A thermosensitive PCNA allele underlies an ataxia-telangiectasia-like disorder

Proliferating cell nuclear antigen (PCNA) is a sliding clamp protein that coordinates DNA replication with various DNA maintenance events that are critical for human health. Recently, a hypomorphic homozygous serine to isoleucine (S228I) substitution in PCNA was described to underlie a rare DNA repair disorder known as PCNA-associated DNA repair disorder (PARD). PARD symptoms range from UV sensitivity, neurodegeneration, telangiectasia, and premature aging. We, and others, previously showed that the S228I variant changes the protein-binding pocket of PCNA to a conformation that impairs interactions with specific partners. Here, we report a second PCNA substitution (C148S) that also causes PARD. Unlike PCNA-S228I, PCNA-C148S has WT-like structure and affinity toward partners. In contrast, both disease-associated variants possess a thermostability defect. Furthermore, patient-derived cells homozygous for the C148S allele exhibit low levels of chromatin-bound PCNA and display temperature-dependent phenotypes. The stability defect of both PARD variants indicates that PCNA levels are likely an important driver of PARD disease. These results significantly advance our understanding of PARD and will likely stimulate additional work focused on clinical, diagnostic, and therapeutic aspects of this severe disease.

Proliferating cell nuclear antigen (PCNA) is a sliding clamp protein that coordinates DNA replication with various DNA maintenance events that are critical for human health. Recently, a hypomorphic homozygous serine to isoleucine (S228I) substitution in PCNA was described to underlie a rare DNA repair disorder known as PCNA-associated DNA repair disorder (PARD). PARD symptoms range from UV sensitivity, neurodegeneration, telangiectasia, and premature aging. We, and others, previously showed that the S228I variant changes the protein-binding pocket of PCNA to a conformation that impairs interactions with specific partners. Here, we report a second PCNA substitution (C148S) that also causes PARD. Unlike PCNA-S228I, PCNA-C148S has WT-like structure and affinity toward partners. In contrast, both disease-associated variants possess a thermostability defect. Furthermore, patient-derived cells homozygous for the C148S allele exhibit low levels of chromatin-bound PCNA and display temperature-dependent phenotypes. The stability defect of both PARD variants indicates that PCNA levels are likely an important driver of PARD disease. These results significantly advance our understanding of PARD and will likely stimulate additional work focused on clinical, diagnostic, and therapeutic aspects of this severe disease.
Proliferating cell nuclear antigen (PCNA) is a sliding clamp protein that acts as a molecular "hub" for chromosomal DNA replication, repair, and maintenance. PCNA consists of three identical subunits that form a closed ring that surrounds DNA (1). Each subunit contains one partner binding region that various factors use to interact with PCNA, allowing the trimeric ring to accommodate up to three partners simultaneously (1,2). Because PCNA is a closed ring, it must be loaded onto DNA by the replication factor C (RFC) clamp loader complex (3,4). After loading, PCNA functions as a molecular tether that increases the processivity of various DNA-acting enzymes ( Fig. 1; (5)). Beyond its role in enzymatic processivity, PCNA functions as a scaffold that congregates various factors that copy, surveil, and repair DNA; thus, PCNA coordinates DNA replication with DNA repair, chromatin remodeling, cell cycle regulation, and apoptosis (5). Proper genome stability requires partners to physically interact with PCNA; mutations that abrogate binding can be embryonically lethal (6,7). Therefore, it is critical to understand how PCNA modulates its interactome and elucidate how deviations from normal function impact human health.
Recently, a rare genetic disease, PCNA-associated DNA repair disorder (PARD; alternative name: Ataxia-Telangiectasia-like disorder 2) was reported in four individuals homozygous for the hypomorphic PCNA-S228I variant (8). PARD shares a variety of symptoms (e.g., UV sensitivity, neurological abnormalities, and growth defects) with other DNA repair disorders such as Ataxia-Telangiectasia, Cockayne syndrome, and Xeroderma Pigmentosum but presents with a much larger breadth of abnormalities (8). Despite its primary role as a replication protein, the S228I variant does not significantly perturb bulk DNA synthesis. In contrast, the S228I substitution increases cells' sensitivity to UV-induced DNA damage, pointing to a defect in DNA repair and/or the DNA damage response. From a structural standpoint, the S228I substitution dramatically shrinks the partner binding site (9,10), which disrupts interactions between PCNA and a subset of its partners (8)(9)(10)(11). Because only certain partner interactions are disrupted by the S228I variant, a view prevailed that PARD is driven by altered specificity for certain binding partners involved in DNA repair (8)(9)(10)(11).
In this study, we describe specific clinical and genetic features of three unrelated children who display apparent PARD symptoms and who are homozygous for a novel PCNA-C148S allele. Like patients with PARD with PCNA-S228I, these individuals suffer from a combination of photosensitivity, growth delays, and neurodegeneration. Unlike PCNA-S228I cells, cells harboring the C148S variant appear sensitive to dsDNA break-inducing agents. Furthermore, the C148S alteration does not induce large conformational changes in the PCNA protein, nor does it impair partner binding. However, we find that both variants induce stability defects in the PCNA protein. Furthermore, we observe that patient-derived cells carrying PCNA-C148S exhibit cellular defects consistent with a loss of PCNA stability. Our results challenge the current prevailing idea that altered binding specificity accounts for PARD. We suggest that PCNA longevity and/or levels of chromatinbound PCNA underlie this rare disease.

Results
Clinical description of patients 1, 2, and 3 We studied three unrelated patients with a phenotype suggestive of PARD ( Fig. 2 and Table 1). Two individuals (patients 1 and 2, Fig. 2) were born to healthy consanguineous parents (first cousins) who had a history of spontaneous abortions. There is no history of consanguinity in the family of patient 3, but both parents came from a small, isolated town in northeastern Brazil (Fig. S1). The three subjects share a phenotype characterized by severe short stature, microcephaly, developmental delays, ataxia, and telangiectasia ( Fig. 2, C-F and Table 1). Clinical evidence of neurodegeneration started at 4 to 5 years of age and was defined by progressive gait instability, muscle weakness, and speech difficulties. Neuroimaging showed cerebellar atrophy (Fig. 2F) in patients 2 and 3 but was normal in the younger patient 1. Recurrent respiratory infection was reported in two subjects (patients 2 and 3). A mild immunoglobulin deficiency (i.e. IgM, IgG, and IgA) was identified in all subjects, with normal white blood cell count. Complete reports of all three individuals are available in the Supplement.
Whole-exome sequencing of patient samples reveals the presence of a homozygous PCNA mutation To identify the genetic cause underlying the observed disorders, we performed whole-exome sequencing on all three patients. In all cases, the average coverage of the target exonic regions was >105× with a coverage range of >96%. We filtered sequence variants by focusing on rare deleterious homozygous variants across all three patients, identifying 6 to 15 missense mutations in each patient (Table S1).
We found that the only potentially deleterious variant shared among all three patients is a homozygous transversion mutation within exon five of the PCNA gene, (variant ref ID: NM_002592.2: c.443G>C). The resulting mutation causes a cysteine-to-serine alteration at position 148 of the PCNA protein ([p.C148S], NCBI dbSNP ID: rs1274412848) (12). We confirmed that this mutation is homozygous in all three patients using Sanger sequencing (Fig. 2B). We also determined the genetic status of healthy family members in families 1 and 2 by Sanger sequencing and found all were heterozygous for PCNA-C148S. This allele is very rare in the general population, as it is only found at a very low frequency in the gnomAD database (minor allele frequency of 0.00002897 in the Latino population). In addition, our analysis revealed all three patients share common polymorphisms within 250 kb (Chr20:4,972,112-5,221,841) around the PCNA gene (Chr20: 5,098,255). While this observation suggests that the three patients share a common ancestor, all three reside in distinct regions of Brazil. The fact that they have no known shared family history (Fig. S1) led us to conclude that they are distantly related relatives. The C148 residue is highly conserved across all metazoan species (Figs. 2G and S1), and the C148S variant is predicted to be pathogenic (Table S2). Taken together, our results collectively suggest that the underlying cause of all three patients' disease is due to the presence of the C148S substitution in PCNA.  (Fig. 3A). As a reference, we used two different healthy donor cells (HDCs) as negative controls and fibroblasts from patients with Cockayne syndrome (pol η deficient) as positive controls. We found that, unlike PCNA-S228I, C148S fibroblasts were as resistant to UV irradiation as HDCs (Fig. 3A). Because cells defective in DNA replication and repair machinery are often sensitive to dsDNA break-inducing agents (13,14), we next investigated whether C148S fibroblasts were sensitive to increasing dosages of ionizing radiation (Fig. 3B). We found that patient-derived C148S cells were more sensitive to ionizing radiation-induced DNA damage compared with HDCs. We confirmed our results by pharmacologically inducing dsDNA breaks with zeocine and again observed an increased sensitivity in C148S cells compared with HDCs (Fig. 3C). Our combined results suggest that the C148S mutation impairs dsDNA break repair, which manifests as reduced viability, but does not increase sensitivity to UVinduced damage. This is a significant distinction between patient-derived C148S and S228I fibroblasts, which we decided to investigate biochemically.

No major conformational change in PCNA-C148S
We tested whether the C148S substitution affects the overall structure of PCNA, as was observed with S228I (10). We determined the crystal structure of PCNA-C148S to 3.1-Å resolution (Protein Data Bank [PDB]: 8E84, Fig. 4A Figure 2. A novel Ataxia-Telangiectasia-like syndrome associated with an alteration in a highly conserved region of PCNA. A, pedigree tree of the three patients. The known genotype of each individual confirmed by Sanger sequencing is below each symbol ("WT" is the reference allele and "Mut" is the altered allele). B, Sanger sequencing confirming the point mutation (c. 442G.C>; pC148S) in the PCNA gene. C-F, clinical presentation of patients with the C148S allele. Patients present with (C) mild facial dysmorphisms: broad forehead, low-set posteriorly rotated ears, broad nasal bridge, thin upper lip, and cutaneous telangiectasia located on face (patient 1, 2.5 years of age), (D) ocular telangiectasia (patient 3, 12 years of age), (E) systemic cutaneous telangiectasia. F, midline sagittal T1-weighted brain scan of patient 3 showing cerebellar atrophy. See Supplemental Text for the complete clinical presentation of each patient. G, protein sequence alignment of PCNA from several species. A cysteine at position 148 is conserved across multicellular eukaryotes. PCNA, proliferating cell nuclear antigen. Table S3). We found that overall the structure of PCNA-C148S is similar to that of PCNA-WT with an RMSD value of 0.85 Å, indicating no significant global changes to the structure. Furthermore, the conformation of residues surrounding C148S does not differ significantly from PCNA-WT. Although the S228I variant has a deformed partner binding cleft (9,10), this region is not obviously altered in PCNA-C148S. It is important to note that a portion of the binding cleft exhibits weak density in our PCNA-C148S structure. This challenge prevented precise modeling of the entire partner binding cleft. Despite this issue, there is unambiguous density for Y133, a residue that dictates the conformation of the partner binding cleft (9,10). Because Y133 is in the same conformation as seen in PCNA-WT, we conclude that the partner binding cleft is unchanged in PCNA-C148S (Fig. 4A).

The C148S substitution does not impair partner binding
Because PCNA-S228I is defective for binding to numerous partner proteins (9, 10), we sought to investigate if a similar defect exists with PCNA-C148S. We focused our efforts on three different PCNA partners: p21 CIP1 , FEN1, and RNaseH2B, as these proteins participate in various DNA replication and/or damage response pathways and their affinity for PCNA-WT and/or S228I were previously reported (9,10,15). We measured partner PIP-box peptide affinity by isothermal titration calorimetry. We found that PCNA-WT and C148S bind partners with similar affinity (Figs. 4B and S2 and Table S4), except for a subtle increase in affinity of C148S-PCNA for RNaseH2B. In contrast, all three peptides weakly interact with PCNA-S228I (9, 10). To further explore the binding properties of both PARD variants, we investigated the binding of PCNA variants to a peptide of the PIP-Box from DNA polymerase δ (pol δ). We found a similar binding affinity of pol δ for PCNA-WT and C148S and slightly weaker affinity for PCNA-S228I. Collectively, these results indicate that the two associated PARD variants do not share the same partner binding properties.
To see if these differences in binding extend to full-length partners, we measured binding to two full-length partners (RFC and FEN1) instead of PIP-box peptides. PCNA is known to stimulate the ATPase activity of RFC (3,11,16), and we therefore leveraged this feature to assess RFC binding to PCNA. PCNA-WT, C148S, and S228I stimulated the ATPase activity of RFC to the same order of magnitude (Fig. 4C). We then determined affinity of PCNA for RFC by measuring ATPase activity as a function of PCNA concentration (Fig. 4D), as previously reported for PCNA-S228I (11). We found that PCNA-WT, C148S, and S228I all have similar affinity for RFC (K d, app 150 nM, 200 nM, and 120 nM, respectively). We next assessed the binding affinity between FEN1 and PCNA through a pull-down experiment using Histagged full-length FEN1. As expected, PCNA-S228I binds FEN1 poorly (9), while both WT-PCNA and C148S have Thermosensitive PCNA variants underlie a rare disorder similar affinities for FEN1 (Fig. 4, E and F). Our combined structural and affinity binding studies indicate that, in contrast to PCNA-S228I, PCNA-C148S does not exhibit the defect in partner binding that was thought to drive PARD.

PARD mutations impair PCNA stability
From a structural point, the C148 residue is buried in a hydrophobic pocket near the interface between subunits and participates in a π-sulfur interaction with F144 (Fig. 4A). These types of interactions have been shown to confer substantial protein stability (17). Thus, we hypothesized that the C148S substitution disrupted protein stability. To investigate this hypothesis, we measured thermostability of PCNA using a thermal shift assay, with tertiary structure monitored by PCNA's intrinsic tryptophan fluorescence. We found that all three PCNA variants display unfolding curves consistent with a two-state unfolding mechanism (Figs. 5 and S3 and Table S5). Both PCNA-C148S and -S228I have dramatically reduced thermal stability compared with PCNA-WT (PCNA-WT T m : 52.0 ± 0.03 C, PCNA-C148S T m : 42.0 ± 0.06 C, and PCNA-S228I T m : 44.0 ± 0.1 C). Based on our data, we estimate that 17 to 22% of PCNA-C148S and -S228I are in the unfolded state at normal physiological temperature. To quantify the energetic consequences of the C148S substitution on PCNA stability, we conducted equilibrium unfolding experiments. We incubated PCNA-WT and C148S with different concentrations of the denaturing agent guanidinium hydrochloride (Gdm-HCl) and monitored secondary structure by circular dichroism (CD) and tertiary structure by tryptophan fluorescence. Both readouts exhibit a steep transition at identical Gdm-HCl concentrations, indicating a single, cooperative transition between the folded and unfolded state (Fig. 5B). In addition, our results showed that PCNA-C148S unfolds at a lower concentration of Gdm-HCl than wildtype (i.e. 0.5 M), verifying a stability defect through independent approaches. Our results suggest that the C148S substitution causes an 0.6 kcal/mol loss in folding free energy; however, the precision of this estimate is not entirely clear due to the noisy baselines in the PCNA-C148S data. Collectively, our results indicate that PCNA-C148S and -S228I are significantly less stable than wildtype and that a fraction of each variant is unfolded at physiological temperatures.

PARD variants are quickly inactivated at elevated temperature
Noting the stability defect of the PARD associated variants, we next asked if the substitutions compromised the formation of functional trimers. We measured PCNA oligomerization after incubation at 4, 37, and 42 C using a native polyacrylamide gel assay (Fig. 6A, (18,19)). As a point of comparison, we used PCNA-D150E, a substitution that is known to disrupt trimerization and lead to monomerization (20,21). We found that PCNA-WT, -C148S, and -S228I all migrated as trimers after incubation at 4 C, while D150E migrated as an expected monomer. In contrast, preincubation at 42 C led to the formation of higher molecular bands indicative of aggregation for all PCNA variants except PCNA-WT, which primarily behaved as a trimer. We did observe that a small fraction of PCNA-S228I persisted as a trimer, consistent with its slightly higher melting point or a potential refolding event. This heat-induced aggregation also occurred at 37 C for both PARD-associated variants, but to a lesser extent than at 42 C, consistent with our tryptophan fluorescence data (Figs. 5A and S4A).
We next investigated how each PARD substitution impacts the functional half-life of PCNA at various temperatures. Because RFC's activity depends on functional-trimeric PCNA, we used its ATPase activity to measure levels of functional PCNA (Fig. 6B). We preincubated each variant for 1 or 24 h at 4, 25, or 42 C and then measured ATPase activity. All PCNA variants retain the ability to stimulate RFC ATPase activity when preincubated at 4 C or 25 C (Fig. S4 Because PCNA is functional only when bound to DNA, we next asked if heat inactivation of each PARD-associated variant impacted their ability to load onto DNA. To address this, we used a bead-based assay that allowed us to monitor the loading of PCNA onto a single-stranded DNA plasmid (Figs. 6D and S5, A and B). After 5 min, a distinct signal for PCNA was detected that is dependent on both RFC and ATP. We confirmed this assay by linearizing the DNA with a restriction enzyme (i.e., BglII), which allows PCNA to slide off the DNA. This eliminated the PCNA signal, indicating that PCNA is specifically loaded onto the DNA substrate. With this system, we monitored loading of PCNA that was either preincubated at 25 C or 42 C. We found that PCNA-WT and -C148S load onto DNA at similar rates when preincubated at 25 C (t 1/2 : 12-15 s; Fig. 6E) and to a similar extent for PCNA-S228I (Fig. S5C). In contrast, after incubation at 42 C, both PARD-associated variants load onto DNA at 10× lower levels compared with WT-PCNA (Figs. 6F and S5D). Collectively, our results suggest that both PARD-associated variants function as temperature-sensitive mutants that form trimeric PCNA at low temperatures but unfold and become inactivated at physiological temperatures and above.

Patient-derived C148S cells have lowered levels of PCNA
Given that both PARD-associated variants have decreased stability, we next asked how these substitutions impact total PCNA levels in patient-derived cells. To test this, we compared PCNA levels from patient-derived cells versus HDCs. We detected significantly lower PCNA levels in patient-derived cells harboring the C148S substitution compared with HDCs (Fig. 7A). We next asked if these lower PCNA levels are altered in specific compartments in the cell. Therefore, we fractionated whole-cell lysates and found consistently lower levels of PCNA in all cellular compartments of patient-derived cells (Fig. 7B). Most strikingly, we detected much lower chromatinbound PCNA in PARD-derived cells than HDCs, suggesting that the C148S substitution impacts the levels of PCNA on DNA in cells.
We next asked if decreased PCNA-C148S stability could drive the lower levels of PCNA in patient-derived cells. Therefore, we performed cycloheximide pulse-chase experiments to monitor PCNA lifetime in cells (Fig. 7C). We found after 24 h of cycloheximide treatment the levels of PCNA in the patient-derived cells decreased much faster than in the HDCs. Therefore, our combined molecular and cellular results indicate that PCNA-C148S is less stable than PCNA-WT and that this defect reduces the levels of PCNA on DNA.

Patient-derived-C148S cells exhibit defects at elevated temperatures
Noting the thermal stability defect of PCNA-C148S, we next asked if the cellular dsDNA sensitive phenotype correlated with PCNA instability. We reasoned that elevated temperatures should impact PCNA activity and therefore increase the susceptibility of PARD cells to dsDNA break-inducing agents.
To test this, we compared the levels of γH2AX, a known marker of dsDNA breaks (22), at both 37 C and 42 C in patient-derived and control cells using flow cytometry (Fig. 8A). We found that PCNA-C148S cells have a slightly higher level of γH2AX at 37 C compared with HDCs, but at 42 C the levels of γH2AX is much higher than in HDCs. These results suggest that the patient-derived cells have higher levels of basal dsDNA breaks, which is exacerbated at higher temperatures.
Because γH2AX signaling can trigger cellular arrest (22), we next tested whether the observed temperature-induced increase in γH2AX expression resulted in decreased cellular growth. Therefore, to monitor the growth of each cell line, we performed a wound-healing assay at both 37 C and 42 C (Figs. 8, B and C and S6). Following cell adhesion, the monolayer was mechanically disrupted using a pipet tip. We then monitored "wound" healing over time by light microscopy. At 37 C, all cell lines were able to grow back into a single monolayer within 48 h. In contrast, at 42 C, only the HDCs showed complete recovery, while the patient-derived cells harboring PCNA-C148S showed minimal growth. Therefore, these results suggest that cells harboring the C148S variant have restricted proliferation at elevated temperatures. Collectively, our findings suggest that elevated temperatures impair cell viability and provides a step toward a molecular understanding for PARD.

Discussion
The PCNA-C148S allele is associated with PARD patients present with an atypical collection of symptoms that include neurodegeneration, skin photosensitivity, cutaneous abnormalities, and developmental disorders. At the cellular level, both variants exhibit different phenotypes. Unlike PCNA-S228I, PCNA-C148S fibroblasts appear insensitive to UV-induced damage. We cannot completely rule out that all cell types carrying PCNA-C148S are insensitive to UV-induced damage, as our study only focused on fibroblasts. In contrast, we find that PCNA-C148S fibroblasts are sensitive to doublestrand break-inducing agents; the effect of double-strand breaks on PCNA-S228I cells remains unknown (8). Regardless, our combined clinical, genetic, and experimental findings strongly suggest that the presence of a homozygous PCNA-C148S is a novel cause of PARD. In general, patients with either PCNA-C148S or -S228I present with symptoms similar to other DNA repair disorders. The sensitivity to sunlight and photophobia typical of PARD is also observed in both patients with Xeroderma Pigmentosum and Cockayne Syndrome, although to a lesser extent (23,24). Furthermore, the ataxic gait and telangiectasia seen in patients with PARD and Ataxia-Telangiectasia are similar (8,25). Like these DNA repair disorders, PARD appears to have a progressive clinical component. For instance, continuous neurodegeneration likely leads to cerebellar atrophy, as these symptoms are present in the two older patients. Similarly, progressive cerebellar ataxia is observed in patients with Ataxia-Telangiectasia (26). These similarities and clinical overlap with other DNA repair disorders underscore the importance of developing genetic testing to distinguish PARD from other DNA repair disorders.
Despite the central role played by PCNA in DNA replication, patients with PARD present with symptoms distinct from those of other DNA replication disorders. For instance, aside from the mild immunoglobulin deficiency seen in patients expressing PCNA-C148S, patients with PARD generally do not have significant immunologic dysfunction (9). This finding contrasts with the severe immunological symptoms observed in other DNA replication disorders such as immunodeficiency 96, which results from ligase 1 missense mutations, and immunodeficiency 54, which is a consequence of MCM4 truncation. These individuals suffer from extreme viral and bacterial infections due to extreme leukopenia (27)(28)(29). Similar immune dysfunctions were reported in two individuals with DNA polymerase δ mutations (30). In patients with PARD, the hypomorphic activity of PCNA may be sufficient to ensure normal immune function, particularly if DNA replication is not affected. However, patients with PARD may have compromised immunologic functions in settings of elevated body temperatures (e.g., during gestation stages or febrile episodes) and require further surveillance. Thus, it is critical to characterize the full spectrum of symptoms that may arise from mutations in replication proteins, despite their overlapping cellular functions.

Impact of the C148S substitution on PCNA stability
Our work provides new insight into how subtle mutations can profoundly affect PCNA stability. The C148 side chain makes a π-sulfur interaction with the neighboring Y144 (1). In general, πsulfur interactions provide considerable stability (0.5 kcal/ mol), comparable with what we observed for PCNA-C148S (17). In addition, converting a buried free cysteine to the more hydrophilic serine has been found to destabilize proteins (31)(32)(33). Thus, two nonmutually exclusive biophysical explanations exist for the stability defect observed in PCNA-C148S. We also observe a stability defect in PCNA-S228I, which is likely caused by the steric disruption of the partner binding cleft (9,10).
Subtle mutations that compromise trimer stability have been discovered in PCNA proteins across metazoan species (18)(19)(20). However, unlike most of the other reported mutations, the S228I and C148S substitutions are not directly involved in forming the interfaces between PCNA subunits. Therefore, our work demonstrates that substitutions in many regions of PCNA can potentially result in a stability defect and have harmful effects on human health. Further genetic surveillance will be necessary to reveal the full repertoire of mutations that can lead to PARD.

PCNA-C148S provides novel insight into PARD
Our study on PCNA-C148S provides new insight into the molecular basis behind PARD. PCNA-S228I has a partner binding defect that impairs interactions with specific partners (8)(9)(10)(11). This does not appear to be the primary defect underlying PARD, as PCNA-C148S appears to interact with its partners with similar affinity as PCNA-WT. However, our work reveals that the only shared phenotype between both PARD mutants is their remarkable thermostability defect (Fig. 5). Furthermore, these variants appear partially impaired at physiological body temperatures (Fig. S4A)   permissive enough to support human life. While our findings do not negate the idea that aberrant partner binding can play a role in PARD (particularly for the S228I variant), they reveal a potential second mechanism that likely underlies this disease: a stability defect leading to lower PCNA levels on chromatin.
Maintaining proper levels of chromatin-bound PCNA is critical for genome stability. During late G1 phase, chromatinbound PCNA levels increase in accordance with the DNA replication machinery (34). PCNA is removed from chromatin either through passive dissociation or actively by the ATAD5-RFC clamp unloader after the ligation of the nascent strands ( Fig. 1) (35, 36). This constant loading and unloading suggests that chromatin bound PCNA levels exist in a "Goldilocks zone" that must be properly maintained. Perturbances to this balance, such as failure to remove PCNA from the chromatin, leads to an increase in recombination and genome instability (20). Furthermore, specific post-translational modifications such as phosphorylation and acetylation can change the levels of chromatin-bound PCNA and result in promiscuous DNA repair activity (37)(38)(39). Thus, altering the balance of chromatin-bound PCNA can interfere with various metabolic and regulatory pathways and profoundly affect genome stability.
The thermostability defect of PARD variants lowers PCNA levels in all compartments, but especially chromatin-bound PCNA. The lower chromatin-bound fraction could be due to a decreased loading rate or an increased innate dissociation rate. A reduction in the lifetime of PCNA on chromatin may partially explain the differences these mutants have on different DNA metabolic pathways. For instance, DNA synthesis appears to be relatively unaffected in PARD (8), potentially because the replicative polymerases are thought to act almost immediately after PCNA loading (40). However, PCNA-dependent activities that occur later or more slowly, such as DNA repair, are likely more sensitive to PCNA's lifetime on DNA. The idea that unstable PCNA variants impair DNA repair is consistent with the finding that substitutions (C81R, E113G, E143K, D150E, and G178S) in yeast PCNA that reduce chromatin-bound PCNA levels cause defects in repair and the DNA damage response (18,19,21,41). Furthermore, degradation of PCNA in human cells leads to γH2AX signaling and Chk1/2 activation (42). Future experiments will test whether the PARD variants have defects in their residence time on DNA and impact the DNA damage response.

Potential therapeutic applications and prospects for PARD
Our biophysical characterization of PCNA-C148S opens new avenues for potential development of PARD therapeutics. Because both PARD variants have a thermostability defect, increasing their stability and/or lifetime on DNA may improve PCNA function. Small molecules that stabilize PCNA could be considered a therapeutic avenue. However, this strategy may be challenging because small molecules could interfere with secondary binding sites that are involved in partner interactions with PCNA. Indeed, many drugs that bind to sliding clamps interact with the partner binding cleft and block partner binding (2,43,44). An alternative therapeutic approach for PARD therapy is to increase the levels or lifetime of the chromatin-bound PCNA. This strategy could be implemented at the preventative care level by limiting the patients' exposure to conditions that dramatically impact their PCNA levels. For instance, UV-induced DNA damage stimulates the monoubiquitination of PCNA (45), which enhances the activity of error-prone DNA polymerases (46,47), and also triggers PCNA removal by ATAD5-RFC (48,49). Therefore, it may be advantageous from both a cellular and molecular level to limit the amount of direct sunlight these patients experience. If loss of chromatin-bound PCNA is the true driver of PARD, we envision that inhibiting the ATAD5-RFC clamp unloader may increase PCNA levels on DNA and alleviate PARD symptoms. In addition, the temperature sensitivity of PARD suggests that close temperature monitoring of patients and fever-reduction strategies may decrease disease progression.
Beyond therapeutics, our work highlights the need for increased global sequencing of the PCNA gene to understand how mutations can impact human health. The identification of a second disease-causing PCNA allele suggests that there are potentially other PARD-causing variants that have yet to be discovered. Identifying these variants could provide a deeper understanding of the cellular pathways compromised in patients with PARD. Furthermore, developing mouse models with each of the PARD-associated variants could illuminate the genetic pathways that lead to a pathophysiological phenotype. Ultimately, understanding how disease-causing variant impacts PCNA function will likely provide new mechanistic insight and lead to the development of effective therapies for PARD.

Study limitations
Although our clinical, cellular, structural, and biochemical data strongly suggest the disease relevance of PCNA-C148S, our study has several limitations. Our cellular experiments were not done with isogeneic pairs of cell lines, and we could not perform all cellular experiments with both primary patient cell lines. Owing to technical issues with rescue experiments and sample limitations, we acknowledge that differences in genetic backgrounds may contribute to the cellular phenotypes we observe. Despite these limitations, our data strongly support that these variants have a thermostability defect and that this impairs proper cellular function at elevated temperature. Further studies will be necessary to fully explore the molecular and cellular causes of this disease.

Molecular genetic analysis
All probands underwent whole-exome sequencing analyses according to previously published protocols (50,51). Briefly, we constructed an exome library using SureSelect Human All Exon V6 Kit (Agilent Technologies) following manufacturer's instructions. We then sequenced the library on a HiSeq 2500 platform (Illumina) using HiSeq SBS V4 cluster generation and sequencing kit (Illumina) run on paired-end mode. Reads were aligned to the hg19 assembly of the human genome using the bwa-mem aligner (52). Duplicate reads were flagged with the bammark duplicates tool from biobambam2, which is publicly available through the GNU GENERAL PUBLIC LICENSE Version 3. Variant calling was performed with Freebayes, and the resulting Variant Calling Format files were annotated with ANNOVAR (53). The family pedigrees revealed consanguinity among its members; thus, we searched the exome data for the presence of homozygous variants in all patients. Such variants were absent in our in-house sequencing data sets and in public databases (gnomAD, http://gnomad.broadinstitute.org/, and Abraom http://abraom.ib.usp.br/) (54,55). Silent mutations were excluded. The assessment of gene function was performed using the Online Mendelian Inheritance in Man (OMIM) and the PubMed databases. Sanger sequencing was performed to validate the candidate variant identified. The pathogenicity of the C148S variant was then scored using several in silico programs: SIFT, PolyPehn2, PROVEAN, CADD, MUtation Assessor, and REVEL (56)(57)(58)(59)(60)(61). PCNA sequence alignment was performed using both ClusterOmega and Aminode (62,63).
Cell viability with UV irradiation, zeocin treatment, and gamma-ray irradiation Cells (5.0 x 10 4 cells) were seeded in 96-well plates 2 days prior to treatment and were washed with preheated PBS prior to UV light (260 nm), gamma (γ) ray, or zeocin treatment. UV dose rates were monitored using a VLX-3W radiometer, 0.1 J/ m 2 /s for low-dose exposures and 0.74 J/m 2 /s for high doses. Unirradiated cells were maintained in PBS for the same time as their irradiated counterparts. For γ-ray irradiation, cells were washed with PBS supplemented with 890 μM CaCl 2 and 500 μM MgCl 2 and subjected to 1.355 Gy/min using an IBL 637 Cesium-137γ-ray. After UV and γ-ray treatments, cells were incubated with fresh medium for 72 h. For zeocine treatment (Invitrogen, ThermoFisher), cells were incubated Thermosensitive PCNA variants underlie a rare disorder with medium supplemented with various concentrations for 72 h. Cell viability was performed using the Cellular Proliferation Kit II (XTT, Roche) according to the manufacturer's recommendations. Cell metabolism and viability were assessed in triplicate by the 492 and 650 nm ratio.

Wound healing assay
Cells were seeded to 90 to 100% confluency at 37 C 24 h before the experiment. The monolayers of cells were then scraped with a sterile pipette tip to generate an open wound. Cells were then shifted to either 37 C or 42 C for 48 h to allow wound repair. We obtained brightfield images using an EVOS XL Core System microscope (ThermoFisher), and the area of the wound was processed in ImageJ (67).

Protein expression and purification PCNA
All PCNA variants were expressed and purified as described (9). Briefly, PCNA variants were expressed from a pET3c vector in Escherichia coli BLR-DE3 that was grown in 2XYT supplemented with 100 μg/ml of ampicillin at 37 C. Once cells reached an A 600 between 0.6 and 0.8 they were induced with 1 mM IPTG overnight at 18 C. One-liter cultures were centrifuged at 4000g and resuspended in Buffer A (25 mM Tris [pH 7.5], 10% [vol/vol] glycerol, and 2 mM DTT) and lysed via cell disruptor (Microfluidics Inc). Lysates were then loaded onto sequential S and 2 × Q 5 ml HiTrap columns (GE Healthcare) that were pre-equilibrated with buffer A. S-columns were removed prior to two-column volume washes with Buffer A. PCNA was eluted using a gradient of Buffer B

Crystallization and structural determination
PCNA-C148S was crystallized by the hanging-drop method. For crystallization, 1 μl of protein (15 mg/ml) was mixed with an equal volume of well solution (175 mM magnesium acetate and 20% PEG 3350). Crystals appeared after 3 to 7 days at room temperature. Crystals were briefly swiped through Paratone N cryo-protectant and frozen at 100 K in a cryostream. Crystallographic diffraction data were collected on a Rigaku system with a Saturn 944 CCD detector. Indexing, integration, and scaling were performed with HKL3000, and the structure was solved via molecular replacement using PHASER with the partner binding region (residues 116:133) of PCNA-S228I deleted as a search model (PDB: 5E0T) (9, 69). Refinement and model building was carried out using phenix.refine (70) and Coot (71). Model statistics are as defined by phenix.refine and summarized in Table S3.
FEN1 in vitro pull-downs FEN1 was expressed in E. coli and the lysate was coupled to cobalt resin as described above. After FEN1 coupling, the resin was washed with 5 × 5 column volumes of Nickel Wash Buffer. The FEN1-coupled resin was then incubated for 1 h with 5 μM PCNA-WT, C148S, or S228I at 4 C. The resin was then washed again with 5 x 5 column volumes of Nickel Wash Buffer to remove excess PCNA. PCNA and FEN1 were then eluted with 0.5 column volumes with Nickel Elution Buffer. Samples were separated on a 12% SDS-PAGE gel and stained with Coomassie. Densitometry was conducted in ImageJ.

Isothermal titration calorimetry
Isothermal titration calorimetry conditions were similar to those previously used to determine peptide binding thermodynamics to PCNA-WT and S228I (9).

Thermal shift assay
We assessed intrinsic tryptophan fluorescence at increasing temperature conditions to obtain a relative T m . PCNA was diluted to various conditions (200 nM or 1 μM) in either PCNA Gel Filtration Buffer prepared using with 25 mM Tris-HCl (pH 7.5) or Hepes (pH 7.5). Tryptophan fluorescence was measured using a Fluoromax 4 Spectrofluorometer (Horiba Scientific). Typical parameters were excitation: 280 nm (slit width 2 nm) and emission: 325 nm (slit width 5 nm), integration time (0.1 s), temperature range 25 to 60 or 70 C; equilibration times of 5 min at each temperature. Fluorescence values were normalized with Y= F -F U /F N -F U where F is the raw fluorescence signal and F U and F N are the theoretically calculated fluorescence signals at each temperature from the unfolded and native states, respectively. F N and F U values were determined from the linear relationship of the first and last five calculated values, respectively. The melting point and thermodynamic values were determined from a Boltzmann sigmoidal curve. To assess the kinetics of decay, PCNA was incubated at a constant temperature for 24 h and monitored every 2 min. Samples were normalized to the first time point and fit with a two-phase decay.

Native gel assay
The stability of the PCNA trimer was determined by using a native gel assay (18). All PCNA variants were incubated for 24 h at 4 C, 37 C, or 42 C. Various concentrations of each variant were then loaded on a 4 to 20% gradient PAGE gel (Bio-Rad, 4561096). Samples were incubated at 4 C for 1 h in Native Gel Buffer (192 mM glycine and 25 mM Tris base) and run for 2.5 h at 100 mV. Gels were then stained with Coomassie.

Equilibrium unfolding experiments
The stability of PCNA-WT and C148S was assessed by circular dichroism (CD) in the presence of incremental guanidinium hydrochloride (Gdm-HCl) concentrations. PCNA was dialyzed overnight at 4 C in 150 mM KPO 4 buffer (pH 7.0) with 1 mM TCEP. Samples (1 μM) were then titrated with different Gdm-HCl concentrations (0-6 M) in 150 mM KPO 4 buffer (pH 7.0) with 1 mM TCEP buffer and allowed to equilibrate for 5 days. CD signals were collected using a Jasco-810 spectropolarimeter (Jasco, Inc) equipped with temperature control system. Equilibrium unfolding monitored from 215 to 260 nm was collected for each sample in a 2-nm cuvette at 25 C with 25 nm bandwidth. CD signals were plotted at 218 nm versus [GdmHCl]. Data were fit with Savuka (72) using a twostate mode (73). The tertiary structure was measured by monitoring intrinsic tryptophan fluorescence using the same samples (excitation: 295 nm [bandpass 2 nm]; emission: 325 nm [bandpass 5 nm]).

ATPase activity assay
DNA-mediated ATPase activity of RFC was measured at room temperature using a previously published coupledenzyme ATPase assay (11,74). Reactions mixtures containing 1 μM PCNA and 150 nM RFC were incubated with a master mix (3U/ml Pyruvate kinase, 3 U/ml Lactate dehydrogenase, 1 mM ATP 670 nM Phosphoenol pyruvate, 170 nM NADH, 50 mM Tris [pH 7.5], 5 mM MgCl 2 , and 400 mM Potassium glutamate) and 1 μM annealed oligos. PCNA concentrations were titrated from 0 to 600 nM for all affinity studies. For all the half-life experiments, PCNA (1 μM) was incubated at the designated temperature for various time points prior to the experiment and diluted to a final concentration of 150 nM in the reaction. All reactions were performed at 25 C with reagents equilibrated for at least 10 min. Absorbance was measured in 96-well plates using an excitation filter at 355 nm and a bandpass of 40 nm. Oligos used in the reactions were 5 0 -TTTTTTTTTTTATGTACTC GTAGTG TCTGC-3 0 and 5 0 -GCAGACACTACGAGTACATA-3 0 .

Preparation of bead-based DNA template and loading assay
We determined the kinetics of PCNA loading by a beadbased PCNA loading assay. ssM13 DNA (7 kbp; New England BioLabs) was coupled to streptavidin beads (Ther-moFisher Dynabeads kilobaseBINDER Kit, 60101) overnight at room temperature using biotinylated oligos following manufacturer's recommendations. Beads were resuspended in autoclaved water, and coupling efficiency was determined by PCR. Loading reaction occurred in a 1× Loading Buffer (50 mM Hepes, 4% glycerol, 0.01% NP-40, 250 mM Potassium Glutamate, 5 mM MgCl and 0.5 mM TCEP) with 50 ng of ssM13, 1 μM PCNA, 150 nM RFC, and 2 mM of various nucleotides. For elevated temperature experiments, PCNA was preincubated at 42 C for 24 h prior to loading. Loading reactions were conducted at 25 C for times indicated and quenched by adding 25 mM EDTA and chilling samples on ice. Following two washes with Loading buffer, we eluted all bound protein with Laemmli buffer. PCNA was detected by Western blot using an anti-PCNA antibody (Abcam, ab29) and quantified by ImageJ (67).

Whole-cell and cell fractionation western blots
Cells were grown till 70 to 80% confluency in 10-or 25-cm dishes, as described above. Cells were harvested and washed with 1× PBS and stored at −80 C until lysis. Whole-cell extracts were prepared by lysing cells in RIPA buffer (Cold Spring Harbor protocol) for 30 min with periodic vortexing. Cytosolic and nuclear extracts were prepared by using NE-PER Nuclear and Cytoplasmic Extraction Reagents (ThermoFisher, 78833) as described by the manufacturer. Chromatin fractions were prepared by sonicating the nuclear pellet in RIPA buffer for 15 min at 4 C. Proteins were detected using the following antibodies: PCNA (Abcam, ab29), Tubulin (Cell Signaling, 2146S), and H2B (Cell Signaling, D2H6). Protein levels were determined in three biological replicates by using ImageJ and compared with expression in HDCs.

Protein half-life experiments
To determine the half-life of each PCNA variant, cycloheximide pulse-chase experiments were performed. Cells were incubated with either dimethyl sulfoxide or 500 μM of cycloheximide (MP Biomedicals) at 37 C, 5% CO2 for 24 h. Following lysis with RIPA buffer, we assessed PCNA levels by Western blot (Abcam, ab29) and measured by densitometry, as above. Values were normalized to each loading control and then normalized against the dimethyl sulfoxide vehicle. Results represent the average of three biological replicates.

Flow cytometric γH2AX signaling
Exponentially growing cells were plated the day before differential temperature treatment and incubated at 37 C. Cells were separated to different temperatures (37 or 42 C) for 24 h. At the indicated time points, cells were collected as described (75). Briefly, cells were fixed on ice with 1% formaldehyde and storage at −20 C in 70% ethanol. Cells were permeabilized and blocked with BSA-T buffer (0.2% Triton X-100 [Sigma-Aldrich] and 1% BSA in PBS). Cells were then incubated with 1/5000 anti-γH2AX (05-636 Millipore) overnight at 4 C. Samples were incubated overnight at 4 C with 1/200 FITC anti-mouse secondary antibody (Sigma-Aldrich). Cells were then stained with propidium iodide solution (PI, 20 μg/ml, 200 μg/ml RNase A, Invitrogen, Life Technologies, and 0.1% Triton X-100) in the dark and at room temperature for 30 min.
Cells were processed on a BD AccuriTM C6 (BD) type cytometer and analyzed using the BD CSamplerTM Analysis software. More than 10,000 events were obtained for the analysis of each sample.

Statistics
In experimental data, error bars represent standard deviation of 2 to 6 independent experiments as indicated in the figure legends. All analyses were done in GraphPad Prism with standard t test, multiple t test comparison, or one-way ANOVAs as indicated in the figure legend. p Values are indicated as follows: <0.05 (*); <0.01 (**); <0.001 (***) <0.0001 (****).

Study approval
This study was approved by the Research Ethics Committee of the Hospital das Clinicas da Faculdade de Medicina da Universidade de of Sao Paulo (Approval number: 37868114.3.0000.0068; January 27, 2015). All the human studies in this article abide by the Declaration of Helsinki principles. All patients' guardians gave written informed consent prior to initiating the genetics studies. Written consent was obtained from the patients' guardians to publish photographs of affected individuals. DNA samples were extracted from peripheral blood leukocytes using standard procedures.

Data availability
All data within this article is within the main text and supplemental. The PDB and map files are deposited in the PDB database (PDB: 8E84).
Supporting information-This article contains supporting information.
responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.
Abbreviations-The abbreviations used are: HDC, healthy donor cell; PARD, PCNA-associated DNA repair disorder; PCNA, proliferating cell nuclear antigen; RFC, replication factor C.