Extreme multivalency and a composite short linear motif facilitate PCNA‐binding, localisation and abundance of p21 (CDKN1A)

Abstract

Cyclin‐dependent kinase inhibitor 1 (CDKN1A; also known as p21) promotes cell cycle arrest and regulates DNA replication and DNA repair by high‐affinity binding to proliferating cell nuclear antigen (PCNA) using a C‐terminal short linear motif (SLiM). High‐affinity binding to PCNA is driven by positively charged flanking regions of the SLiM, but the molecular details of their interaction as well as their roles for other p21 functions are not known. Using biophysics to study the interaction between PCNA and p21 variants with different Lys/Arg compositions in the flanking regions, as well as using D‐amino acids, we find that the flanking regions of p21 bind to PCNA likely through an interaction driven by complementary charges without specific contacts. Although the exact Lys/Arg composition of the p21 flanking regions is unimportant for high‐affinity PCNA binding, these positions are conserved in p21 orthologs, implying a conserved biological function. Accordingly, in cell‐based experiments, we find that, while the flanking regions affect p21 abundance, both the context and the Lys/Arg composition of the N‐terminal flanking region are crucial for p21 nuclear localisation. Such integration of SLiMs into a composite SLiM may be a widespread phenomenon and complicates the separation of function and drug development.

Regions flanking the PCNA‐binding motif in p21 contribute to high‐affinity binding through complementary charges without specific contacts. These regions affect p21 abundance, but their exact Lys/Arg composition is unimportant for PCNA binding. Conversely, both context and Lys/Arg composition of the N‐terminal flanking region are crucial for p21 nuclear localisation, suggesting overlapping short linear motif (SLiMs). Such composite SLiMs may be widespread in intrinsically disordered regulatory proteins, complicating separation of function. Figure generated using ChimeraX (Meng et al., Protein Sci. 32(11):e4792 (2023)).

Abbreviations

APIM

AlkB homologue 2 PCNA‐interacting motif

APSB

adaptive Poisson‐Boltzmann solver

CDK

cyclin‐dependent kinase

CDKN1

cyclin‐dependent kinase inhibitor 1

ESM

evolutionary scale modelling

IDP

intrinsically disordered protein

IDR

intrinsically disordered region

IRES

internal ribosome entry site

ITC

isothermal titration calorimetry

NLS

nuclear localisation signal

NMR

nuclear magnetic resonance

PCNA

proliferating cell nuclear antigen

PIP

PCNA‐interacting protein

PQC

protein quality control

QCDPred

quality control degron prediction

SLiM

short linear motif

Introduction

Intrinsically disordered proteins (IDPs) and protein regions (IDRs) are widespread in biology where they undertake important functions in highly regulated processes [1]. They interact with other proteins, as well as with RNA, DNA, membranes and ions, exploring different interaction mechanisms resulting in different degrees of persistent disorder in the complexes [2, 3, 4]. Many IDPs interact via short linear motifs (SLiMs), which are short (2–12 residues) conserved stretches of residues typically positioned in an otherwise nonconserved region of IDRs [5, 6, 7]. SLiM interactions often lead to folding upon binding resulting in diverse structures within the complexes, depending also on partner topology [8]. SLiMs are prevalent in cellular date‐hubs, characterised by a single protein integrating signals from a larger set of different proteins [9, 10, 11, 12]. Here, the same SLiM, carrying minor sequence variability, exists in all partners. Thus, date‐hub selectivity for a specific partner is likely not necessarily encoded solely by the SLiM itself. Instead, a larger discriminative power may lie within the sequences outside the omnipresent SLiM. Such so‐called context effects can emerge from different features and properties [13], and recent work has highlighted the importance of SLiM‐flanking regions in determining binding affinities as well as positive and negative partner selectivity [14, 15]. The flanking regions do not per se constitute part of SLiMs as these may not be conserved at the residue level, although features of nonconserved IDRs can be overall conserved [16].

A central hub in DNA homeostasis is proliferating cell nuclear antigen (PCNA) [17, 18]. It is a homotrimeric DNA clamp, which is loaded onto DNA to act as a processivity factor and hub for a large set of proteins essential for DNA replication and repair [19, 20, 21, 22, 23, 24, 25, 26], chromatin remodelling [27] and epigenetics [28]. Most PCNA ligands interact with PCNA using one of three different SLiMs—the PCNA‐interacting protein (PIP) box (QxxφxxΩΩ) [14, 29], the PIP degron (QxxφTDΩΩxxx(R/K)) [30], or the AlkB homologue 2 PCNA‐interacting motif (APIM) ([K/R]‐[F/Y/W]‐[L/I/V/A]‐[L/I/V/A]‐[K/R]) [31]. Here, φ is an aliphatic residue, Ω an aromatic residue, and x is any residue. Despite variation in the SLiM sequences, the three motifs fold upon binding to form superimposable structures consisting of a short 3₁₀‐turn packing the conserved residues into two binding pockets: one harbouring the hydrophobic and aromatic residues and one the glutamine (the Q‐pocket) (Fig. 1A).

Fig. 1.

p21 binds to PCNA via the PCNA interaction protein (PIP)‐degron short linear motif (SLiM) and its positively charged flanking regions. (A) Electrostatic surface charge mapped onto PCNA using the adaptive Poisson‐Boltzmann solver (APSB) electrostatics plugin in PyMOL. Red surface indicates negative charge and blue indicates positive charge. In the crystal structure, the p21 peptide is represented as grey cartoon with SLiM interacting residues shown as sticks (PDB ID: 1AXC [38]). (B) Sequences of the peptide variants of the p21 PIP‐degron used in the current study including positively charged flanking regions (coloured blue). Lower case letters indicate d‐amino acids. Consensus SLiMs are shown for comparison.

A well‐studied ligand for PCNA is the cyclin‐dependent kinase inhibitor, p21 (also known as p21^WAF1/Cip1) [32, 33], which by itself is a hub with many different interaction partners. It is a potent cyclin‐dependent kinase inhibitor (CDKI) functioning as a regulator of cell cycle progression [34] facilitated by the binding of p21 to cyclin‐CDK complexes through its N‐terminal domain [35, 36, 37]. p21 also plays a regulatory role in DNA replication and DNA damage repair and binds to PCNA via its C‐terminal PIP‐degron [38], blocking the binding of processivity factors necessary for PCNA‐dependent DNA synthesis [39]. Cell cycle progression is controlled by CDKs through activation by cyclin binding and phosphorylation, and inhibition by CDKIs. However, the protein level of CDKs does not fluctuate substantially during the cell cycle. Instead, regulation occurs through changes in expression level and the half‐lives of cyclins and CDKIs [40]. p21 is a well‐characterised target of the E3 ubiquitin‐protein ligase CRL4^Cdt2 resulting in controlled degradation of p21 in S‐phase [30, 41, 42, 43, 44]. It has been postulated that both the CRL4^Cdt2‐complex and the target protein bind to separate binding pockets on chromatin‐bound PCNA via their PIP‐binding motifs, thus linking CRL4^Cdt2 to its targets for effective degradation [45]. Aside from CRL4^Cdt2, p21 has been shown to be targeted for degradation by the SCF^Skp2 complex [46, 47] and APC/C^Cdc20 [48] during G1/S‐phase transition and G2‐phase, respectively, or by direct interaction with the proteasome in either a ubiquitin‐dependent or independent manner [49, 50].

The nanomolar affinity of p21 for PCNA makes it one of the strongest binding PCNA ligands known [51], suggesting p21 to be a master regulator of PCNA. Therefore, to understand the molecular origin of this very high affinity, and because most PCNA ligands carry a PCNA‐binding SLiM, we recently explored the role of the flanking regions surrounding the p21 PIP‐degron for properties that would confer this high affinity [14]. We found the many positive charges located in the immediate flanking regions of the p21 PIP‐degron to be essential for the high affinity, and their replacement by hydrophilic and negatively charged residues decreased the affinity for PCNA by a remarkable 4000‐fold [14]. The charge properties of the p21‐flanking regions match those of the surface of PCNA, which is overall highly negatively charged in the area surrounding the binding pockets (Fig. 1A). Other studies have shown that removing either the N‐ or C‐terminal flanking region decreases the affinity by −65 and ~10‐fold, respectively, but the complex still retains an affinity in the high nanomolar range [52]. In addition, the N‐terminal flanking region has been shown to harbour a nuclear localisation signal (NLS) [53].

In the present study, we further investigate to what extent the interaction between PCNA and p21 is driven by the flanking regions, and if the flanking regions bind to PCNA via specific interactions or play roles in interaction with other partners. Using a set of p21 variants and combining biophysics and bioinformatics with cell‐based studies, we find the flanking regions to be the major determinants for binding of p21 to PCNA with a strongly electrostatically driven interaction. Remarkably, the interaction is independent on charge type (Arg/Lys) and charge positions and even consents to ambidextrous binding. This highly dynamic and adaptive interaction is likely important in allowing access for binding partners of the degradation machinery to invade the PCNA:p21 complex, securing fast regulation despite high affinity. Additionally, we find that an NLS previously identified in the N‐terminal flanking region [53], as well as degron activity are imposing strong evolutionary constraints on the sequence of the flanking regions, resulting in a composite SLiM architecture. Our findings have implications for understanding competition in cellular hubs such as PCNA and p21, and we expect similar flanking region dynamics exist in other highly regulated systems that depend on SLiMs for recognition.

Results

Binding of p21 to PCNA is electrostatically driven with modest SLiM contribution

To investigate to what extent the flanking regions of the p21 PIP‐degron contribute to PCNA interaction, as well as delineate the mechanism of this interaction, we first used isothermal titration calorimetry (ITC) to measure interaction with the isolated p21 core motif QTSMTDFYHS (p21_core), removing both positively charged flanking regions (Fig. 2A). The core motif of p21 yields an interaction of micromolar affinity (Table 1). Thus, removing the positively charged flanking regions (p21_core) resulted in a ~500‐fold increase in K _D compared to the longer wild‐type (WT) p21 peptide (p21_140–156) (33 μM and 68 nm at 100 mm NaCl, respectively) (Figs 1B, 2A). Hence, as the core motif contributes only partly to affinity, this imposes a larger role for the affinity of the flanking regions. We therefore assessed whether the PCNA:p21 interaction would be driven by electrostatics involving the charged flanking regions of p21 and quantified the affinities of p21_140–156 for PCNA at different NaCl concentrations using ITC (Fig. 2B). The affinity of the complex was highly dependent on the ionic strength, ranging from a K _D of 19.3 nm at 0 mm NaCl to a K _D of 606 nm at 400 mm NaCl (Table 1) corresponding to an ~30‐fold decrease in affinity. The affinity was weakened by 1.8–3.5‐fold for each 100 mm increase in NaCl concentration. After 400 mm NaCl, no further decrease in affinity was observed. This is likely a result of the charges being fully screened by the added counter ions.

Fig. 2.

Ionic strength dependence of the PCNA:p21 interaction. (A) Representative ITC analysis of PCNA interacting with p21_140–156 (left) and p21_core (right) at 100 mm NaCl. Both isothermal titration calorimetry (ITC) isotherms were fitted to a ‘one set of sites’‐model. (B) K _D values obtained from ITC plotted against NaCl concentrations. (C) The logarithm of the K _D values of p21_140–156 plotted against the logarithm of the mean ionic activity, a_±, of NaCl. The line represents the fit of the data to the Lohman‐Record equation (see methods) and the slope of the fit approximates the number of counterions released upon binding. (D) K _D values of the interaction between PCNA and p21_140–156, p21_SWAP, p21_All–R and p21_All–K determined from ITC plotted against the NaCl concentration. The values are an average of three replicas (n = 3) and errors are SEM values between the triplicates. Positively charged amino acids are coloured blue. Note that the p21_140–156 data in D are the same as in B and are shown for easy comparison.

Table 1.

Thermodynamics of PCNA interaction with different p21 peptide variants.

Peptide	[NaCl] (mm)	K D (nm)	ΔH (kJ·mol−1)	−TΔS (kJ·mol−1)	N
p21140–156	0	19 ± 2	−73.9 ± 0.9	29.8 ± 0.7	0.77 ± 0.08
	100	68 ± 10	−75.9 ± 0.3	34.9 ± 0.6	0.86 ± 0.02
	200	182 ± 20	−90 ± 9	51 ± 9	0.86 ± 0.05
	300	341 ± 30	−89.6 ± 0.4	52.6 ± 0.2	0.76 ± 0.01
	400	606 ± 58	−93 ± 3	57 ± 3	0.80 ± 0.03
	500	579 ± 24	−89 ± 2	54 ± 1	0.72 ± 0.03
p21core	100	32 900 ± 3800	−56 ± 7	30 ± 8	0.700 ± 0.004
	300	34 300 ± 2300	−88 ± 14	62 ± 14	0.71 ± 0.08
p21All–R	100	45 ± 7	−87 ± 1	45 ± 2	0.80 ± 0.05
	200	139 ± 8	−84 ± 3	45 ± 3	0.73 ± 0.01
	300	273 ± 17	−83.5 ± 0.9	46 ± 1	0.68 ± 0.01
	400	398 ± 30	−81 ± 3	44 ± 3	0.61 ± 0.02
p21All–K	100	87 ± 9	−68 ± 2	28 ± 2	0.92 ± 0.01
	200	347 ± 56	−74 ± 4	37 ± 4	0.75 ± 0.02
	300	456 ± 30	−80 ± 3	41 ± 3	0.71 ± 0.02
	400	653 ± 44	−76 ± 3	41 ± 4	0.80 ± 0.02
p21SWAP	100	87 ± 3	−60.8 ± 0.9	20.5 ± 0.9	0.95 ± 0.04
	200	206 ± 11	−76 ± 2	38 ± 2	0.79 ± 0.09
	300	407 ± 6	−72 ± 4	36 ± 4	0.82 ± 0.04
	400	547 ± 60	−74 ± 2	38 ± 2	0.79 ± 0.01
p21DL	0	12 ± 5	−74 ± 2	28 ± 3	0.66 ± 0.04
	100	85 ± 9	−75 ± 1	35 ± 1	0.74 ± 0.07
	200	275 ± 36	−90 ± 4	53 ± 4	0.60 ± 0.01
	300	596 ± 37	−86 ± 3	50 ± 3	0.78 ± 0.02
	400	854 ± 15	−89 ± 2	54 ± 2	0.73 ± 0.01

All isothermal titration calorimetry (ITC) experiments were repeated three times and fitted to a ‘one set of sites’‐model. Errors are the SEM values between the triplicate values. Representative ITC analysis plots are illustrated in Fig. 2A.

In contrast to the ionic strength dependence of p21_140–156 binding to PCNA, the affinity for p21_core was independent of the ionic strength between 100 mm and 300 mm (Table 1). Interestingly, the affinity of p21_core was much weaker than the affinity of p21_140–156 at high salt concentrations, where an effect on K _D upon addition of NaCl is no longer observed, but the effect on hydrophobicity may begin to contribute. This could indicate that the flanking regions retain some interactions with PCNA even at high salt concentrations. As lysine and arginine besides their charge contain significant non‐polar features, these may contribute to binding at high salt because especially arginine increases in hydrophobicity with increasing salt concentrations [54]. We observed no effect on the solubility of the peptides or PCNA across the various salt concentrations.

Plotting log(K _D) vs. log(a_±), where a_± is the mean ionic activity, yields a straight line where the slope approximates the number of released counterions upon complex formation [55, 56]. Approximately two (1.7) ions were released upon p21_140–156 binding to PCNA (Fig. 2C). Since we expect the histidine residues of both p21 and PCNA to be deprotonated at this pH (pH 7.4), this leaves seven positive charges and one negative charge per p21 peptide. Comparing the number of counterions released to the number of charges of the p21 peptide (+6) suggests that fewer charged residues on the surface of PCNA are involved in complex formation. Accordingly, there appears to be more complexity in the interaction than would be expected from the formation of distinct salt bridges between the partners.

p21 flanking regions likely interact with PCNA through a mean‐field‐type interaction

Next, we examined whether the interaction between the p21‐flanking regions and PCNA is mediated by specific contacts, that is, distinct salt bridges, or by other mechanisms driven by, for example, an interaction between the positively charged flanking regions and the negative surface area of PCNA, as observed in highly charged protein complexes [2, 57] described by a mean‐field model [2, 58]. To address this, we designed peptides in which the positively charged residues in both flanking regions were replaced either by only Arg (p21_All–R), only Lys (p21_All–K) or where these were internally swapped (p21_SWAP) relative to the wild‐type sequence (p21_140–156) (Fig. 1B). We then measured the affinities and thermodynamics of binding by ITC and compared across the variants. Generally, the interaction is enthalpically driven and complex formation is associated with an entropic penalty. At low salt (100 mm), the affinities of the four peptides only differed very slightly up to ~2‐fold with K _D values between 45 nm and 87 nm (Table 1, Fig. 2D). Similarly, at increasing NaCl concentrations, the affinities of all four peptides decreased (Fig. 2D), comparably to p21_140–156 (Fig. 2B). However, we note a slight difference between Arg and Lys in their response to increasing ionic strength. Evident for all salt concentrations is that the p21_All–R peptide has the highest affinity, while the p21_All–K peptide has the lowest affinity. This could be a result of the difference in geometry between Arg and Lys where arginine has more atoms capable of carrying a charge, increasing the likelihood of forming electrostatic interactions [59]. Furthermore, Arg is more hydrophobic than Lys [54]. Despite this difference, the measured K _Ds for all peptides at a given salt concentration are within a ~2‐fold range. Thus, no strong preference for any positively charged residue exists, and since the number of counter ions released is much lower than the number of charges, these data suggest that it is more likely that a distributed set of electrostatic interactions generated by the charged side chains is responsible for the enhanced affinity allowing fast exchange of contacts and dynamic rebinding events [60].

Looking into the thermodynamics of the interactions, we see that the entropy is generally more favourable for all variants at low salt concentrations (0–100 mm NaCl), except for the p21_All–R peptide. This may suggest that the flanking regions are more dynamic in complex with PCNA at lower salt concentrations, but are more restricted at high salt, likely owing to the screening of the electrostatic interactions that become more hydrophobic instead, also explaining the increase in favourable enthalpy. These thermodynamic parameters could also suggest that arginine makes contacts that are less dynamic and provides more favourable enthalpic contributions to binding compared to lysine. Although there seems to be slight differences in the binding affinity and thermodynamics between lysine and arginine, this does not affect the conclusion that the p21 flanking regions at physiological salt concentrations likely interact through a mean‐field‐type interaction.

The flanking regions bind ambidextrously

To test the hypothesis that it is the overall distribution of charge rather than the chemical specificity of the location of the charges that is important, we designed a peptide where the flanking regions were made entirely of the unnatural d‐amino acids, while the core motif was kept in the form of the natural l‐amino acids (p21_DL) (Figs 1B, 3A). Given recent work on interactions with disordered peptides composed of d‐amino acids [61], the interactions of the PIP motif with the binding pockets of PCNA would be expected to require complementary stereochemistry. However, if the interactions by the flanking regions are driven only by a complementary charge–charge interaction with no specific or fixed contacts, the affinity should not change when changing the stereochemistry of the flanking residues. Comparing the interaction of p21_140–156 and p21_DL with PCNA, we find that p21_DL retains high affinity at low NaCl concentrations (Fig. 3A). We note that the K _D of p21_DL is slightly increased at high salt concentrations; however, the difference remained within a twofold range (Fig. 3A). This further supports the importance of a distributed set of electrostatic interactions by the flanking regions of p21.

Fig. 3.

The flanking regions bind ambidextrously with a high impact on serum stability. (A) K _D values from isothermal titration calorimetry (ITC) of the interaction between PCNA and p21_DL as a function of NaCl concentrations. Lower case letters represent D‐amino acids. The values are averages of three replicas (n = 3), errors are the SEM values between the triplicates. Note that the p21_140–156 data points are the same as in Fig. 2 and are shown for easy comparison. (B) Serum peptide abundance of p21_140–156 (black) and p21_DL (purple) after incubation in 25% human serum in PBS relative to an internal standard measured by mass spectrometry and plotted against the incubation time. p21_140–156 data were fitted to an exponential one‐phase decay. The purple dashed line shows the average p21_DL abundance.

Peptides containing D‐amino acids tend to be more proteolytically stable compared to their l‐amino acid counterparts [62, 63]. We therefore examined if there was any difference in the proteolytic stability of p21_140–156 and p21_DL. The peptides were incubated in 25% human serum in PBS, and the peptide abundances were analysed with mass spectrometry. The p21_140–156 peptide was rapidly degraded with a half‐life of ~11 min, whereas the p21_DL peptide remained stable during the 1‐hour incubation time (Fig. 3B). Collectively, this data demonstrate that p21_140–156 and p21_DL interact with PCNA with similar affinities indifferent to the stereochemistry of the flanking regions. However, the peptides vary greatly in their proteolytic stability. This observation could provide new possibilities for developing drugs targeting proteins that interact through a mean‐field‐type interaction.

The p21 SLiM‐flanking regions are conserved

The electrostatically and ambidextrous driven binding of the flanking regions of p21 to PCNA would imply the residues here to be interchangeable, imposing little evolutionary conservation on the sequence. To test this, we used the protein language model ESM‐1b [64] to predict variant effects of human p21 (CDKI 1, UniProt ID: P38936) (Fig. 4A, top). As an alternative approach, we also computed evolutionary conservation scores using the global epistatic model for predicting mutational effects (GEMME webtool [65] by first generating a sequence alignment of 1540 homologous p21 sequences, which served as input for the GEMME model (Fig. 4A, bottom).

Fig. 4.

Positively charged flanking regions of the PCNA interaction protein (PIP)‐degron are conserved in p21. (A) Heatmaps showing the predicted mutational effect from evolutionary scale modelling (ESM)‐1b (top) and GEMME (bottom). A per‐residue median score is shown on top of the heatmaps along with the human p21 sequence with the PIP‐degron flanking residues highlighted in red. The colour scales indicate the predicted mutational impact, with more negative scores indicating unfavourable mutations. Wild‐type residues are marked in white. (B) Hexbin plot comparing ESM‐1b and GEMME model scores, showing a strong positive correlation (Pearson = 0.77, Spearman = 0.68). (C) Per‐residue median scores from ESM‐1b (left) and GEMME score (right), mapped onto the AlphaFold2 structure of human p21 (UniProt ID: P38936) obtained from the AlphaFold Protein Structure Database; structural visualisation and annotation were performed using UCSF ChimeraX v1.8. We note that this structure is not meant to represent the structural ensemble of p21 in solution. Residues in the p21 PIP‐degron flanking regions are shown with sidechains and annotated in red. (D) Sequence logo generated from 515 orthologous p21 sequences from jawed vertebrates extracted from NCBI (https://www.ncbi.nlm.nih.gov/gene/1026/ortholog/?scope=7776, 2024‐06‐17). Conserved PIP box motifs (QxxφxxΩΩ; φ = A/V/I/L/M, Ω = F/Y/W) were extracted from the sequences using regular expression search and were aligned relative to the motif, including 10 flanking residues on either side. The sequence logo was generated using seqlogo v5.29.9 (https://github.com/betteridiot/seqlogo). Reporting is consistent with MIAPA guidelines [86] Numbers correspond to the positions in the human p21 sequence. The residues are coloured according to categories: Positive (H, K, R) as blue, negative (D, E) as orange, neutral (N, Q, S, T) as green, other (C, G, P) as yellow, aliphatic (A, I, L, M, V) as black and aromatic (F, Y, W) as purple.

The variant effect scores from the two methods were overall similar (Fig. 4B) with both the sequences of the p21 PIP‐degron and its flanking regions generally conserved (Fig. 4A). While the flanking regions are generally predicted to be relatively intolerant to sequence changes, we find—in line with our biochemical experiments—that R‐K and K‐R substitutions in the flanking regions are predicted to be tolerated. Mapping the median ESM‐1b masked marginal scores per position onto the AlphaFold2 structure of p21 (Fig. 4C) revealed regions predicted to be sensitive to substitutions to localise to the structured N‐terminal region and the C‐terminal PIP‐degron. The alignment‐based GEMME predictions revealed a similar trend (Fig. 4C). To gain further insights into the Lys/Arg conservation in the flanking regions, we searched for p21 (CDNK1A) orthologs in jawed vertebrates using the NCBI Gene tool. We searched the query sequences for the PIP box motif (QxxφxxΩΩ), fixing the motif in the search, and included 10 residues on each flanking side to generate a sequence logo (Fig. 4D). This revealed that the PIP‐degron and the positively charged flanking regions are conserved. Furthermore, we observe that the central TD motif (T148 and D149) of the PIP‐degron is somewhat conserved. It is also evident that Lys141 and Arg142 in the N‐terminal flanking region, and Lys154 and Arg156 in the C‐terminal flanking region are highly conserved. At position 140, leucine is slightly more prevalent than arginine, and position 143 is most often occupied by Lys or Arg, although Gly is also common here. The arginine at position 155, which is part of the PIP‐degron motif, is almost exclusively an Arg, although a small population of orthologs contains a Lys at this position.

Given that alterations in the Lys/Arg composition do not substantially affect the affinity for PCNA, the strict conservation at this position must have evolved for other reasons. The sequence logo shows that the N‐terminal flanking region of the p21 PIP‐degron (residues 140–143) is especially conserved among orthologs. This is interesting, as residues 140–142 (RKR) have been shown to be essential for nuclear localisation of p21 [53]. However, the impact of Lys/Arg substitutions for nuclear localisation is unknown.

The flanking regions impact nuclear localisation of p21

Since the p21 N‐terminal flanking region carries an NLS [53], we decided to assess the importance of the positively charged flanking regions for p21 nuclear localisation. We generated a series of full‐length p21 variants (Fig. 5A), based on the peptide variants described above, as GFP fusion proteins for expression from an integrative mammalian expression vector (Fig. 5B). We also included a full‐length variant where all the residues of the SLiM were mutated to alanine, generating a variant that as a peptide would bind very weakly to PCNA [66]. As a control, mCherry was synthesised from an internal ribosome entry site (IRES) downstream of the p21 mRNA. The expression vectors were introduced into human HEK293T cells containing a genomic landing‐pad site allowing for Bxb1‐catalysed site‐specific integration of the plasmid [67, 68]. As the expression plasmid does not contain a promoter, single‐copy expression is achieved.

Fig. 5.

The p21 flanking regions impact nuclear localisation and abundance of p21. (A) List of the full‐length p21 variants investigated. Flanking regions are shown in bold and single‐site mutations to alanine are shown in red. (B) Schematic illustration of the expression system. Upon Bxb1 catalysed site‐specific recombination of the plasmid into the genomic landing pad, the green fluorescent protein (GFP)‐fused p21 variants are expressed from a tetracycline/doxycycline‐regulated promoter (Tet‐ON). As a control, mCherry is produced from an internal ribosome entry site (IRES). Figure made with BioRender.com. (C) Fluorescence microscopy of live cells expressing the p21 variants listed in panel (A) (n = 2). Hoechst staining was used to visualise the nucleus. The scalebar measures 10 μm. Contrast has been adjusted individually for each image. (D) Western blotting of whole‐cell lysates of HEK293T landing‐pad‐containing cells expressing the indicated p21 variants (n = 2). (E) Western blotting of whole‐cell lysates of HEK293T landing‐pad‐containing cells expressing the indicated p21 variants either treated (+) or untreated (−) with 15 μm of the proteasome inhibitor bortezomib (BZ) for 16 h (n = 2).

The transfected cells were analysed by fluorescence microscopy, and as expected, we found that WT p21 localises to the nucleus. Replacing the four key residues of the PIP‐degron essential for PCNA binding [66] with alanine (p21_A–motif) still resulted in nuclear localisation of the proteins (Fig. 5C). This indicates that PCNA binding of p21 is not a determining factor for nuclear localisation of p21.

Deletion of the entire PIP‐degron and both flanking regions (p21_Δcore,ΔFR) resulted in retention of the protein in the cytosol, as was also the case for p21_ΔFR, cementing the importance of the flanking regions for nuclear localisation. In agreement with previous studies [53], deletion of the N‐terminal flanking region (p21_{ΔFR N–term}—residues R140–R143) impaired nuclear localisation. In contrast, the protein maintains nuclear localisation after deletion of the C‐terminal flanking region (p21_{ΔFR C–term}), indicating that this flanking region is conserved for other reasons. Interestingly, we notice that deletion of the core alone (p21_Δcore), which still leaves the described NLS intact, also fails to localise the p21 variant to the nucleus. This indicates that the context around the NLS or the distance between the two flanking regions is optimised for nuclear localisation. To our knowledge, this has not been observed before, and since deletion of the C‐terminal flanking region results in WT‐like localisation, it is unlikely that the p21 NLS is a bipartite NLS despite some resemblance [69]. Interchanging the R/K composition in the flanking regions resulted in a complete (p21_{All K} and p21_{All R}) or partial (p21_swap) impairment of nuclear localisation. Together, these experiments suggest not only that the N‐terminal flanking region is required for nuclear localisation, but that the exact position and the Lys/Arg in the flanking region are necessary for nuclear localisation. We conclude that some sequence specificity, especially in the N‐terminal flanking region, is required for efficient nuclear localisation of p21, which may explain the conservation of these residues. This is consistent with previous studies [53], which have identified residues 140–142 (RKR) as being an NLS. However, our data also indicate that this previously described NLS is not sufficient on its own but additionally depends on context.

The flanking regions play a minor role for p21 abundance

The cellular abundance of p21 is tightly regulated in response to cellular stimuli that control cell cycle progression and to DNA damage [40]. This prompted us to investigate whether the observed conservation of the PIP‐degron and its flanking regions is a result of its role in the regulation of the abundance of p21. As studies have described both nuclear‐specific degradation of p21, as well as the necessity of cytoplasmic translocation for successful degradation [70], the full‐length variants described above (Fig. 5A) were expressed both with and without an NLS fused to the N‐terminus of GFP. When comparing the abundance of the variants with and without the introduced NLS, we generally observed lower abundance of the variants with an NLS. Comparing the steady‐state abundance of all variants in the nucleus and the cytosol, we observe the same tendency between variants in both compartments (Fig. 5D). Compared to the WT, we observe no effect on protein levels upon deletion of the core motif on its own (p21_Δcore). However, deletion of both flanking regions (p21_ΔFR) and the entire PCNA‐binding site (p21_Δcore,ΔFR) resulted in a slight increase in abundance, indicating that the flanking regions possibly contribute to the degron potential of this region. We observed WT‐like abundance upon deletion of the N‐terminal (p21_{ΔFR N–term}) and the C‐terminal (p21_{ΔFR C–term}) flanking regions alone (Fig. 5D). In addition, replacing the Lys residues in the flanking regions with Arg (p21_{All R}) had no effect on the abundance of p21. This is surprising as this removes two out of the four Lys known to be ubiquitylated prior to degradation, including Lys141 in the N‐terminal flanking region known as the primary ubiquitylation site for SCF^Cdk2‐mediated degradation [71]. Finally, replacing all Arg with Lys (p21_{All K}), as well as interchanging the residues in the flanking regions (p21_swap) resulted in WT‐like abundance. Upon treatment with the proteasome inhibitor bortezomib (BZ), we observed an increase in the abundance of p21 (Fig. 5E), in accordance with WT p21 being a proteasome target in both the nucleus and the cytosol [40]. This effect was independent of the core motif, suggesting that p21 is degraded also independently of PCNA binding. Accordingly, a prediction of protein quality control (PQC) degrons using QCDPred [72] indicates the presence of a strong PQC degron at position 70 (Fig. 6), which potentially contributes to the degradation of both the WT and p21_Δcore proteins.

Fig. 6.

Prediction of protein quality control (PQC) degrons in p21 using Quality Control Degron Prediction (QCDPred). Plot displaying the QCDPred degron probability [72] for the p21 protein sequence (UniProt ID: P38936). Note the presence of two high probability protein quality control (PQC) degron regions positioned around residues 70 and towards the C terminus (152–155).

Discussion

In this study, we examined the biochemical and biological implications of the positively charged flanking regions surrounding the PCNA‐interacting PIP‐degron in p21. The flanking regions are critical for the nanomolar affinity to PCNA and bind with similar affinity irrespective of their Lys/Arg composition or if they are D/L stereoisomers. Given that the flanking regions appear to interact via nonspecific contacts and that a low number of counter ions are released upon interaction, this type of interaction could be interpreted as an interaction with extreme multivalency, where nearby charges in the flanking regions are available for fast re‐association and dissociation, due to their high local effective concentration. This mode of interaction has strong resemblance to the interactions dominating the complex between the highly oppositely charged proteins, prothymosin α and histone H1.0 [2], and to the effects of phosphorylation on the interaction between Sic1 and Cdc4 [73]. The high‐affinity binding may be achieved by lowering the entropic penalty upon complex formation and through an avidity effect of the PIP degron and the flanking regions [74]. Furthermore, electrostatic steering of the positively charged p21 flanking regions towards the negatively charged PCNA binding pocket may result in increased association rates, not quantified in the present study. A similar type of dynamic complex is observed for a disordered ligand interacting with a folded PDZ domain. In this case, residues outside the PDZ binding SLiM remain dynamic upon binding PDZ yet, make important contributions to the affinity mostly by electrostatic interactions [75]. p21 has been demonstrated to be specifically ubiquitinated by E3s at Lys141, Lys154, Lys161 and Lys163 [76], which are in both flanking regions. Thus, despite high‐affinity binding, the dynamics of the complex allow invading partners, such as the CRL4^Cdt2 complex, to take over, facilitated by the accessibility of the flanking regions, enabling subsequent ubiquitination by the E3 ligases. Whether this involves a transient formation of trimers to enhance the dissociation via competitive substitution [60] remains to be explored.

Although the Lys/Arg composition of the flanking regions does not greatly influence the binding affinity to PCNA at physiological salt concentrations, we still observe conservation of some positions among p21 orthologs (Fig. 4). Thus, the flanking regions retain important biological roles other than PCNA binding, and indeed, we observe that the N‐terminal flanking region (R140–R143) is essential for nuclear localisation, in accordance with previous studies [53]. Classical nuclear localisation sequences, such as the NLS in p21, bind strongly to nuclear import receptors, such as importin α, and the complex dissociates in a GTP‐dependent manner once inside the nucleus. In the known complexes of importin α, the Arg and Lys of NLSs make side chain‐specific interactions, as illustrated for the complexes involving the NLS from ChREBP and SV‐40 [77]. Thus, the exact content of the N‐terminal flanking region of p21 is important for its nuclear localisation function (Fig. 5C), which may support a more enthalpic interaction, where the region assumes fixed contacts to the nuclear transport protein. This is further supported by the observation that alterations in the Lys/Arg composition of the flanking regions influence the nuclear localisation of p21 (Fig. 5C). Thus, a more conventional interaction with specific side chain interactions with the transport protein may explain why the specific residues are conserved in the N‐terminal flanking region.

We observed no effect of the C‐terminal flanking region (residues 154–156) on nuclear localisation, suggesting other reasons for its conservation. Studies have shown that p21 is ubiquitylated at Lys141 and Lys154 [71, 76], possibly providing an explanation for the high degree of conservation observed for Lys among p21 orthologs. Additionally, previous studies have shown that the central TD residues within the core of the PIP degron, in addition to the positively charged residues at position +3 (K154) and +4 (R155) after the PIP box, are important for the recruitment and binding of CRL4^Cdt2 to the chromatin‐bound PCNA in complex with its PIP degron‐containing partners [30, 78]. Arg155 of p21 is recognised as being important for binding to CDK2 and CDK4 [79], while Arg156 is found to be methylated, which influences the subcellular localisation of p21 [80], hinting to why these residues may be conserved. Our data indicate that the flanking regions contribute to p21 abundance and thus may influence the degron potential of the region. However, since we observe proteasomal degradation of the p21_Δcore, this indicates additional E3s and/or degrons also contribute to p21 degradation. This agrees with p21 being targeted by other E3s, such as SCF^Skp2 and APC/C^Cdc20 [40], and as we predict that p21 contains an exposed PQC degron in the region around position 70, PQC E3s may also contribute. In addition, we note that p21 carries an Rxx C‐degron targeted by the Cul4^TRPC4AP E3 [81], thus further broadening the potential mechanisms for regulating p21 abundance. However, as the abundance of p21 and its binding to PCNA are both independent of the Lys/Arg composition of the flanking regions, the explanation behind the evolutionary conservation of the C‐terminal flanking region remains to be determined. The underlying reason for the conservation could be too subtle to allow us to capture them in our assays that rely on in vitro experiments and overexpression. Hence, for probing the effects of the p21 flanking regions further, it may be advisable to mutagenise p21 at its endogenous locus.

Collectively, the results describe that the PCNA‐interacting region, which include the binding motif, the degron motif and the NLS, is all located in the same short C‐terminal region of p21. It is intriguing to speculate why several biologically important elements cluster within this short region, while large parts of p21 (from residues 80–140) are much more divergent and are predicted to be tolerant towards mutations (Fig. 4). One explanation could be that functional clustering may confer an evolutionary advantage as this minimises the risk of loss‐of‐function mutations. As p21 functions as a master regulator of PCNA and checkpoint activator [35, 37, 38], neither its localisation, its binding to PCNA, nor its degradation via binding to the CRL4^Cdt2 complex and the E3‐ligase can be allowed to be compromised. Such integration of separate SLiMs into a composite SLiM may be a widespread phenomenon, but decomposition of the individual components is challenging and may complicate the development of drugs targeting specific p21 functions.

Materials and methods

Protein expression and purification

The purification of PCNA followed essentially the protocol used in previous studies [82]. His₆‐tagged PCNA in a pQE32 vector was transformed into either BL21(DE3) or BL21(DE3) pLysS cells and grown on agar plates containing either 50 μg·mL⁻¹ kanamycin (Kan) (BL21(DE3) cells) or 50 μg·mL⁻¹ Kan + 35 μg·μL⁻¹ chloramphenicol (Cam) (BL21(DE3) pLysS cells). An HRV 3C cleavage site followed the His₆‐tag (MGSSHHHHHHSSGLEVLFQGPH). From the plate, a single colony was added to 10 mL LB media with antibiotics, and the culture grown overnight (o/n) at 37 °C, 180 rpm. For protein expression, 10 mL o/n culture was transferred to 1 L LB media added antibiotics and grown at 37 °C, 180 rpm. At OD₆₀₀ ~ 0.6–0.8, 0.5 mm IPTG was added. The protein was expressed for ~5 h at 37 °C, 180 rpm and cells were harvested by centrifugation at 5000  g for 15 min. These were resuspended in 30 mL Buffer A (20 mm sodium phosphate, 1 m NaCl, 50 mm imidazole, 5 mm β‐mercaptoethanol, pH 7.4) supplemented with a complete EDTA‐free protease inhibitor cocktail tablet (Roche Diagnostics GmbH, Mannheim, Germany). Cells were lysed by sonication at 80% amplitude, 1 cycle for 10x 15 s with 15 s pause. The lysate was cleared by centrifugation at 20 000  g for 30 min and the supernatant loaded onto a gravity column containing 5 mL Ni²⁺‐NTA resin (Cytiva AB, Uppsala, Sweden) pre‐equilibrated with Buffer A. The column was washed with 50 mL Buffer A, and the protein eluted with 10 mL Buffer B (20 mm sodium phosphate, 1 m NaCl, 500 mm imidazole, 5 mm β‐mercaptoethanol, pH 7.4). The eluate was concentrated using a 15 mL 10.000 MWCO spin filter (Merck Millipore Ltd., Darmstadt, Germany) and precipitated protein removed by centrifugation at 20 000  g for 15 min. The sample was loaded onto a HiLoad 16/600 Superdex 200 size exclusion chromatography (SEC) column equilibrated with 20 mm sodium phosphate, 100 mm NaCl, 1 mm dithiothreitol (DTT), pH 7.4, and protein purity checked by SDS/PAGE with relevant fractions pooled and stored at 4 °C.

Peptides

All peptides were N‐terminally acetylated and C‐terminally amidated and were of <95% purity. p21_140–156 (Ac‐RKRRQTSMTDFYHSKRR‐NH2), p21_All–R (Ac‐RRRRQTSMTDFYHSRRR‐NH2), p21_All–K (Ac‐KKKKQTSMTDFYHSKKK‐NH2) and p21_SWAP (Ac‐KRKKQTSMTDFYHSRKK‐NH2) peptides were purchased from TAG Copenhagen (Søborg, Denmark). DL‐p21 (Ac‐rkrrQTSMTDFYHSkrr‐NH2) and core‐p21(Ac‐QTSMTDFYHS‐NH2) peptides were from BioSynth (Gardner, MA, USA).

Plasmids

p21 cDNA was codon optimised for expression in human cells and introduced into an integrative VAMP‐seq expression vector directly fused to GFP by a short linker sequence [67] (GenScript, USA). Single‐site variants of p21 were generated by GenScript (Piscataway, NJ, USA). The Bxb1 expression vector has been described previously by Matreyek et al. [67].

Isothermal titration calorimetry (ITC)

PCNA was buffer exchanged and concentrated on a 15 mL 10.000 MWCO spin filter (Merck Millipore, Darmstadt, Germany). Buffers used for ITC varied in NaCl concentration, and all buffers contained 20 mm sodium phosphate, 0–500 mm NaCl, 1 mm tris (2‐carboxyethyl)phosphine (TCEP), pH 7.4. Peptides were dissolved in the same buffers as PCNA, and the pH was checked and adjusted if needed. The concentrations of PCNA and peptides were determined using A₂₈₀ measured on a NanoDrop ND‐1000 instrument (Thermo Fisher Scientific, Waltham, MA, USA). Extinction coefficients of PCNA and peptides were calculated using ExPASy ProtParam (https://web.expasy.org/protparam/). All ITC experiments were carried out with PCNA in the cell and peptide in the syringe. Protein and peptide samples were centrifuged at 20 000  g for 15 min prior to the experiments to degas the samples. All ITC experiments were recorded at 25 °C on a Malvern MicroCal PEAQ‐ITC instrument (Malvern, Cheshire, UK) using a stir speed of 750 rpm, and the data were fitted to a one‐site binding model using the MicroCal PEAQ‐ITC Software. All ITC experiments were repeated at least 3 times.

Calculation of counterion release

In the presence of NaCl, chloride anions ($n_{-}$) and sodium cations ($n_{+}$) associated with PCNA and p21 peptides are released upon complex formation:

$$\text{PCNA}+\mathrm{p}21\rightleftharpoons \text{PCNA}:\mathrm{p}21+n_{+}\mathrm{N}{\mathrm{a}}^{+}+n_{-}\mathrm{C}{\mathrm{l}}^{-}$$

The total number of counterions ($\Delta n$) released upon PCNA:p21 binding is related to the dependence of the observed dissociation constant, K _D, as a function of the mean ionic activity, $a_{\pm }$, of salt at different concentrations [55, 56]:

$$\Delta n\approx \frac{d\log \left(K_{\mathrm{D}}\right)}{d\log \left(a_{\pm }\right)}$$

where $a_{\pm }$ is the mean ionic activity coefficient. The mean ionic activity coefficient, $a_{\pm }$, of a salt is given by:

$$a_{\pm }=m_{\pm }{\gamma }_{\pm }$$

where $m_{\pm }$ is the mean ionic molality, and ${\gamma }_{\pm }$ is the mean ionic activity coefficient. The mean ionic molality, $m_{\pm }$, is given by:

$$m_{\pm }=m{\left(v_{+}^{v_{+}}{v}_{-}^{v_{-}}\right)}^{\frac{1}{v_{+}+v_{-}}}$$

where $v_{+}$ and $v_{-}$ are stoichiometric coefficients of the cation and anion. For example, the mean ionic molality of a 100 mm NaCl concentration is calculated as $m_{\pm }=0.1{\left(1^1{1}^1\right)}^{\frac{1}{1+1}}$. Due to the low NaCl concentrations used in the ITC experiments, molal and molar concentrations are essentially the same, and we therefore use molar concentrations in our calculations. Tabular values of the mean ionic activity coefficient, ${\gamma }_{\pm }$, of NaCl [83] were used to calculate the mean ionic activity $a_{\pm }$.

Stability measurements by mass spectrometry (MS)

10 μm peptide was added to 25% human serum in PBS and kept at 37 °C. Samples were collected 0, 2.5, 5, 10, 20, 40 and 60 min after treatment by transferring peptide‐treated serum to 0.2% formic acid (1 : 4) containing an internal standard. Samples were kept at 4 °C until analysis by MS. The samples were analysed on a Waters Acquity UPLC‐TQ‐XS triple quadrupole mass spectrometer system operated in positive ESI mode. Separation was performed on an Agilent Poroshell 300SB‐C18 2.1 × 7.5 mm column with gradient elution using 0.25% formic acid in water (A) and acetonitrile (B) as mobile phases. A peptide with a similar molecular weight was used as the internal standard, and abundances were reported as the response factor relative to the internal standard using the transition 383.66/84.03 as quantifier for the p21_140–156 and p21_DL peptides.

ESM‐1b

We adopted a similar approach to that described in a previous study [84] using the ESM‐1b model [64] on the target protein, cyclin‐dependent kinase inhibitor 1 (UniProt ID: P38936), using its primary sequence. Specifically, we employed the pretrained ESM‐1b model comprising of 650 million parameters (ESM‐1b_t33_650M_UR50S, available at https://github.com/facebookresearch/esm) and applied a masked marginal likelihood approach to estimate variant effect scores. This involved masking the residue at the target position, computing the predicted probabilities for all possible amino acid residues and deriving the score as the ratio between the probability of the wild‐type residue and that of the substituted amino acid [84] to assess the effects of amino acid variants using the ESM‐1b model [64].

GEMME

We ran the Global Epistatic Model for predicting Mutational Effects (GEMME) [65] available at the GEMME webpage http://www.lcqb.upmc.fr/GEMME/Home.html, using a sequence file containing the protein sequence of cyclin‐dependent kinase inhibitor 1 (UniProt ID: P38936). The pipeline was run with default settings. Initially in the pipeline, an input multiple sequence alignment was generated by running five iterations of jackhmmer [85] against the Uniref100 database using the EV couplings framework (https://github.com/debbiemarkslab/EVcouplings). The conservation level was calculated using two iterations. We used GEMME to calculate the full single‐site mutational landscape, which combines evolutionary conservation, evolutionary fit and site‐independent frequencies.

Sequence logo generation

We generated a sequence logo plot by identifying orthologs of human cyclin‐dependent kinase inhibitor 1 (CDKN1A/p21, UniProt: P38936) in jawed vertebrates using the NCBI Gene tool (https://www.ncbi.nlm.nih.gov/gene/1026/ortholog/?scope=7776 https://www.ncbi.nlm.nih.gov/gene/1026/ortholog/?scope=7776), retrieving 539 sequences as of June 17^th, 2024. We searched for conserved PIP box motifs using a regular expression for the canonical motif (QxxφxxΩΩ) (φ = A/V/I/L/M, Ω = F/Y/W) and identified 515 orthologs containing a matching motif. For each hit, the motif was aligned by position, and the surrounding 10 flanking residues on either side were extracted. No formal multiple sequence alignment was performed. The resulting motif‐flanking regions were used to generate a sequence logo using seqlogo v5.29.9 (https://github.com/betteridiot/seqlogo). This analysis follows reporting guidelines outlined in the MIAPA framework retrieving 539 sequences as of June 17^th, 2024. We searched for conserved PIP box motifs using a regular expression for the canonical motif (QxxφxxΩΩ; φ = A/V/I/L/M, Ω = F/Y/W) and identified 515 orthologs containing a matching motif. For each hit, the motif was aligned by position, and the surrounding 10 flanking residues on either side were extracted. No formal multiple sequence alignment was performed. The resulting motif‐flanking regions were used to generate a sequence logo using seqlogo v5.29.9 (https://github.com/betteridiot/seqlogo) [86].

Cell culture

The landing‐pad containing HEK293T cells (RRID: CVCL_0063) (cell line: TetBxb1BFPiCasp9) [68] were propagated in Dulbecco's modified Eagle's High‐Glucose medium (DMEM) (Sigma‐Aldrich, St. Louis, MO, USA) supplemented with 10% fetal bovine serum (Sigma, St. Louis, Missouri, USA), 2 mm glutamine (Sigma), 5000 UI·mL⁻¹ penicillin G (BioChemica, Monza, Italy) and 5 mg·mL⁻¹ streptomycin sulfate (BioChemica) at 37 °C in a humidified incubator with 5% CO₂. Cells were passaged using trypsin at 80–90% confluency. The cells tested negative for mycoplasma (Mycostrip, InvivoGen, San Diego, CA, USA) and were regularly authenticated by selection for recombinants with 10 nm AP1903 (MedChemExpress, Monmouth Junction, NJ, USA) and 2 μg·mL⁻¹ doxycycline (Sigma‐Aldrich). For transfection of p21 constructs in a 96‐well plate (and proportionally more for larger plates), FugeneHD (Promega, Madison, WI, USA) was used following the manufacturer's instructions: 40 ng p21 expression plasmid, 3 ng Bxb1 expression plasmid and 0.2 μL FugeneHD in a final volume of 8 μL Optimem (Thermo Fisher Scientific). Selection for recombinant cells was performed by treatment with 10 nm AP1903 (MedChemExpress) with simultaneous induction of protein expression from the landing‐pad site using 2 μg·mL⁻¹ doxycycline (Sigma‐Aldrich) 48 h after transfection. All cells were analysed after an additional 4 days of culturing. Cells were treated with bortezomib (LC Laboratories, Woburn, MA, USA) for 16 h at a final concentration of 15 μm.

Fluorescence microscopy

A 96‐well microplate with a polymer coverslip bottom (Ibidi) was coated for optimal cell adherence by incubation with 50 μL poly‐D‐lysine (50 μg·mL⁻¹) (Gibco, Waltham, MA, USA) in PBS (10 mm Na₂HPO₄, 1.8 mm KH₂PO₄, 137 mm NaCl, 3 mm KCl, pH 7.4) for 1 h. Subsequently, wells were washed with distilled water to remove any residual poly‐D‐lysine and left to air‐dry. Once the wells were dry, 10 000 transfected cells were seeded in the wells. After 24 h in growth medium supplemented with doxycycline, the cells were treated with 5 μm Hoechst 33342 staining dye (Abcam, Cambridge, UK) directly in the growth medium and incubated at 37 °C for 1 h. The cells were analysed by live cell imaging on an ImageXpress Confocal HT.ai microscope (Molecular Devices, San Jose, CA, USA) equipped with a 20× water‐immersion objective and MetaXpress software. Subsequently, images were analysed using CellProfiler [87] and fiji [88].

SDS/PAGE and western blotting

Cells were washed with PBS and directly lysed in SDS sample buffer (3% SDS, 93 mm Tris/HCl pH 6.8, 18% (v/v) glycerol, 0.02% (w/v) bromophenol blue, 2.5% (v/v) β‐mercaptoethanol). The samples were sonicated 3 × 10 s and boiled for 3 min. The proteins were resolved on 7 × 8 cm 12.5% (w/v) acrylamide separation gels with 3% (w/v) stacking gels using a constant voltage of 125–150 V for 1 h in reservoir buffer (50 mm Tris, 0.4 M glycine, 0.1% SDS). PageRuler prestained protein ladder (Thermo Fisher Scientific) was used as a molecular weight marker. Next, the proteins were transferred onto a 0.2 μm nitrocellulose membrane (Advantech, NJ, USA) in blotting buffer (50 mm Tris, 100 mm glycine, 0.01% (w/v) SDS, 20% (v/v) 96% ethanol) at 100 mAmp/gel for 1.5 h. The transferred protein was stained by Ponceau S (0.1% Ponceau S (Sigma‐Aldrich), 5% (v/v) acetic acid). Next, the membranes were incubated in 5% (w/v) dry milk powder in PBS containing 2.5 mm NaN₃ and 0.1% (v/v) Tween‐20 for 30 min. The membranes were washed in PBS and incubated o/n in their respective primary antibodies at 4 °C. After three rounds of 5 min incubation in wash buffer (50 mm Tris/HCl pH 7.4, 150 mm NaCl, 0.01% (v/v) Tween‐20) the membranes were incubated in their respective horse radish peroxidase (HRP)‐conjugated secondary antibodies for 1 h. Following another 3 rounds of incubation with wash buffer, the membranes were treated for 3 min with ECL detection reagent (Amersham GE Healthcare, Amersham, UK) and developed using a Bio‐Rad ChemiDoc MP Imaging System (Bio‐Rad, Hercules, CA, USA). The antibodies used were anti‐GFP (diluted 1:1000) (Chromotek, 3H9, Martinsried, Germany), anti‐mCherry (diluted 1:1000) (Chromotek, 6G6, Martinsreid, Germany), HRP‐conjugated anti‐rat (diluted 1:5000) (Invitrogen, 31 470, Waltham, MA, USA) and HRP‐conjugated anti‐mouse (diluted 1:5000) (Dako, Glostrup, Denmark, P0260).

Conflict of interest

KL‐L holds stock options in and is a consultant for Peptone Ltd. All other authors declare no competing interests.

Author contributions

SS, FBL, RHP and BBK designed the study. SS purified proteins and performed the ITC experiments and analysis, supervised by BBK. FBL performed all in‐cell experimental work, supervised by RHP. CKS, MO and PB performed and analysed the stability measurements by MS. NJ performed the computational analysis and ensured the data availability of the computational data for the article with supervision from KLL. SS, FBL, RHP and BBK wrote the manuscript with input from all other authors.

Peer review

The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer‐review/10.1111/febs.70133.

Data availability statement

All data and software generated for this article are available on GitHub: https://github.com/KULL‐Centre/_2024_simonsen‐larsen_pcna.