potently inhibited APC/C CDH1 E3 ligase activity, as determined by the prolonged half-lives and elevated levels of APC CDH1 substrates in the absence or presence of the CDK4 inhibitor (Extended Data Fig. 5d-f). By contrast, PIN1 inhibitors led to significant decreases in PIN1 and Geminin along with other APC/C CDH1 substrates in a time-dependent manner., corresponding to an increase of G1 cells (Fig. 5d, Extended Data Fig. 5g-j). To maximize the effects of PIN1 inhibitors on PIN1 degradation, i.e., dramatic PIN1 degradation, we applied a 3-day treatment for in vitro experiments.
To further confirm these results, we measured APC/C CDH1 activation kinetics in single cells upon the CDK4 or PIN1 inhibitors treatment by analyzing changes in the degron reporter levels.
CDK4 inhibition caused activation of APC/C CDH1, as evidenced by starkly decreased reporter intensity and cell cycle arrest. Only a few cells, presumably those that had already passed the G1/S checkpoint when Palbociclib was administered, proceeded to complete another round of cell division after adding the CDK4 inhibitor, and none had multiple divisions (Fig. 5e, middle panel).
Hence, CDK4 inhibition likely causes an immediate reactivation of APC/C CDH1 reflected by decreased reporter intensity along with G1 arrest rather than cell death. Consistently, proteasome inhibition blocked Palbociclib-induced degradation of APC/C CDH1 substrates (Extended Data Fig.  5k). Of note, both PIN1 inhibitors (Sulfopin and AApin) led to prolonged S/G2 phases and cell death in some cells, but eventually a reactivation of APC/C CDH1 to induce G1 arrest (Fig. 5e,  inhibition, PIN1 inhibition has a broader activity to arrest cells both in G1 and during mitosis. Significantly, the combination of the CDK4 inhibitor and the PIN1 inhibitor-induced striking activation of APC/C CDH1 and G1 arrest (Extended Data Fig. 5o).
It has been reported that Cyclin D/CDK4 either inactivates APC/C CDH1 directly 8 or indirectly by phosphorylating RB and triggering the onset of E2F-dependent expression of Cyclins E and EMI1, followed by Cyclin E/CDK2-and EMI1-mediated inactivation of APC/C CDH1 to commit the G1/S transition 7, 9 (Extended Data Fig. 6a). Therefore, we examined whether PIN1 inhibitor and CDK4 inhibitor-induced activation of APC/C CDH1 is dependent on Cyclin D, RB, CDK2 or EMI1.
Knockout or knockdown of each of these genes did not block PIN1 inhibitor-and CDK4 inhibitor-  Fig. 6i, j). We found that WT-to-CDH1-KO protein ratio, but not WT-to-CDH1-KO mRNA ratio significantly decreased upon PIN1-and CDK4/6-inhibitors treatment (Fig. 5f). Thus, these data strongly supported that PIN1 and CDK4 inhibitors reactivate APC/C CDH1 through RBindependent, CDH1-dependent mechanisms, and simultaneous loss of RB and CDH1 was sufficient to overcome the synergistic effects induced by PIN1 and CDK4 inhibitors (Fig. 5g). To further separate transcriptional regulation from post-translational one, we measured protein stability of APC/C CDH1 substrates by the cycloheximide chase assay. CDH1 KO significantly prolonged the protein half-lives of PIN1, and APC/C CDH1 substrates with or without PIN1 and CDK4 inhibitors treatment (Fig. 5h, Extended Data Fig. 6k). Combining the PIN1 inhibitor with the CDK4 inhibitor caused even more pronounced PIN1 poly-ubiquitination, which was diminished by CDH1 KO (Extended Data 6l).
These above results not only uncover that APC/C CDH1 E3 ligase activity is inhibited by PIN1catalyzed trans-to-cis isomerization of CDH1 after CDKs-mediated phosphorylation but also suggest that, in addition to conventional cell-cycle regulation, PIN1 inhibition blocks trans to cis prolyl isomerization of phosphorylated CDH1 and cooperates with CDKs inhibitors to reduce CDH1 phosphorylation, leading to re-activation of APC/C CDH1 . APC/C CDH1 activation, in turn targets PIN1 for degradation via its D-box, thereby creating a negative feedback loop between PIN1 and APC/C CDH1 and inducing an insurmountable G1 arrest (Fig. 5i, j). It was reported that Abemaciclib inhibits kinases other than CDK4/6 including CDK2 57 , which is also a CDH1 upstream kinase. Therefore, Abemaciclib has the therapy advantage over Palbociclib. As for safety and tolerability, the two-drug regimen showed no bone marrow suppression and was well tolerated with maintenance of body weight (Extended Data Fig.   8d-m). In the RB-deficient mouse model, the tumors didn't significantly benefit from Palbociclib, but the combination of Sulfopin and Palbociclib elicited a complete inhibition of tumor growth ( Fig. 6j, k). Thus, PIN1 inhibitors synergize with CDK4 inhibitors against TNBC in human cells and mouse xenograft models.

PIN1 and CDK4/6 inhibitors achieve synergistic anti-tumor activity against aggressive TNBC in immune-competent mouse models
Next, we thought to confirm their synergistic anti-tumor activity against TNBC in immunecompetent mouse models, which is more clinically relevant to human patients. To this end, we generated two cohorts of immune-competent genetically engineered TNBC mouse models: K14cre; p53wt/f; Brca1wt/f_BT1 and K14cre; p53wt/f; Brca1wt/f_BT3, which were Rb-deficient and Rbproficient respectively and resembled aggressive human TNBC with the growth of highly proliferative and poorly differentiated mammary carcinomas in syngeneic immune-competent Fig. 8n, o, Supplementary Table 2).
To ascertain the suitability of our models, we first confirmed the interactions of Cdh1 and Pin1 or Cdk4 in these mouse tumors (Extended Data Fig. 8p). We transplanted the transgenic tumors orthotopically in nude mice or FVB mice to generate allogeneic tumor mouse model (nude mice) and syngeneic tumor mouse model (FVB mice), respectively, Notably, syngeneic BC models are historically much more challenging to treat with chemotherapy or targeted agents than PDOX and allograft models in immune-compromised hosts 59 . Indeed, CDK4 inhibitors or PIN1 inhibitors had limited efficacy in this hard-to-treat, highly undifferentiated, and proliferative murine TNBC.
However, their combination was highly effective and well-tolerated and significantly delayed tumor progression and increased overall survival compared to either monotherapy ( Fig. 6l-q).
Collectively, our results indicate that the combination of PIN1 and CDK4/6 inhibitors achieves synergistic anti-tumor activity against RB-proficient or RB-deficient TNBC in immune-compromised or immune-competent mouse models, with an excellent safety profile, making it a strong candidate for clinical development.

Discussion
Epithelial cells execute an active program to maintain interphase that relies on reducing the levels of continuously accumulating pro-mitogenic proteins, including cyclins, through ubiquitinmediated degradation 1, 60 . A key ubiquitin ligase for the maintenance of interphase is APC/C CDH1 , which targets a range of pro-mitogenic proteins for degradation 7,8,61 . In cancer, APC/C CDH1 has been identified as a tumor suppressor 62,63 . Inactivation of CDH1 is achieved through multi-site Pro-directed phosphorylation 43 , but whether CDH1 is subject to further regulation after phosphorylation is unknown. On the other hand, through isomerization of Pro-directed phosphorylation, PIN1 promotes tumorigenesis by acting as a master cancer signaling regulator activating over 70 oncoproteins and inactivating over 30 tumor suppressors 16, 17 . However, the role and regulation of Pin1 in cell cycle progression remain elusive, even though PIN1 was originally identified as a mitotic regulator 26 .
Here we report for the first time that the two opposing enzymes, APC CDH1 , a tumor suppressor that promotes maintenance of interphase and can be phosphorylated by CDK4, and PIN1, an oncoprotein that promotes mitotic progression, directly interact with and negatively regulate each other. More importantly, this reciprocal inhibitory mechanism between APC/C CDH1 and PIN1 creates a therapeutic vulnerability that can be harnessed to greatly enhance the efficacy of CDK4/6 inhibitors, especially in RB-deficient tumors.
In its active, dephosphorylated form, CDH1 binds the D-box of PIN1, which is buried within its PPIase domain and may be exposed by PIN1 inhibitors, and promotes PIN1 ubiquitinationdependent proteolysis. However, when CDK activity garners momentum at the end of G1, CDH1 is phosphorylated specifically at S163, which is the preferred high-affinity binding motif for the leading to an insurmountable G1 arrest, which translates into synergistic anti-tumor activity against triple-negative breast cancer both in immune-compromised and -competent and/or RB-deficient or -proficient mouse models. Thus, the mechanism we uncovered is important for cancer cell proliferation when PIN1 is overexpressed and CDKs activity is high.
In summary, our work uncovers a novel reciprocal inhibitory mechanism of PIN1 and APC/C CDH1 to regulate the G1/S checkpoint, whose aberration causes APC/C CDH1 inhibition and PIN1 overactivation in a vicious feedback loop, leading to unchecked cell cycle proliferation and cancer. Moreover, we further develop a new therapeutic strategy using clinically available PIN1 inhibitors and CDK4 inhibitors to break this vicious cycle synergistically to induce anti-cancer activity against not only RB-proficient but also RB-deficient TNBC, paving the way for new clinical trials to evaluate their clinical impact on patients with TNBC.

Acknowledgments
We        Fig. 1k. e, Frequency of G1 arrest (ratio of G1 arrested cells to total cells).
The error bar indicates 95% confidence interval determined by bootstrapping. Data in graphs are mean ± s.d., analyzed by unpaired two-sided t-test. ***P < 0.001.  The samples were measured by data-independent acquisition (DIA) mass spectrometry method as described previously [71][72][73] . The Orbitrap Eclipse Tribrid mass spectrometer (Thermo Scientific) instrument coupled to a nanoelectrospray ion source (NanoFlex, Thermo Scientific) and EASY-nLC 1200 systems (Thermo Scientific, San Jose, CA). A 120-min gradient was used for the data acquisition at the flow rate at 300 nL/min with the temperature controlled at 60 °C using a column oven (PRSO-V1, Sonation GmbH, Biberach, Germany). All the DIA-MS methods consisted of one MS1 scan and 33 MS2 scans of variable isolated windows with 1 m/z overlapping between windows. The MS1 scan range is 350 -1650 m/z and the MS1 resolution is 120,000 at m/z 200.
The MS1 full scan AGC target value was set to be 500 % and the maximum injection time was 100 ms. The MS2 resolution was set to 30,000 at m/z 200 with the MS2 scan range 200 -1800 m/z and the normalized HCD collision energy was 28%. The MS2 AGC was set to be 4000 % and the maximum injection time was 50 ms. The default peptide charge state was set to 2. Both MS1 and MS2 spectra were recorded in profile mode. DIA-MS data analysis was performed using Spectronaut v16 [74][75][76] with directDIA algorithm by searching against the Uniprot 77 downloaded human fasta file. The oxidation at methionine was set as variable modification, whereas carbamidomethylation at cysteine was set as fixed modification. Both peptide and protein FDR cutoffs (Qvalue) were controlled below 1% and the resulting quantitative data matrix were exported from Spectronaut. All the other settings in Spectronaut were kept as Default.

STED imaging
Except where indicated otherwise the steps were performed at room temperature. Cells were rinsed with PBS twice and fixed with 2% PFA for 15 min. Fixative was removed by washing with PBS 3 times. Cells were then permeabilized with 0.1% Triton for 10 min. After removing Triton, cells were blocked with 5% BSA for 1 hour and then incubated with anti-PIN1 (Abcam) and anti-CDH1 (Santa Cruz) antibodies overnight at 4°C. After three washes with PBS, the cells were then incubated with Alexa Fluor ® 514 Goat Anti-Mouse (Invitrogen) and Alexa Fluor ® 568 Goat Anti-Rabbit (Abcam) antibodies for 1 hour. Following the incubation, the cells were washed 3X with PBS and mounted for STED imaging. Colocalization rates were calculated using the LAS X software (Leica).

Real-time PCR
Total RNAs were extracted using the QIAGEN RNeasy mini kit. cDNA synthesis was performed using Maxima Universal First Strand cDNA Synthesis Kit from Thermo Scientific. qPCR reactions were performed with FastStart Universal SYBR Green Master (Rox) from Roche. The experiments were performed according to the manufacturer's instructions. The sequences of the primers used for qRT-PCR analyses were provided in Supplementary Table 3.

RNA sequencing and data analysis
Total RNAs were extracted from the BC cell lines WT and PIN1 KO MDA-MB-231, MCF-7 and MDA-MB-468 respectively. RNA-sequencing samples were prepared as previously described 78 .
Gene set enrichment analysis (GSEA) was performed using GSEA software (Broad). Normalized counts of PIN1 KO versus WT cells were used for GSEA analysis against the biological process related gene sets. Normalized enrichment scores (NES) were used to generate bar graphs for visualization of the functional transcriptional outputs of the three cell lines. Cells were then plated 24 hours before starting the microscope acquisition. PIN1 inhibitors or CDK4/6 inhibitor were added in the medium and imaged using a Nikon Eclipse TE-2000 inverted microscope with a 10X Plan Apo objective and a Hammamatsu Orca ER camera, equipped with environmental chamber controlling temperature, atmosphere (5% CO2) and humidity. Images were acquired every 30 min using the MetaMorph Software. For each condition filmed, 4 different fields were selected.

Drug combination test and synergy calculations
p53Cinema single cell analysis package was used for semiautomatic tracking of individual cells in live cell imaging datasets as described previously 80 . Tracking data were then used to quantify intensity of fluorescent reporters from background subtracted images by averaging 10 pixels within the cell nucleus. Cells were tracked using only information about a constitutively expressed nuclear marker, such as H2B-Turquoise, and were thus blind to the dynamics of molecular players of interest, such as mCherry-Geminin. Only cells that remained within the field of view throughout the entire duration of the experiment were considered for downstream analyses.
We defined the frequency of G1 arrest as those cells that arrested in G1 phase for at least 20 hours after drugs were added. S/G2 durations were calculated by the time that cells spent in S/G2 phase after drugs were added.

Cell synchronization and cell cycle profiling
Cells synchronized by double thymidine block or nocodazole block as described previously 81 were collected at the indicated time points and suspended in cell cycle kit (Beckman Coulter) according to the manufacturer's instructions. Stained cells were sorted with CytoFLEX LX1 Flow Cytometer.
The results were analyzed by FSC Express software.

Annexin V-FITC-PI double staining
For detection of apoptosis, cells treated with indicated inhibitors were co-stained with Annexin V-FITC and PI (Dead Cell Apoptosis Kit, Invitrogen) according to the manufacturer's instructions.
Stained cells were sorted with CytoFLEX LX1 Flow Cytometer.

Immunoblot and immunoprecipitation analyses
In vivo Ubiquitination Assay was performed as described previously 83 . 293T cells were transfected with His-ubiquitin and the indicated constructs. Thirty-six hours after transfection, cells were treated with 2 μM MG132 for 12 hours and lysed in buffer A (6 M guanidine-HCl, 0.1 M Na 2 HPO 4 /NaH 2 PO 4 , and 10 mM imidazole pH 8.0). After sonication, the lysates were incubated with Ni-NTA beads (QIAGEN) for 3 h at 4 °C. Subsequently, the His pull-down products were washed twice with buffer A, twice with buffer A/TI (1 volume buffer A and 3 volumes buffer TI), and once with buffer TI (25 mM Tris-HCl and 20 mM imidazole pH 6.8). The pull-down proteins were resolved by SDS-PAGE for IB.

In vitro kinase assay
In vitro kinase assay was performed as previously described 84 . Briefly, HA-tagged CDH1 WT and mutants were transfected into HEK293T cells, followed by being immunoprecipitated with monoclonal Anti-HA-Agarose antibody (Sigma, A2095). The purified HA-CDH1 proteins were then incubated with 500 uM of AT P gS (Abcam, ab138911) and 0.5 ug of recombinant human cyclin D1+CDK4 proteins (Abcam, ab55695) in the kinase reaction buffer (50mM Tris-HCl, 10mM MgCl 2 , 0.1mM EDTA, 2mM DTT, 0.01% Brij 35, pH 7.5) for 30 min at room temperature. Then adding 2 mM of PNBM (Abcam, ab138910) and allowing the alkylating reaction proceed for additional 2h at room temperature. The reaction was then terminated by adding 5x SDS loading buffer and boiled for 10 min. Samples were then subjected to IB using anti-Thiophosphate ester antibody (Abcam, ab92570).

In vivo therapy for patient-derived xenografts
All animal experiments were approved by the IACUC of the Beth Israel Deaconess Medical Center.
Triple-negative BC patient-derived xenograft (TM00096) was purchased from Jackson Laboratories. Pieces from PDOX tumors were subcutaneously implanted into the mammary fat pads of 6-week-old BALB/c female nude mice. Tumor sizes were measured every three days by caliper after implantation and tumor volume was calculated by the modified ellipsoidal formula: tumor volume = ½ length × width 2 . Treatments were started once the tumors reached 3-5 mm in diameter and continued until tumors reached 15 mm in any direction. Mice were randomly assigned to six groups with comparable average tumor size. Sulfopin treatment was given by intraperitoneal injection with a dosage of 40 mg/kg (dissolved solution: 5% DMSO in D5W, 7 days/week), Palbociclib treatment was given by oral gavage with a dosage of 100 mg/kg (dissolved solution: saline, 5 days/week), Abemaciclib treatment was given by oral gavage with a dosage of 100 mg/kg (dissolved solution: 0.5% CMC-Na, 5 days/week), or drug combinations in which each compound was administered at the same dose and scheduled as a single agent. The investigators were not blinded to allocation during experiments and outcome assessment.

In vivo therapy for immunocompetent TNBC mouse models
All animal experiments were approved by the IACUC of the Beth Israel Deaconess Medical Center.
Maximum permitted the longest dimension of tumors was 20 mm. Pieces from breast tumors generated in K14cre; p53wt/f; Brca1wt/f female mice were transplanted into the mammary pads of 6-week-old FVB female mice. For survival studies, treatments were started once the tumors reached 3-5 mm in diameter and continued until mice were symptomatic or tumors reached 20 mm in any direction, at which point mice were euthanized. For time point analysis, mice were sacrificed two weeks post-treatment initiation. Sulfopin treatment was given by intraperitoneal injection with a dosage of 60 mg/kg (7 days/week), Abemaciclib treatment was given by oral gavage with a dosage of 100 mg/kg (7 days/week), or drug combinations in which each compound was administered at the same dose and scheduled as a single agent. Tumor sizes were measured every three days by caliper after implantation and tumor volume was calculated by the modified ellipsoidal formula: tumor volume = ½ length × width 2 . The investigators were not blinded to allocation during experiments and outcome assessment.

Experiment-guided Model
The chemical shift perturbation was interpreted as ambiguous iterative restrains used for docking a random conformation of the phosphopeptide on PIN1 (PDB: 1PIN) 51 using HADDOCK2.2 webserver 85 . The restrains were derived by marking two strongly perturbed PIN1 residues, R17 and W34 as active residues and three moderately perturbed residues S18, Y23 and E35 as passive residues. The peptide was assumed to be fully flexible with the phosphoserine, pS163, and its adjacent proline, P164, being the active residues that interact with PIN1. In subsequent runs, the model was refined using ambiguous distance restraints based on the interpretation of previously solved crystal structures of similar phosphopeptides bound to the WW domain of PIN1 86 .

Proline Isomerization Study
Commercially synthesized specific 13 C, 15 N-P164 labeled CDH1 phosphopeptide was used to facilitate direct quantitative determination of the cis and trans proline populations. The strong 13 C-HSQC peaks originating from Pro164 can be easily distinguished from the weak peaks due to ~1% natural abundance 13 C present in the rest of the peptide. Two isolated sets of peaks were observed for P164. Based on the interpretation of the chemical shifts, the major peaks were assigned as trans isomer and the minor peaks were assigned as cis isomer 87 . The proline resonance assignments were further confirmed using a 2D-13 C-HSQCTOCSY experiment while no attempts were made to stereospecifically assign proton resonances, thus the assignment of HB2 and HB3, HG2 and HG3, and HD2 and HD3 are interchangeable. 58 µM free peptide and its complex with a 4-fold molar excess of PIN1, dissolved in the above-mentioned NMR buffer were used to estimate the cis and trans isomer populations at 25 o C.