Constitutional POLE variants causing a phenotype reminiscent of constitutional mismatch repair deficiency

Heterozygous POLE or POLD1 germline pathogenic variants (PVs) cause polymerase proofreading associated polyposis (PPAP), a constitutional polymerase proofreading deficiency that typically presents with colorectal adenomas and carcinomas in adulthood. Constitutional mismatch‐repair deficiency (CMMRD), caused by germline bi‐allelic PVs affecting one of four MMR genes, results in a high propensity for the hematological, brain, intestinal tract, and other malignancies in childhood. Nonmalignant clinical features, such as skin pigmentation alterations, are found in nearly all CMMRD patients and are important diagnostic markers. Here, we excluded CMMRD in three cancer patients with highly suspect clinical phenotypes but identified in each a constitutional heterozygous POLE PV. These, and two additional POLE PVs identified in published CMMRD‐like patients, have not previously been reported as germline PVs despite all being well‐known somatic mutations in hyper‐mutated tumors. Together, these five cases show that specific POLE PVs may have a stronger “mutator” effect than known PPAP‐associated POLE PVs and may cause a CMMRD‐like phenotype distinct from PPAP. The common underlying mechanism, that is, a constitutional replication error repair defect, and a similar tumor spectrum provide a good rationale for monitoring these patients with a severe constitutional polymerase proofreading deficiency according to protocols proposed for CMMRD.


| INTRODUCTION
Three distinct mechanisms ensure accurate DNA replication before cell division. The DNA polymerases ε and δ (Pol ε, Pol δ) ensure high replication fidelity by the base selectivity of their active sites as well as their 3′ to 5′ exonuclease proofreading activity, which detects and excises misincorporated nucleotides. Post-replication, the DNA mismatch repair (MMR) system corrects errors that escape proofreading (Burgers & Kunkel, 2017;Ganai & Johansson, 2016;Jiricny, 2006;Rayner et al., 2016). Compromised polymerase proofreading and MMR deficiency have been observed in a variety of tumor types.
MMR deficiency results from biallelic inactivation of one of four MMR genes (MLH1, MIM# 120436; MSH2, MIM# 609309; MSH6, MIM# 600678; PMS2, MIM# 600259) and is associated with the well-described mutator phenotype of microsatellite instability (MSI), an increased frequency of insertion and deletion mutations (indels) in short tandem repeats. MMR deficiency is observed in approximately 15% of colorectal and 26% of endometrial cancers, and at a lower frequency in other tumors (Ryan et al., 2019;Vilar & Gruber, 2010).
Although typically observed in neoplastic cells, constitutional MMR deficiency (CMMRD) and constitutional polymerase proofreading deficiency, called polymerase proofreading associated polyposis (PPAP), are rare hereditary cancer predisposition syndromes Wimmer et al., 2014). Over 200 CMMRD patients, most often caused by germline biallelic PVs in PMS2 or MSH6, have been published, and CMMRD is now a recognized recessively inherited, childhood cancer syndrome (MIM# 276300). The tumor spectrum includes hematological malignancies and brain tumors, as well as colorectal adenomas and carcinomas, and other Lynch syndrome-associated cancers. The median age at diagnosis of the first malignancy is <10 years . Nearly all CMMRD patients have characteristic, nonmalignant clinical features.
Of these, multiple café-au-lait macules (CALMs) and other alterations of skin pigmentation are the most prevalent (Wimmer, Rosenbaum et al., 2017). The Care for CMMRD (C4CMMRD) consortium has integrated these features into a clinical scoring system to guide diagnostic workup for CMMRD in pediatric and young adult cancer patients .
Less than a decade ago, it was shown that germline PVs in the exonuclease domains of POLE and POLD1 cause PPAP, a dominantly inherited colorectal polyposis and cancer predisposition syndrome . Up to now, some 149 individuals from 55 unrelated families with PPAP have been reported (Table S1). Approximately half of these families (n = 28) have the recurrent POLE PV p.Leu424Val. Endometrial, ovarian, pancreatic, and breast cancers, as well as high-grade gliomas, have also been reported in patients with PPAP. The median age at the first malignancy was 43 years (range: 16-67 years) in 122 individuals carrying a POLE PV and was 41 years (range 21-64 years) in 25 individuals with a POLD1 PV.
Recently, two pediatric cancer patients have been reported with a clinical presentation highly suggestive of CMMRD but caused by constitutional POLE PVs (Lindsay et al., 2019;Wimmer, Beilken et al., 2017). Interestingly, the constitutional POLE PVs identified in these cases have not been reported to cause PPAP despite being known somatic mutations in hyper-mutated tumors. Here, we report three further cases with a CMMRD-like phenotype caused by constitutional POLE PVs not associated with PPAP but previously found as somatic mutations in ultra-mutated tumors (TMB >100 mut/Mb).

| Editorial policies and ethical consideration
All clinical and genetic analyses of the cases described here were performed as part of their diagnostic work-up. The patients/or their legal representative gave informed consent to all analyses and written informed consent to germline genetic testing. Written informed consent has been obtained from all patients and/or their legal representatives to publish clinical information and pictures of their skin pigmentation alterations.

| Tumor analysis
Tumor tissue fixation and sectioning, histology, and immunohistochemistry (IHC), as well as RNA extraction, DNA extraction, and MSI analysis, all followed standard diagnostic practices (please see Supporting Information for details).
Tumor DNA-methylation analysis used the Illumina Infinium EPIC/ 850 k BeadChip array (Illumina) according to the manufacturer′s instructions and using 250 ng of sample DNA. The unprocessed IDAT files were uploaded to the DKFZ/Heidelberg neuro classifier (https://www. molecularneuropathology.org/mnp) and subsequently compared to methylation data of a reference cohort in the Neuro methylation classifier v.11b4. Matching is obtained if the score is >0.9.
For whole-exome comparative tumor sequencing, germline and tumor DNA (500 ng) were fragmented to 300 bp using Covaris S2 (Agilent), and adaptor ligation was performed on a Sciclone G3 (Perkin Elmer) using

KAPA HTP Library Preparation Kit (Roche). Exomes were enriched with
SureSelectXT Clinical Research Exome kit (Agilent). Somatic singlenucleotide variants and small indels were called using GATK 4.1.7 best practices (Van der Auwera & O′Connor, 2020). In brief, sequencing reads were trimmed using bbduk v.38.26 (https://sourceforge.net/projects/ bbmap/) and aligned to the human reference genome (hg19/GRCh37) using BWA v.0.7.15 mem (Li & Durbin, 2009). Aligned reads were filtered using GATK PrintReads and base quality scores were recalibrated using Mutational signatures were explored using the latest (as of June 2021) version of the R/Bioconductor package MutationalPatterns (Blokzijl et al., 2018). Briefly, count matrices were derived for all three different types of mutations (single base substitutions, double base substitutions, and indels) and their cosine similarity to the mutational signatures of the COSMIC reference signature collection v3.2 (Alexandrov et al., 2020) was calculated. COSMIC cosine similarity scores were normalized against a background cohort of 95 somatic whole-genome sequencing samples to account for distinct inter-signature cosine similarity distributions: Specifically, each cosine similarity was normalized by dividing the difference between them and the median cosine similarity by the interquartile range of cosine similarity values for that signature. For presentation, signatures were grouped by etiology. CMMRD-negative genetic diagnoses were confirmed using a sequencing-based MSI assay that can detect the low-frequency microsatellite length variants in DNA extracted from non-neoplastic PBLs of patients with CMMRD (Gallon et al., 2019). The assay was performed with adaptions as described in (Perez-Valencia et al., 2020), and an MSI score >2.00 was used to classify a patient as CMMRD-positive.

| RESULTS
An extensive description of the clinical course and follow-up of each case reported here can be found in Supporting Information.

| Case 1
Case 1 presented at the age of 4 years and 4 months with a large tumor (5.1 × 3.8 × 3.1 cm) in the right cerebellar hemisphere. On physical examination, the patient had one hypo-and multiple hyperpigmented skin spots distributed over the entire body (Figure 1a-d). More than six were reminiscent of CALMs and had a diameter over 0.5 cm. CALMs were absent in the non-related parents and in two healthy older siblings. Both parents were healthy with no history of cancer. A paternal aunt was diagnosed with colon cancer at the age of 30 years. The maternal grandfather died at the age of 35 years with leukemia (diagnosed at 29 years of age) and the maternal grandmother had Hodgkins' lymphoma at the age of 65 years and non-Hodgkins lymphoma at the age of 66 years. The patient scored at least four points according to C4CMMRD criteria (Table 1) and, hence, was further evaluated for CMMRD.
Pathological analysis of the brain tumor, including genome-wide methylation (850K) analysis, using the Heidelberg classifier, showed it to be a WHO grade IV anaplastic medulloblastoma of subtype SHH (score 0.97) and subclass SHH A (score 0.82), with desmoplastic/ nodular features and without amplification of MYC or MYCN

| Case 2
Case 2 was diagnosed with IDH1-wildtype glioblastoma (WHO IV) at the age of 29 years. At the age of 20 years, the patient had been diagnosed with a colorectal adenocarcinoma at the splenic flexure (G2 pT2), and multiple (50-60) colorectal adenomas (mainly of a tubulovillous histology, and several showed high-grade dysplasia). At the age of 28 years, the patient had a calcifying epithelioma/pilomatrixoma (0.5 × 1.5 × 1.4 cm) removed from his upper arm ( Figure 1g,h). The patient reported that a similar cyst was removed in childhood. Physical examination of the patient at the age of 30 years showed multiple café au lait-colored skin spots distributed over the entire body and several skin papules and cysts on his forehead and trunk (Figure 1e,f). Their family history revealed that the maternal grandmother and a maternal great-uncle had colorectal cancer in their 70s, but no further cancer diagnoses were reported. Taken together, at the time of glioblastoma diagnosis the patient scored at least nine points according to C4CMMRD criteria (Table 1)

| Case 3
Case 3 presented at 13 years of age with a rectal mass of 6.0 × 5.0 × 2.0 cm and numerous (>100) adenomatous polyps in the colon, the largest measuring 3.0 × 2.0 × 1.9 cm (Figure 1j). The final staging of the rectal tumor was T3N1M0, consistent with stage IIIB rectal carcinoma. On physical examination, the patient showed several CALMs over the legs, abdomen, back and one on her face ( Figure 1i). A review of her records revealed that she had been followed in the neurosurgery clinic due to the incidental discovery of an unbiopsied tectal plate glioma two years before her diagnosis of rectal cancer (Figure 1k). During the time of her evaluation for colorectal cancer, she was also seen by an orthopedic surgeon for a 10-year-long history of a right shoulder mass that was surgically removed and found to be an intramuscular venous malformation ( Figure 1l). The family history of this only child was reviewed in detail.
The paternal grandmother had breast cancer diagnosed in her seventies or eighties, but there were no cases of colorectal cancer or young-onset tumor. According to C4CMMRD criteria, the patient scored 8 points (Table 1)  affects a highly conserved amino acid in the Polε exonuclease domain and has been described as a somatic mutation consistently associated with a high TMB (Campbell et al., 2017;Johnson et al., 2017) with the typical mutational signature of Pol ε proofreading deficiency.
Retrospectively performed exome sequencing revealed a TMB of 169 mut/Mb for the rectal carcinoma and a high cosine similarity to the Pol ε proofreading deficiency SBS signature 10b ( Figure S3). T A B L E 1 (Continued) Germline DNA analysis showed that this POLE variant is absent in the healthy mother. The father was not available for testing.

| DISCUSSION
The three cases described here add to two published case reports (Lindsay et al., 2019;Wimmer, Beilken, et al., 2017)  Analysis of POLD1 should also be considered as, like POLE, variants have been found as somatic driver mutations in hyper-mutated tumors and in the germline of PPAP patients (Table S1). Whilst the causative POLE PVs were confirmed to be de novo in two of these five cases (the father was not available for genetic testing in the other three), offering predictive testing of siblings may be advisable to take ciency. This should be taken into account when performing genome or exome sequencing of tumors to direct treatment and identification of germline genetic defects as is advocated for pediatric brain tumors (Campbell et al., 2017;Gröbner et al., 2018). The clinical and molecular similarities between CMMRD and these patients, as well as a common constitutional defect in replication error repair, also provide a good rationale for these patients to be monitored according to protocols proposed for CMMRD (Durno et al., 2017;Tabori et al., 2017;Vasen et al., 2014).
The phenotypes of these five cases appear to be distinct from and more severe than PPAP. For example, although Cases 2 and 3, and the first reported case in the literature (Wimmer, Beilken et al., 2017) had colorectal tumors consistent with a polyposis phenotype, these were diagnosed at exceptionally young ages. Furthermore, Case 1 and the case reported by Lindsay et al. (2019) were diagnosed with a pediatric medulloblastoma, which does not fall into the known tumor spectrum of PPAP but is seen in approximately 5% of CMMRD patients (Wimmer, Rosenbaum et al., 2017). All of the cases reported here have other features, such as CALMs and pilomatrixomas used in the C4CMMRD scoring system to identify potential CMMRD patients. Case 2 also had an intramuscular venous  Table S1).
This suggests that these POLE PVs may have an exceptional constitutional penetrance.
Indeed, in vivo and in vitro yeast experiments have shown that different polymerase exonuclease domain variants equivalent to PVs in human Pol ε can have very different impacts on replication mutation rate (Barbari et al., 2018;Hamzaoui et al., 2020;Xing et al., 2019). Of note, the highest mutation rates exceeded those observed in exonuclease domain negative yeast strains, suggesting that mutation rate is not solely determined by exonuclease activity (Hamzaoui et al., 2020;Xing et al., 2019 Functional data on the POLE PVs identified in the three cases with CMMRD-like phenotypes described here, and the two previous cases from the literature (Lindsay et al., 2019;Wimmer, Beilken et al., 2017), are limited and, as the results of Galati et al. (2020) and others (Barbari et al., 2018;Hamzaoui et al., 2020;Mur et al., 2020) show, in vitro assays should be interpreted with caution due to different impacts of equivalent polymerase variants between humans and model organisms. However, the functional impact of these and some PPAP-associated variants have been assessed using human tumor sequence data, allowing the comparison of mutation rates between them (Table S1). Using a tumor mutational signature-based  Table S1). POLE p.(Pro436Ser) (P436S) was found as both a somatic variant in hyper-mutated tumors and as a PPAP-associated germline variant. Arrows and red coloration indicate which POLE variants were found in patients with a CMMRD-like phenotype suggest that PPAP-associated POLE PVs may have a weaker "mutator" effect than those found only in hyper-or ultra-mutated tumors and in the CMMRD-like cases described here. However, it should be noted that assessing variant impact using tumor sequence analysis is complicated by the natural history of the tumors: Tumors may have multiple polymerase variants and other endogenous or exogenous exposures, such as MMR deficiency, that influence mutation rate and signature (Campbell et al., 2017). Furthermore, PPAP-associated variants may be predominantly associated with hyper-mutated tumors, such as POLE p.(Pro436Ser) (Hamzaoui et al., 2020;Spier et al., 2015) in the study of Campbell et al. (2017), where all (5/5) tumors carrying this variant were hyper-mutated ( Figure 3 and Table S1). Therefore, identification of more patients and the development of novel functional assays and models may be needed to clarify the genotype-phenotype relationship of constitutional POLE variants.
In the literature, we found additional patients with constitutional POLE PVs and potentially CMMRD-like phenotypes ( Figure 3 and  (Hamzaoui et al., 2020).
However, these three patients have PPAP pedigrees, and their POLE variants have not been observed in tumor analyses or were found in isolated tumors making interpretation of their functional impact uncertain ( Figure 3 and Table S1). Therefore, they may represent extremes of the PPAP phenotypic spectrum, possibly caused by genetic or environmental modifiers. Another patient carrying a POLE p.(Tyr458Asn) variant was diagnosed with a colorectal cancer aged 25 years and has an affected daughter diagnosed with multiple colorectal adenomas at age 17 years (Shickh et al., 2021). Despite not raising suspicion for CMMRD according to C4CMMRD criteria, these are very early-onset phenotypes. POLE p.(Tyr458Asn) has not been described in tumors or in a PPAP family to our knowledge ( Figure 3 and Table S1), hence this father and his daughter may also fit the suggested genotype-phenotype relationship of POLE variants.
In conclusion, the three cases presented here, and two previously published case reports (Lindsay et al., 2019;Wimmer, Beilken et al., 2017) may represent a distinct pediatric cancer syndrome caused by constitutional polymerase proofreading PVs that are not associated with PPAP but are associated with hyper-mutated tumors.
Regardless, the clinical similarities of these cases to CMMRD suggest constitutional polymerase proofreading deficiency should be considered as a differential diagnosis to CMMRD, and these patients should, based on the known phenotype thus far, be managed according to clinical guidelines for CMMRD.