The key feature of the current model for Pin1-mediated PKC regulation is the Pin1-catalyzed cis-trans isomerization of the turn motif (TM) (Figure 1A). TM is a conserved feature among many AGC kinases that presents a phosphorylatable Ser or Thr residue in the C-terminal domains of these enzymes (Pearce et al., 2010). Phosphorylation of the TM is part of the PKC maturation process, and it is essential for enzyme stability (Newton, 2010; Gao and Newton, 2002). In conventional PKC isoforms (α, βI/II, γ), the TM is followed by a proline residue and this pThr-Pro motif is proposed to be the isomerizable Pin1 target (Zhou et al., 1999; Figure 1A). Therefore, the first step toward understanding how Pin1 regulates PKC was to determine the interaction mode between Pin1 and TM. To that end, NMR-detected binding experiments were conducted using uniformly ¹⁵N-enriched ([U-¹⁵N]) WW domain and full-length Pin1, and two synthetic peptides (pTMα and pTMβII) that correspond to the phosphorylated TMs of the α and βII PKC isoforms, respectively (Figure 1B). Addition of increasing concentrations of pTMα and pTMβII resulted in drastic shifts in the 2D ¹⁵N-¹H heteronuclear single-quantum coherence (HSQC) spectra of [U-¹⁵N] Pin1 and WW (Figure 1—figure supplements 1–4). The residue-specific chemical shift perturbation (CSP) values were calculated for pairs of spectra corresponding to the maximum concentrations of pTM in each titration series vs. apo Pin1. The four CSP plots of Figure 1C clearly demonstrate that the pTM binding site resides on the WW domain, and that the interaction mode is not significantly influenced by the presence of the PPIase domain in full-length Pin1. The fast exchange regime of the pTM-protein interactions enabled us to construct binding curves for all resolved residues with CSPs larger than the mean (see ”Construction and analysis of the chemical shift-based binding curves” in the Methods section for details on the selection criteria and fitting procedures). The curves fit well globally with a single-site binding model (Figure 1D, Supplementary file 2). The reported global K_d values fall into the micromolar range typical for WW-substrate affinities (Verdecia et al., 2000; Born et al., 2019; Wilson et al., 2013). The similarities of the global K_d values between WW and full-length Pin1 (Figure 1D) further support the conclusion that the PPIase domain does not significantly affect the pTM-WW interactions, and that the TMs of the α and βII PKC isoforms bind preferentially to the Pin1 WW domain.

The lack of high-affinity interactions of the PPIase domain with substrate does not preclude a catalytic mode of action for Pin1 as demonstrated previously for several Pin1 substrates (Wang et al., 2015a). We therefore tested the ability of Pin1 to isomerize the TM. These experiments monitored activity against the isolated TM and in the context of a larger C-terminal region that harbors a second phosphorylated motif found in most AGC kinases – the hydrophobic motif (HM). HMs in PKC isoforms are described by the sequence FXXF(S/T)(F/Y) where the underlined Ser/Thr is constitutively phosphorylated as part of the kinase maturation process (Newton, 2010; Edwards and Newton, 1997). HM is located downstream of the TM with 14 amino acid residues separating the two motifs. Cell biological studies suggest that the HM is involved in PKC interactions with Pin1 (Abrahamsen et al., 2012). However, the functional role of the HM is unclear as it does not fit the definition of a canonical Pin1 substrate due to the absence of a Pro residue that follows pSer/Thr. In these and all subsequent experiments, we used two synthetic double-phosphorylated peptides (pV5α and pV5βII; Figure 1B) to model the C-term PKC regions that contain both phosphorylated HM and TM.

Pin1 catalytic activity is readily measured using the ¹H-¹H exchange spectroscopy (EXSY) of substrates in the presence of catalytic amounts of the enzyme (Smet et al., 2005; Jinasena et al., 2019; Mercedes-Camacho et al., 2013). EXSY experiments are applicable to characterization of chemical exchange processes with the exchange rates between 0.1 s^–1 and 100 s^–1 (Palmer et al., 2001; Kawale and Burmann, 2023). Pin1 brings the slow cis-trans isomerization process of pSer/Thr-Pro bonds into biologically relevant timescales by enhancing the isomerization rate ~10³- to 10⁴-fold. This rate enhancement gives rise to characteristic cross-peaks in the 2D ¹H-¹H EXSY NMR spectra. To our surprise, addition of catalytic amounts of Pin1 to either isolated pTMβII or to the entire phosphorylated pV5βII region failed to generate cross-peaks between the amide ¹H resonances of pThr641 in the trans(t) and cis(c) conformations of the pThr641(–1)-Pro642(0) bond in the pThr641(–1)-Pro642(0)-Pro643(+1) segment (Figure 2A). We intentionally used a long mixing time (0.5 s) to enable the detection of slow processes. Yet, no evidence of significant rate enhancement was observed, suggesting that even upon addition of Pin1 the exchange rate between the isomers remains slower than 0.1 s^–1.

The lack of appreciable enhancement of the isomerization rate was similarly obtained for the pThr638(–1)-Pro639(0)-Pro640(+1) segment of the pTMα and pV5α regions (Figure 2B). Of note, we were able to detect Pin1-catalyzed cis-trans isomerization of the peptidyl-prolyl bond between Gln634 and Pro635 in pTMα. This reaction was manifested by the appearance of cross-peaks between the amide ¹H_N of Gln634 in the cis and trans conformations (Figure 2B). Although Gln634-Pro635 is not a canonical Pin1 substrate, the catalytic action of Pin1 brings the reaction rate into a detectable range with k_EX of ~1 s^–1, where k_EX is the sum of the forward and reverse kinetic rate constants of the isomerization reaction, k_tc and k_ct. The Gln634-Pro635 data serve as direct evidence that weak interactions of the TM region with the PPIase domain are sufficient for Pin1 to appreciably catalyze the cis-trans isomerization of the non-canonical Gln-Pro substrate but not the pThr-Pro segment of the TM. For both PKC α and βII isoforms, the HM has no detectable effect on Pin1 catalytic activity against these substrates.

We hypothesized that the proline at position +1 rendered the pTMα and pTMβII motifs poor substrates for PPIase-catalyzed isomerization. Thus, we generated the P640A variant of pTMα by replacing Pro at +1 position with Ala. Pin1-catalyzed isomerization of the pThr638-Pro639 bond in the P640A variant is readily detectable as reported by the appearance of distinct cross-peaks between the cis and trans conformers (Figure 2C). The k_EX value of 29.6 s^–1 obtained from the time dependence of the cross-peak intensities (see “¹H-¹H exchange spectroscopy” in the Methods section) is typical for the Pin1 substrates. Similarly, replacement of Pro at the (+1) position with Thr in the short LpTPPD peptide common to the TM regions of α, βII, and γ PKC isoforms also resulted in significant enhancement of the cis-trans isomerization rate (k_EX value of 46.2 s^–1; Figure 2D). We conclude that Pro at the (+1) position of TMs is incompatible with these motifs serving as efficient substrates for isomerization by Pin1.

Our collective data further suggest that, contrary to prevailing models, Pin1 does not control downregulation of PKCα and βII isoforms by catalyzing isomerization of the PKC tail. Rather, Pin1 does so via a non-catalytic mechanism. The novelty of our findings prompted us to more thoroughly investigate the biophysical and structural basis of Pin1 interactions with the C-terminal domains of α and βII PKC isoforms. We focused on three aspects: (i) identifying the role of the HM, (ii) defining the binding mode, and (iii) establishing the effect of the C-terminal domain phosphorylation state on interactions with Pin1. To those ends, we conducted a total of 19 NMR-detected binding experiments between Pin1 and the relevant regions of the C-term domains of PKC α and βII isoforms with different phosphorylation states (Supplementary file 1). All information regarding the notations, sequences, and affinities is given in Supplementary file 2. The Pin1 domains are identified with the subscript ‘iso’ and ‘Pin1’ in their isolated forms and full-length Pin1 contexts, respectively.

The HM (PKC α/βII) sequence FEGFpSF/Y does not fit the canonical definition of a Pin1 substrate, and its role in the PKC-Pin1 interactions is unknown. Addition of the phosphorylated HM regions, pHMα and pHMβII, to [U-¹⁵N] full-length Pin1 resulted in large CSPs of the amide N-H_N cross-peaks for many Pin1 residues, thereby providing direct and site-specific evidence for the Pin1-HM interactions (Figure 3A and Figure 3—figure supplement 1A and B). We were surprised to find that residues in both Pin1 domains were significantly affected, suggesting the existence of more than one pHM binding site in Pin1 (Figure 3A and Figure 3—figure supplement 2A). To identify which Pin1 domain harbors the HM binding site(s), we titrated isolated [U-¹⁵N] WW and PPIase domains with pHMβII and α (Figure 3B and Figure 3—figure supplement 2B). The CSP patterns of full-length Pin1 and the sum of isolated domains are essentially identical. These data indicate that Pin1 contains two sites that bind the phosphorylated HM: one residing in the WW domain and the other in the PPIase domain (Figure 3A and B and Figure 3—figure supplement 2).

The fast-exchange regime of pHM binding to Pin1 (Figure 3C) enabled us to construct the binding curves for all responsive residues and fit them globally using the two-site binding model (“Construction and analysis of the chemical shift-based binding curves” in the Methods section). The fitting produced domain-specific K_d values within full-length Pin1 (Figure 3D). The pHMα motif has comparable affinities to both domains (725 μM to WW_Pin1 and 701 μM to PPIase_Pin1, Supplementary file 2), whereas the pHMβII motif has a 3.5-fold greater affinity for PPIase_Pin1 than it does for WW_Pin1 (K_d values of 75 μM and 257 μM, respectively; Supplementary file 2). To determine if the presence of the upstream unphosphorylated TM influences pHM-Pin1 interactions, we tested the binding of the V5α-TM-pHM and V5βII-TM-pHM regions to full-length Pin1. The NMR spectra and the measured K_d values are identical to those for the pHM regions only (Figure 3E). Thus, unphosphorylated TM has little effect on the HM interactions with Pin1. The collective data demonstrate that, despite not being a canonical substrate, the phosphorylated HM of PKC isoenzymes can interact with Pin1 via two independent binding sites that reside on the WW and PPIase domains, respectively.

Both HM and TM are phosphorylated in mature PKC. Addition of the pV5βII and pV5α regions that harbor both phosphorylated motifs to full-length Pin1 produced large CSP values in both Pin1 domains (Figure 4A, Figure 4—figure supplements 1 and 2A). The CSP plot of full-length Pin1 complexed to pV5βII(α) matches the sum of the CSP plots obtained for the isolated WW domain complexed to pTMβII(α), and the isolated PPIase domain complexed to pHMβII(α) (Figure 4A and B; Figure 4—figure supplements 1 and 2). These data provide unambiguous evidence that full-length Pin1 engages in bivalent interactions with the PKC C-terminal tail where WW binds the TM and PPIase binds the HM. The directionality of these interactions is imposed by the binding preferences of the TM. As the TM primarily interacts with the WW domain (Figure 1C), the only site available for HM binding resides on the PPIase domain. Neither motif is an isomerizable Pin1 substrate.

Since the Pin1-pV5βII interactions fall into the intermediate-to-fast binding regime, NMR lineshape analysis was applied to obtain the K_d value of 1.5 μM (“Determination of binding affinities using lineshape analysis” in the Methods section, Supplementary file 2). These interactions are only moderately affected by the C113S mutation that reduces Pin1 catalytic activity ~40-fold (Behrsin et al., 2007). The catalytically deficient C113S Pin1 variant shows a clear bivalent interaction mode with pV5βII (Figure 4—figure supplements 3 and 4). Compared to the wild-type (WT) Pin1, C113S has only ~2-fold weaker affinity to pV5βII (K_d value of 3.4 μM, Supplementary file 2). The low micromolar K_d values obtained for WT Pin1 and the C113S variant reflect the thermodynamic advantage that the bivalent interaction mode imparts on the interactions with the C-term tail.

We illustrate the thermodynamic effect of bivalency using the binding curves of two representative residues (Ser115 and Phe125) in three distinct system compositions (Figure 4C). The affinity of monovalent PPIase-pHMβII interactions does not appreciably depend on whether pTMβII is pre-bound to the WW domain: the corresponding K_d values are 133 μM and 117 μM, respectively (Figure 4C, Supplementary file 2). Moreover, the K_d values for the monovalent pHMβII-WW and pTMβII-PPIase interactions show little dependence on whether the isolated Pin1 domains or full-length Pin1 were used in the binding experiments (Figure 4D). However, when both motifs are presented to full-length Pin1 on the same polypeptide chain, we observe an ~90-fold enhancement in affinity (K_d = 1.5 μM). These general findings hold for the Pin1-pV5α interactions (Figure 4—figure supplement 5). In aggregate, these data provide a thermodynamic view of bivalent interactions between Pin1 and the C-term PKC tail, and attribute the respective ~10(3)-fold and ~90(60)-fold enhancements of pTMβII(α) and pHMβII(α) binding to Pin1 to bivalency.

The action of phosphatases on the HM and TM generates monophosphorylated and dephosphorylated PKC species that are detected in Pin1 pull-down assays (Abrahamsen et al., 2012). To establish how these C-term modifications affect the Pin1 binding mode, we examined the interactions of Pin1 with the α and βII C-term regions where either TM (V5-pTM-HM), HM (V5-TM-pHM), or neither (V5) were phosphorylated. Since clear spectroscopic signatures are available from our data on the PKC motif interactions with full-length Pin1 and its domains (Figures 1C, 3A, B, 4A ,B), they can be used to identify the interaction modes of different V5 species. We illustrate this point using two monophosphorylated regions of the βII isoforms bound to full-length Pin1 (Figure 5A–D).

The V5-pTM-HM binding to the full-length Pin1 produced a CSP pattern in the WW domain that is identical to that of the isolated pTM (compare Figures 5A and 1B) and to that of the pV5βII region (Figure 5B, WW_Pin1 correlation plot). Since pTM binding to WW_Pin1 does not cause any CSPs in PPIase_Pin1 (Figure 1C), the CSPs in the PPIase_Pin1 upon V5-pTM-HM binding could only be caused by the interactions with the unphosphorylated HM. The PPIase_Pin1 chemical shift correlation plot of Figure 5B reflects the influence of the HM phosphate group on the backbone chemical shifts of the isomerase domain. In conclusion, phosphorylated TM imposes a unidirectional bivalent binding mode by employing its high-affinity interaction with WW_Pin1 to direct the unphosphorylated HM to the binding site on PPIase_Pin1.

The C-term region monophosphorylated at HM (V5-TM-pHM) produced a CSP pattern (Figure 5C) that is essentially identical to that of the isolated pHM binding to full-length Pin1. This is illustrated by the chemical shift correlation plots constructed for WW_Pin1 and PPIase_Pin1 (Figure 5D). Since the unphosphorylated TM does not appreciably interact with WW_Pin1, it is unable to impose the unidirectional bivalent binding mode, and the phosphorylated HM is free to occupy both the WW- and PPIase-localized binding sites. We conclude that for the V5-TM-pHM region, the mode of interaction with Pin1 is monovalent and lacks directionality.

Using these types of analysis, we identified the binding modes of the V5 regions of α and βII isoforms (Figure 5E–H). The chemical shift data (Figure 5A and C; Figure 5—figure supplement 1) are mapped onto the extended conformation of Pin1 and the binding modes are illustrated with cartoon representations. Of note, the unphosphorylated C-term region interacts with Pin1 weakly. The CSPs are extremely small for the α isoform but are sufficiently large for βII to arrive at a K_d value of ~960 μM (Supplementary file 2). Based on our data, we conclude that the phosphorylation status of TM determines the valency of the Pin1-PKC C-term interaction mode.

The K_d values determined for all C-term regions enabled us to estimate the thermodynamic gain associated with the phosphorylation of TM and HM in the context of bivalent interactions (Figure 5I and Supplementary file 2). The affinity enhancement due to the TM phosphate is 62-fold, which corresponds to ΔΔG°_pTM of –2.4 kcal/mol. The affinity enhancement due to the HM phosphate is 14-fold, which corresponds to ΔΔG°_pHM of –1.6 kcal/mol. Both ΔΔG° values are within the range reported for the formation of the phosphate-mediated salt bridges in proteins (Errington and Doig, 2005). Summing up the contributions from the two phosphate groups produces the overall ΔΔG° of –4.0 kcal/mol, or ~900-fold enhancement of the Pin1 affinity to the phosphorylated C-term of PKCβII.

The phosphate groups occupy the canonical phosphate binding sites of the Pin1 domains. We reach this conclusion by comparing the ¹⁵N-¹H HSQC spectra of Arg sidechains in the apo and pV5βII-complexed Pin1 (Figure 5J). Based on the structural data obtained previously for monovalent substrates (Verdecia et al., 2000; Hu et al., 2017), Arg17_WW and the Arg68-Arg69_PPIase motif of the PPIase catalytic loop form salt bridges with the phosphate groups. While these three residues are exchange-broadened in apo Pin1 due to dynamics, their cross-peaks reappear upon pV5βII binding – consistent with their direct interactions with these phosphate groups (Figure 5J). To describe the bivalent interaction mode and identify the role of individual PKC and Pin1 residues, we determined the high-resolution structure of the complex.

While both α and βII isoforms show similar bivalency patterns in their Pin1 interaction modes (Figure 5A–D), the 8-fold higher binding affinity of pV5βII to Pin1 informed the choice of the Pin1-pV5βII complex for structural work. Since exhaustive screening of crystallization conditions failed to yield crystals suitable for X-ray diffraction, we pursued a solution NMR-based approach. The NMR structural ensemble of the Pin1-pV5βII complex was calculated with CYANA using ¹H-¹H NOEs (Figure 5—figure supplement 2A and B), hydrogen bond, and torsional angle restraints, and refined in explicit solvent using XPLOR-NIH (“NMR structure calculation and refinement” in the Methods section). The Pin1-pV5βII interface is defined by 75 inter-molecular NOEs whose assignment required the use of specifically labeled amino acids incorporated into the peptide substrate; a total of nine different Pin1 complexes were prepared to ensure sufficient data redundancy (Supplementary file 3). The overall ensemble has a backbone and all-heavy-atom RMSDs of 0.9 Å and 1.2 Å for the ordered regions, respectively (Supplementary file 4 and Figure 5—figure supplement 2C). Of note, the linker region connecting the two domains retains its flexibility in the complex and confers some variability onto the relative position of the WW and PPIase domains, as is evident from the ensemble representation of Figure 5—figure supplement 2C.

The complex exhibits novel structural features that distinguish it from all other structures of Pin1 complexes known to date. These features are highlighted in Figure 6 using the lowest-energy structure of the ensemble. First, the Pin1 substrate-binding mode is unusual in that pV5βII traverses the entire protein and interacts with both domains in a bivalent arrangement. The terminal pV5βII anchoring points are the two phosphorylated motifs, the pSer660 of the HM that binds to the PPIase domain and the pThr641 of the TM that binds to the WW domain (Figure 6A). In addition, the pV5βII region in between the two phosphorylated motifs forms an extensive network of interactions with both WW and PPIase domains (Figure 6—figure supplements 1 and 2). The second noteworthy aspect of the structure is the extensive conformational change that Pin1 undergoes upon pV5βII binding. The WW and PPIase domains are brought into proximity primarily via interactions with the N-terminal pV5βII region. This compact conformation of pV5βII-complexed Pin1 is distinct from the ‘closed’ conformation observed in available crystal structures of monovalent Pin1 complexes (Figure 6B). The major differences are: (i) the ~70° rotation of the WW domain relative to the PPIase module; and (ii) the repositioning of the α4 helix and the α4-β2 loop of the PPIase domain to accommodate pV5βII. The linker region connecting the WW and PPIase domains retains its flexibility and is not involved in pV5βII interactions.

To facilitate structural analysis of the Pin1-pV5βII interface, we separated pV5βII into four segments: the ‘pTM anchor’ (639–643, S1), the ‘turn’ (644–650, S2), the ‘groove’ (651–655, S3), and the ‘pHM anchor’ (656–661, S4) (Figure 6A, Video 1). The five-residue ‘pTM anchor’ harboring the pThr641-Pro642-Pro643 segment is positioned at the interface between the WW and PPIase domains (Figure 6C). Its backbone runs almost parallel to the α1 helix of the PPIase domain. The key interactions are the salt bridge between the Thr641 phosphate and guanidinium groups of Arg17_WW, and the hydrophobic contacts of the pThr641 methyl group and Pro642 pyrrolidine ring with the sidechains of Trp34_WW, Asn90_PPIase, Ile93_PPIase, and Gln94_PPIase. In addition, the NHε group of the Trp34_WW forms a hydrogen bond with the carbonyl oxygen of pThr641. These interactions are present in 65–100% of the Pin1-pV5βII ensemble structures.

The Pro643-Asp644 segment reorients the backbone such that it runs along the β1 strand of the WW domain. This marks the beginning of the seven-residue ‘turn’ segment (Figure 6D) that completes the inter-domain arm of pV5βII and is responsible for the realignment of the PPIase-WW domain interface. Prior structural work on Pin1 identified residues 137–141_PPIase, 148–149_PPIase, and 28–32_WW as being involved in the dynamic inter-domain interface (Born et al., 2019; Wilson et al., 2013). In the Pin1-pV5βII complex, many residues from this subset are now engaged in interactions with ‘turn’ segment pV5βII (Figure 6D and Figure 6—figure supplement 1). The turn itself is stabilized by intra-pV5βII hydrogen bonds and salt bridges. Arg649 plays a particularly prominent role in that regard as it participates in both types of interactions. The intra-molecular salt bridge formed by the Arg649 and Glu646 sidechains (present in 75% of the ensemble structures) is unique to the βII isoform (Figure 6—figure supplement 3) because these Arg and Glu residues are replaced by hydrophobic Leu and Ala/Met residues in other conventional PKC isoforms (Figure 6D and Figure 1A). The contributions of Arg sidechain-mediated interactions are likely responsible for the higher affinity of pV5βII for Pin1 relative to the affinity of pV5α.

The turn is followed by the five-residue pV5βII ‘groove’ segment that is rich in polar residues and interacts exclusively with the PPIase domain. The groove segment threads through a deep groove in the Pin1 PPIase domain formed by helix α4 and the β-sheet comprised of strands β3, β4, and β1 (Figure 6E and Figure 6—figure supplement 4). This configuration drives the repositioning of helix α4 and the α4-β2 loop relative to the structures of Pin1 complexed to the monovalent substrate (Figure 6B). The Pin1-pV5βII interface involves hydrophobic contacts, present in all 20 ensemble structures, of the only hydrophobic residue of this pV5βII segment (Ile651) with Pin1 residues Leu141_PPIase, Ile159_PPIase, and the Arg161_PPIase methylenes. Among polar residues, pV5βII residues Ser654 and Glu655 are within H-bonding distance with the three Pin1 residues of the conserved parvulin tetrad (Wang et al., 2015b; Mueller et al., 2011; Barman and Hamelberg, 2014): His59_PPIase and His157_PPIase, and Thr152_PPIase, respectively (Figure 6—figure supplement 2).

The last segment of pV5βII (the ‘pHM anchor’) emerges from the PPIase groove and occupies the catalytic site. This ‘anchor’ segment contains the entire PKCβII HM whose key residues (Phe656, Phe659, and pSer660) interact with the amphiphilic environment of the Pin1 catalytic site (Figure 6F). Specifically, the sidechains of Phe656 and Phe659 are accommodated by the hydrophobic environment formed by Met130_PPIase, Phe125_PPIase, and Leu122_PPIase. The Phe659 aromatic ring can potentially engage in stacking interactions with His59_PPIase. The pSer660 phosphate is anchored to the Pin1 catalytic loop via salt bridges with a triad of positively charged residues (Arg68_PPIase, Arg69_PPIase, and Lys63_PPIase). Those interactions, present in all 20 ensemble structures, rigidify the loop as is evident from the reappearance of Arg68_PPIase and Arg69 _PPIase resonances in the NMR spectra upon complex formation (Figure 5J). Phe661, the third Phe of the HM, is not involved in any persistent interactions with the Pin1 PPIase module.

We then compared the binding poses of pHM with that of the D-peptide – a potent unnatural peptide inhibitor of Pin1 that binds specifically to the catalytic site of the PPIase domain (Zhang et al., 2007; Figure 6—figure supplement 5A and B). Structural overlay of the complexes shows that pHM and the D-peptide occupy the same PPIase region, and the positions of the ligand phosphate groups coincide (Figure 6—figure supplement 5C). Notable differences include the arrangement of the ligand hydrophobic ring moieties in the catalytic site, and the position of the Gln131_PPIase sidechain relative to the ligand. Specifically, the space occupied by the Gln131_PPIase sidechain in the Pin1::pV5βII complex is occupied by the Gln5 sidechain of the D-peptide in the Pin1::D-peptide complex (Figure 6—figure supplement 5C). Gln131 is the C-terminal residue of the α4 helix that undergoes the most significant rearrangement upon the formation of the bivalent Pin1::pV5βII complex (Figure 6B).

The structural data revealed: (i) a bivalent mode of Pin1 interactions with the C-term tail of PKCβII, (ii) that the phosphate groups occupy the canonical binding sites of the Pin1 WW and PPIase domains, and (iii) that the intervening residues can form an extensive network of interactions with the residues of both Pin1 domains. Moreover, the similarities of the NMR CSP patterns between pV5βII and pV5α binding to Pin1 report that the bivalent interaction mode is shared by both the α and βII PKC isoforms. To interrogate our biophysical and structural conclusions regarding Pin1 function in a cellular context, we developed a system for assessing Pin1-mediated PKCα regulation without the contribution of endogenous Pin1 activity in a HEK293T cell model. HEK293T cells were chosen because these cells are efficiently transfected and therefore broadly used in such experiments. PKCα was chosen for these analyses because we were able to detect endogenous PKCα by immunoblotting in HEK293T cells, whereas PKCβII overexpression was required for detection in this cell line.

Previous studies reported Pin1 downregulates PKC levels in serum-starved cells stimulated with PDBu (Abrahamsen et al., 2012). Consistent with this general concept, efficient (~80%) siRNA-mediated knockdown of Pin1 expression resulted in an ~45% elevation in steady-state PKCα levels in cells cultured in the presence of serum. That is, cells stimulated by natural agonists under more physiological conditions than those stimulated by exogenous agonist PDBu under serum starvation (Figure 7—figure supplement 1A). This relationship was further corroborated by CRISPR screens that produced Pin1 null HEK293T cell lines (“Isolation of a clonally-derived Pin1 null cell line” in the Methods section; Figure 7—figure supplement 1B). Clonally derived cells recovered from the CRISPR challenge exhibited a wide range of Pin1 expression levels, and immunoblotting confirmed an inverse correlation between endogenous Pin1 and steady-state PKCα levels (Figure 7A; R=–0.92, p<0.0005). The linearity of this relationship suggests Pin1 is a limiting component in downregulation of PKCα levels.

The CRISPR approach produced three HEK293T clones with little or no detectable Pin1 antigen, and those lines exhibited the highest steady-state levels of PKCα. Clone 3 represented a particularly attractive candidate for a Pin1 null cell line as it was devoid of detectable Pin1 antigen. This was confirmed by DNA sequencing. HEK293T cells carry four copies of the Pin1 gene (Lin et al., 2014), and DNA sequence analyses indicated clone 3 harbored three frameshift alleles. All three alleles altered the targeted exon 2 that encodes the Pin1 PPIase domain and included: (i) a 5 bp deletion that interrupts the Pin1 amino acid sequence after Thr79, (ii) an 8 bp insertion that interrupted the Pin1 sequence after Arg80, and (iii) a 10 bp deletion that disrupts the Pin1 amino acid sequence after Lys77 (Figure 7—figure supplement 1C). The resulting translation products were prematurely terminated after addition of another 22, 10, and 7 residues, respectively. Although PKCα steady-state levels were upregulated approximately 2-fold in clone 3 relative to parental WT HEK293T cells under both serum-free and serum-replete conditions (Figure 7B; 2.3 ± 0.2- and 2.3 ± 0.3-fold, respectively), the Pin1 null cells were not phenotypically perturbed. Their dimensions (diameter, surface area), viabilities, and proliferation rates were indistinguishable from those of the parental cells (Figure 7—figure supplement 1D–G). Thus, the elevation in steady-state PKCα levels identified a baseline Pin1 loss-of-function phenotype.

To assess the effects of Pin1 mutants on steady-state PKCα expression without the contribution of endogenous Pin1 activity, stable transgenic cell lines individually expressing epitope-tagged versions of WT Pin1; substrate-binding mutants: Pin1^W34A, Pin1^R68A,R69A, Pin1^{W34A,R68A,R69A}; or the catalytic-deficient Pin1^C113S were derived from clone 3 cells. As shown in Figure 7C (top left panel), WT Pin1 and all three of the Pin1 substrate-binding mutants were overexpressed some 4- to 10-fold relative to endogenous Pin1 levels in these stable transgenic lines. Assessment of steady-state PKCα levels in the reconstituted cell lines showed that WT Pin1 expression significantly reduced PKCα steady-state levels relative to the Pin1 null condition (Figure 7C, right panel). By contrast, reconstituted expression of neither Pin1^W34A, Pin1^R68A,R69A, nor of Pin1^{W34A,R68A,R69A} restored Pin1 function (Figure 7C, right panel). As expected, the differences in the triple mutant data in comparison to the Pin1 null mutant data were not statistically significant (p=0.07). However, even though the data did not reach a threshold of statistical significance, PKCα levels in the Pin1^{W34A,R68A,R69A}-expressing cells were consistently higher than those in Pin1 null cells. One speculation is the triple mutant imposes dominant negative effects on some limiting factor in PKCα degradation that are revealed in the Pin1 null background. More work is required to resolve this issue.

Interestingly, we were unable to produce stable cell lines that overexpressed the Pin1^C113S ‘catalytic-dead’ mutant to the same levels achieved for WT Pin1 and the three substrate-binding mutants. Reconstituted Pin1^C113S expression was consistently elevated only some 3-fold relative to endogenous Pin1 levels (Figure 7C, bottom left panel). Yet, this relatively modest level of Pin1^C113S expression was as effective as WT Pin1 expression in reducing steady-state PKCα levels in otherwise Pin1 null HEK293T cells (Figure 7C, right panel). These collective results support a physiologically relevant mechanism for Pin1-mediated regulation of PKCα that requires a bivalent interaction mode of Pin1 with PKCα, but operates independently of the Pin1 prolyl-isomerase catalytic activity.

Previous experimental data suggested that Pin1-binding motifs with Pro at position +1 are disfavored substrates because of their failure to bind the catalytic PPIase domain of Pin1 (Innes et al., 2013). Our results provide direct experimental support for that conclusion, as both pTMα and pTMβII motifs bind with high affinities to the Pin1 WW domain but lack high-affinity interactions with the PPIase domain. In addition, EXSY experiments clearly show that Pin1 is unable to catalyze the cis-trans isomerization of the pThr-Pro motifs of PKCα and PKCβII. The unsuitability of these motifs as Pin1 substrates is due solely to the presence of Pro at the +1 position as evidenced by our demonstrations that replacement of this Pro residue restores their activities as substrates for isomerization by Pin1. Of note, the inability of Pin1 to catalyze cis-trans isomerization was previously observed for the pT²³¹PP site in the Tau protein (Smet et al., 2005; Eichner et al., 2016). Interestingly, this observation prompted the speculation that Tau is not a genuine target for regulation by Pin1 – a speculation that our data indicate requires reconsideration (see below).

Why is Pro at position +1 incompatible with Pin1-catalyzed cis-trans isomerization of what is otherwise a signature substrate motif? Computational studies predict unfavorable overall binding free energies for Pin1 engagement with the transition state and the cis-conformations of the pTPP substrates (Momin et al., 2018). A definitive answer as to why Pro(+1)-containing motifs are incompatible with Pin1 catalysis remains elusive as there is currently no consensus on the precise catalytic mechanism of Pin1. The most recent ‘twisted-amide’ catalysis model (Mercedes-Camacho et al., 2013) envisions a transition state where distortion of the substrate is stabilized by the H-bond between the C=O of the substrate Pro at position (0) and the backbone -NH of Pin1 residue Gln131. As formation and stabilization of the transition state relies on a dynamic network of H-bonds involving conserved residues in the Pin1 active site (Mueller et al., 2011; Barman and Hamelberg, 2014; Xu et al., 2014), Pro at position +1 might impose geometric constraints on the substrate where formation or stabilization of the amide twist is disfavored.

The demonstration that Pin1 S/TPP binding motifs are not isomerizable has interesting implications. A survey of AGC kinases reveals that these motifs are surprisingly common in this important class of enzymes. Moreover, as is the case for the PKC kinases, S/TPP motifs are present in the TM regions for both Akt and PKN kinases (Figure 8A). These arrangements suggest that Pin1 binding to non-isomerizable substrates through its WW domain defines an unappreciated, yet broadly deployed, regulatory strategy that operates independently of Pin1 cis-trans isomerase activity. This point is discussed in further detail below.

Our NMR structure and NMR-detected binding experiments demonstrate that Pin1 interacts with the C-terminal tails of PKCα and PKCβII in a bivalent mode. To our knowledge, this is the first structure of the full-length Pin1 complexed to the substrate that engages both Pin1 domains. The Pin1 WW domain binds the TM, a canonical Pin1 pThr-Pro recognition motif, whereas the Pin1 PPIase domain engages the HM, a non-canonical substrate that lacks Pro after the pSer residue. The specificity of TM binding exhibited by the WW domain imposes a unidirectionality to the Pin1-PKC interaction by directing the HM to the PPIase domain. This bivalent binding mode results in Pin1 adopting a mostly compact conformation, with the ‘pTM anchor’ positioned at the interface between two Pin1 domains. Based on our structure, the linker length separating two phosphorylated Ser/Thr in bivalent Pin1 substrates is projected to be an important factor. In the PKCβII V5 domain, the phosphorylated TM and HM sites are separated by 18 amino acids. This length allows the simultaneous binding of the two sites by Pin1 without causing conformational strain on Pin1 itself. It is of interest to consider these results in light of what has been observed in the interaction of the yeast Pin1 homolog Ess1 with the RNA polymerase II C-terminal tail (Namitz et al., 2021). The Ess1 linker that joins the WW and PPase domains is a rather rigid element, and CTD repeats exceeding 27 amino acid residues are required to separate the two phosphorylated Ess1-binding sites for bivalent interactions (Namitz et al., 2021). By contrast, the cognate Pin1 linker is unstructured, and its flexibility might afford Pin1 the potential to interact with a broader range of substrates. Moreover, the retention of linker flexibility in the Pin1::pV5βII complex suggests that the complexed Pin1 can potentially adopt ‘extended’ states that would not be readily detectable by the ensemble-averaged methods such as solution NMR.

Bivalent interactions afford significant advantages to biologically relevant protein-protein interactions as these are capable of increasing binding affinity and specificity, and inducing the appropriate conformational rearrangements in binding partners (Mammen et al., 1998; Erlendsson and Teilum, 2020). It is in this manner that the two substrate-binding domains of Pin1 govern its ability to engage in bivalent binding interactions that enhance the binding affinity and specificity to multi-phosphorylated client proteins (Daum et al., 2007; Rogals et al., 2016). Several Pin1 substrates that contain neighboring canonical pSer/Thr-Pro motifs in their unstructured regions have been characterized: IRAK1 kinase (Rogals et al., 2016), Cdc25c (Innes et al., 2013), Tau (Eichner et al., 2016), and STAT3 (Lufei et al., 2007). Of those, the IRAK1-derived peptide was the only reported case of bivalent binding to Pin1 based on the NMR CSP data analysis, and this interaction involved two isomerizable pSer/Thr-Pro motifs (Rogals et al., 2016). Our Pin1::pV5βII complex structure represents the first natural bivalent substrate-bound Pin1 structure. The two phosphate groups occupy the substrate-binding site of the WW and PPIase domains, and the intervening residues stabilize the domain interface and bring the two domains into proximity. The linker that connects the Pin1 WW and PPIase domains maintains its flexibility and does not participate in interactions with the PKCβII C-terminal tail. The structure of this complex provides a guide for interpreting the potential binding modes of other Pin1 substrates that contain multiple pSer/Thr-Pro and non-canonical Pin1 motifs. For example, cyclin E contains both canonical pThr³⁸⁰-Pro³⁸¹ and non-canonical pSer³⁸⁴-Gly³⁸⁵ motifs (Yeh et al., 2006). The pSer³⁸⁴-Gly³⁸⁵ motif serves as an important determinant of cyclin E interactions with Pin1 and its Pin1-mediated downregulation.

With regard to Pin1-mediated regulation of PKC, we note the linear relationship between Pin1 and PKCα steady-state levels up to the point of physiological levels of Pin1 expression – as documented in analyses of the clones derived from the CRISPR screen (Figure 7A). However, additional elevations in Pin1 expression did not result in further reductions in PKCα steady-state levels as evidenced by the Pin1 reconstitution experiments. Our interpretation of the data is that physiological expression of Pin1 in HEK293T cells is normally the limiting factor in the stimulated PKCα degradation pathway. When Pin1 is expressed in excess of physiological levels however, some other component of the PKCα degradation pathway becomes limiting in those contexts. It is posited that Pin1 isomerase activity promotes ubiquitination and subsequent degradation of conventional PKC enzymes (Abrahamsen et al., 2012). This proposal rests on the observation that challenge of stimulated cells with a small molecule inhibitor of Pin1 abolishes agonist-induced ubiquitination of PKC. Interpretation of those data is neither ironclad, nor are those results inconsistent with our conclusions. It remains to be established that the inhibitor does not compromise the Pin1-PKC interaction. As described above, our collective biochemical and biophysical data suggest Pin1 downregulates PKC via a mechanism independent of its prolyl cis-trans isomerization activity. Pin1 reconstitution experiments provide direct support for this concept as expression of the ‘catalytic-dead’ Pin1^C113S mutant rescues the Pin1 null condition in a HEK293T cell model.

How might Pin1 operate via a non-catalytic mechanism to downregulate PKC activity? Upon activation, the membrane-associated conformation of PKC is sensitive to dephosphorylation – first at the HM site by PHLPP and subsequently at the TM site and activation loop by PP2A (Gao et al., 2008; Baffi et al., 2019; Hansra et al., 1996; Baffi and Newton, 2022). One possibility is that Pin1 discharges a scaffolding or substrate chaperoning function where it aids (directly or indirectly) in the recruitment of protein phosphatases to the Pin1-PKC complex. Indeed, interactions between Pin1 and PP2A have been detected in pulmonary eosinophils (Shen et al., 2008). Such an activity, when coupled to a substrate chaperoning function, might ‘organize’ and stimulate ordered dephosphorylation of the PKC C-terminus with subsequent degradation of the kinase. A recruitment mechanism of this nature is attractive in that it provides a means for channeling activated PKC to specific protein phosphatase(s) in a temporally and spatially appropriate manner (Figure 8B). Alternatively, sequestration of the HM site by Pin1 might prevent the formation of the autoinhibitory state (Yang et al., 2018; Antal et al., 2015a) and thereby ‘trap’ PKC in an open ‘activated’ conformation. This activated conformer would be particularly susceptible to the action of phosphatases that trigger its ultimate degradation. The significance of phosphorylated HM in maintaining the stability of AGC kinases is well documented (Baffi et al., 2019; Baffi and Newton, 2022; Mora et al., 2004).

It is tempting to speculate that non-catalytic functions for peptidyl-prolyl isomerases are more broadly represented in this group of enzymes. Pin1 and Escherichia coli trigger factor (TF) are both peptidyl-prolyl cis-trans isomerases – although Pin1 is a member of the Parvulin family of peptidyl-prolyl isomerases whereas TF is a member of the FK506-binding protein class. The enzymatic activity of TF is dispensable for its function as a protein folding chaperone (Kramer et al., 2004). A TF^F198A mutant competent for client protein binding, but defective in peptidyl-prolyl cis-trans isomerization activity, remains active as a folding-promoting chaperone in vivo (Kramer et al., 2004). Regardless, our data indicate one can no longer confidently infer whether a protein is a target of Pin1 regulation solely based on whether or not the putative substrate presents isomerizable pSer/Thr-Pro motifs.

The oncogenic properties associated with dysregulation of Pin1 activities identify the isomerase as an attractive target for the development of new therapeutic approaches for cancer treatment (Zhou and Lu, 2016; Wu et al., 2022; Chuang et al., 2021). Unfortunately, although numerous small molecule Pin1 inhibitors have been identified (Moore and Potter, 2013; Yu et al., 2020), these have been plagued by serious off-target toxicities. Resolution of this issue presents a significant obstacle to productive development of the Pin1-targeted drug pipeline (Moore and Potter, 2013; Fila et al., 2008; Nam et al., 2001). In that regard, the screens used to identify Pin1 inhibitors typically rely on readouts of catalytic activity. The data we report herein suggest alternative strategies. That is, screens for ligands that target the bivalent interaction mode rather than the Pin1 catalytic activity. Such strategies hold the potential for enhancing inhibitor specificity and affinity and thereby reducing the toxicities associated with off-target effects. The structure of the Pin1::pV5βII complex now offers a precise template for guiding the design and development of bivalent inhibitors that simultaneously bind to both the Pin1 WW and PPIase domains.

A total of three Pin1 protein constructs were used in this study. The following genes with a codon-optimized DNA sequence were cloned into a pET-SUMO vector (Invitrogen): full-length Pin1 (residues 1–163), the WW domain (residues 1–50), and the PPIase domain (residues 50–163). Pin1 and isolated domains were heterologously expressed in E. coli BL21(DE3) and BL21(DE3)pLysS strains, respectively. For the natural abundance protein preparation, cells were grown in Luria broth (LB) to an OD₆₀₀ of 0.6 prior to induction of protein expression with 0.5 mM IPTG. The cells were grown for additional 4–5 hr at 37°C. For the expression of isotopically enriched proteins, we used an LB to M9 minimal media resuspension method (Marley et al., 2001). To generate uniformly ¹⁵N-enriched (U-[¹⁵N]) or ¹⁵N, ¹³C-enriched (U-[¹⁵N,¹³C]) protein samples, the M9 media contained either 1 g/l of ¹⁵NH₄Cl and 3 g/l of natural-abundance D-glucose, or 1 g/l of ¹⁵NH₄Cl and 3 g/l of ¹³C-D-glucose, respectively. The expression of isotopically enriched Pin1 in BL21(DE3) cells was induced for 15 hr at 15°C. The expression of isotopically enriched individual domains in BL21(DE3) pLysS strain was induced for 5 hr at 37°C.

The cells were harvested by centrifugation (4000 rpm, 30 min) at 4°C. Cell pellets were resuspended in a buffer containing 20 mM Tris-HCl (pH 7.5), 0.5 M NaCl, 5 mM imidazole, and 10 mM β-mercaptoethanol. The 6×His-tagged SUMO fusion proteins were purified using a HisTrap HP Ni²⁺ affinity column (GE Healthcare Life Sciences). The fractions containing fusion protein were pooled and exchanged into a SUMO protease cleavage buffer (20 mM Tris-HCl at pH 8.0, 0.15 M NaCl) using a HiPrep 26/10 column (GE Healthcare Life Sciences). The cleavage reaction was initiated by adding 6×His-tagged SUMO protease to a final concentration of ~8 μg/ml to the protein solution. After 30 min at room temperature, the reaction mixture was loaded onto to the second HisTrap HP Ni²⁺ affinity column to purify the desired protein from the 6×His-tagged SUMO and 6×His-tagged SUMO protease.

Purified proteins (Pin1, WW, or PPIase) were buffer-exchanged into an NMR buffer using either 5 kDa or 3 kDa MWCO centrifugal concentrators (Vivaspin, Sartorius). The NMR buffer contained 10 mM d₄-imidazole at pH 6.6, 100 mM KCl, 1 mM TCEP, 8% D₂O, and 0.02% NaN₃. The purity was assessed with SDS-PAGE conducted on samples with serial dilutions. Protein concentrations were determined by measuring the absorbance at 280 nm and using the following extinction coefficients: 20970 M^–1 cm^–1 (Pin1), 13980 M^–1cm^–1 (WW), and 6990 M^–1 cm^–1 (PPIase). The molecular masses of purified proteins were determined by MALDI-TOF mass spectrometry.

All 18 peptides used in this study were derived from the C-terminal V5 regions of either βII, α, or βI PKC isoenzymes (Homo sapiens) and synthesized commercially (Supplementary file 1). The modifications included acetylation and amidation at the N- and C-termini, respectively, to avoid the influence of positive and negative charges terminal charges on binding and catalysis. All natural abundance peptides were purchased as crude mixtures and subsequently HPLC-purified on a C18 column (Waters). The buffers used for purification were 5 mM NH₄HCO₃ in 100% water (buffer A) and 5 mM NH₄HCO₃ in 82% acetonitrile/18% H₂O (buffer B). A linear gradient of buffer B ranging from 0% to 20–40%, depending on the peptide, was applied during the elution step. The molecular weight and purity of the peptides were verified using ESI and MALDI-TOF mass spectrometry (Figure 1—figure supplement 5). Isotopically enriched peptides, #14 and #15 in Supplementary file 1, were purchased in already purified form and used as is. For the binding experiments (vide infra), stock solutions of peptides were prepared by dissolving the lyophilized powders in ddH₂O. The pH was adjusted to 6.6 either with NH₄OH or HCl. The peptide concentration was determined using the absorbance at 205 nm; for the phosphorylated peptides, phosphate assay was additionally used (Fogg and Wilkinson, 1958).

NMR-detected binding experiments were conducted by adding aliquots of concentrated peptide stock solutions to the [U-¹⁵N] protein (Pin1, WW, or PPIase) samples, and recording [¹⁵N,¹H] HSQC spectra for each peptide concentration point. The spectra were collected at 25°C on Bruker Avance III spectrometers operating at the ¹H Larmor frequencies of 600 MHz or 500 MHz. The concentration of protein samples was between 90 μM and 100 μM. A total of 19 binding experiments were conducted; the parameters are summarized in Supplementary file 2. The binding data were analyzed in three ways, as described in the following subsections.

The ability of Pin1 to catalyze the isomerization of the TM was probed using 2D EXSY (Jeener et al., 1979). The NMR samples contained 50 μM natural-abundance Pin1 and V5 peptides with concentrations ranging between 1 mM and 2 mM. The EXSY experiments were conducted on the total of six peptides: pV5βII, pV5α,pTMβII, pTMα, pTMα-P640A, and SP-2 (Supplementary file 1).

For the -pTPP-containing V5 regions, the maximum mixing time was set to 500 ms. The mixing times t_mix for pTMα-P640A were: 20, 35, 50, 75, 100, 140, 200, 280, 400, and 500 ms. The mixing times t_mix for the SP-2 were: 10, 25, 30, 40, 70, 100, 150, 200, 300, 400, and 500 ms. The cis-trans isomerization of the pThr-Pro bond was monitored by following the intensity of the diagonal and cross-peaks corresponding to the ¹H_N resonance of Thr641 and Thr638 in the βII and α peptides, respectively. The time dependence of the ratio of intensities of the cis→trans (I_ct) and diagonal trans-trans peak (I_tt), I_ct/I_tt, is given by the following equation: (10)IctItt=ktc[1−exp⁡(−kEXtmix)]kct+ktcexp⁡(−kEXtmix)

where k_ct and k_tc are the forward and reverse rate constants for cis→trans isomerization reaction, k_EX is the exchange rate constant that is equal to k_ct+k_tc, and t_mix is the mixing time. Equation (10) was used to fit the data, with k_tc and k_ct as the adjustable parameters. The populations of the cis- and trans-conformers of the peptidyl-prolyl pThr-Pro bond were estimated using the SP-2 peptide (LpTPTD), where the cis-cis diagonal peak is well resolved. The populations of the trans- and cis-conformers of the pThr-Pro bond were I_tt/(I_tt+I_cc)=91% and I_cc/(I_tt+I_cc)=9%, respectively.