The SPS cohort comprised 16 families including 39 patients (≥2 patients per family) diagnosed with SPS and fulfilling the 2010 World Health Organization (WHO) criteria (Snover et al., 2010), as the new WHO guidelines released in 2019 (Rosty et al., 2019) were not available when this study was developed. The complete clinical and somatic characterization of this cohort is available at (Soares de Lima et al., 2021). The presence of germline alterations in APC, MUTYH and DNA mismatch repair (MMR) genes was discarded for all probands.

One patient (AA3531, family SPS.7, Figure 1) presented a loss-of-function variant in the POLD1 gene (c.1941delG; p.Lys648fs*46). The variant co-segregated in other six family members, three of them affected with SPS (Figure 1; Supplementary Figure S1). A summary of clinical characteristics of family SPS.7 is shown in Table 1.

The index case (III.1) was affected with 2 synchronous CRC at age 57 as well as more than 100 serrated polyps, and important proportion of them having a large size (>20 mm). Individual III.5 was diagnosed at 46 y. o. with MALT lymphoma (affecting gastrointestinal tract, breast and lung). A familial history of cancer was present with a paternal grandfather affected with stomach cancer at 48 y. o. and a maternal aunt diagnosed with cancer of unknown origin at 78 y. o.

Regarding the phenotype in the affected siblings, III.2 presented 11 polyps at 56–65 y. o., corresponding to one hyperplastic polyp in the rectum (3 mm), five sessile serrated lesions proximal to the rectum (5–8 mm) and 5 T/LGD (tubular, low-grade dysplasia) adenomas (4–8 mm) distributed all over her colon. III.3 presented 20 polyps <1 cm at 53–60 y. o., including 5 serrated polyps proximal to the rectum (5–6 mm), being four hyperplastic polyps and one sessile serrated lesion without dysplasia, and 15 T/LGD adenomas (<1 cm). III.4 presented 11 polyps at 49–57 y. o., comprising four serrated polyps being two sessile serrated lesions proximal to the rectum (>5 mm, <1 cm), and 7 T/LGD adenomas (one 1 cm at rectum, the rest <1 cm).

The study received the approval of Hospital Clínic de Barcelona Clinical Research Ethics committee (registration number 2013/8286). Written informed consent was obtained in all cases.

Details on germline and tumoral whole-exome sequencing, quality control and alignment, variant calling and variant annotation have been already described in (Soares de Lima et al., 2021).

Germline DNA and cell lines’ DNA was extracted using the QIAamp DNA Blood kit (Qiagen, Redwood City, CA, United States). Somatic DNA was obtained from formalin-fixed paraffin-embedded tissue using the QIAamp Tissue kit (Qiagen, Redwood City, CA, United States). POLD1 variant validation was performed by PCR amplification using the GC-Rich PCR System (Roche, Basel, Switzerland) followed by Sanger sequencing (Eurofins Genomics).

Whole blood was collected into PAXgene Blood RNA tubes (PreAnalytiX, Hombrechtikon, Switzerland), and automated purification of total RNA was performed using the QIAcube and the PAXgene Blood RNA Kit (Qiagen), according to the manufacturer’s protocol.

Total RNA from cultured cells was isolated with the RNeasy Mini Kit, according to the manufacturer’s instructions (Qiagen, Hilden, Germany).

Peripheral Blood Mononuclear Cells (PBMCs) were isolated from whole blood by density gradient centrifugation (Ficoll^® Paque, GE Healthcare Life Sciences). Ten ml of whole blood were mixed 1:1 with PBS, layered on top of 3 ml of density gradient media, and separated by centrifugation at 1800 rpm for 25 min at room temperature with the brake turned off. After recovering the buffy coat, PBMCs were washed three times with cold PBS. Pelleted PBMCs were finally cryopreserved for subsequent protein expression testing.

RNA reverse transcription was performed with the High-Capacity cDNA reverse Transcription kit (Applied Biosystems). Quantitative PCR was run on a QuantStudio1 System (Applied Biosystems) by using Taqman^® Gene Expression probes against POLD1-FAM (Hs01100821_m1) and GAPDH-VIC (4326317E), the latter for normalization purposes. Relative quantification was performed with the –∆∆Ct method.

Pelleted PBMCs or cultured cells were lysed with RIPA buffer solution (Sigma-Aldrich, MA, United States) supplemented with cOmplete Protease Inhibitor Cocktail and PhosSTOP (Roche, Basel, Switzerland).

Protein extracts were run in NuPAGE™ gels according to the manufacturer’s protocol (ThermoFisher, Waltham, MA) and transferred into PVDF membranes (Millipore, Bedford, MA). Blots were probed with anti-POLD1 (EPR15118, #ab186407, Abcam) or anti-GAPDH (clone 14C10, #2118, Cell Signaling) primary antibodies diluted 1:5000, followed by the incubation with the fluorescent Dylight 800 anti-rabbit secondary antibody (SA5-10036). Protein detection was carried out using Odyssey Imaging System (LI-COR, Lincoln, NE).

Human colorectal SW837 cells (diploid and MMR-proficient, Cat No. 91031104, ECACC, Sigma Aldrich) were cultured in RPMI-1640 medium (Gibco, Waltham, MA) supplemented with 10% fetal bovine serum. Human HEK293T cells (Cat No. CRL-3216, ATCC) were cultured in DMEM medium (Gibco, Waltham, MA) supplemented with 10% fetal bovine serum. Both cell lines were maintained under standard growth conditions (37°C, 5% CO₂) and routinely tested for mycoplasma contamination using the Mycoplasma Gel Detection kit from Biotools (Madrid, Spain).

The Benchling (http://benchling.com) CRISPR tool was used to design suitable single guide RNAs (sgRNA) and homology-directed repair (HDR) templates flanking the POLD1 p.Lys648 region. Additionally, as a positive control, we designed a sgRNA and HDR template to model the POLD1 p.Leu474Pro mutation, a pathogenic founder mutation present in Spanish population (Accession ClinVar: VCV000144003.5) (Valle et al., 2014; Ferrer-Avargues et al., 2017).

sgRNA top and bottom strands were purchased from IDT (Coralville, IA) and cloned into the Esp3I site of the lentiCRISPRv2-Puro plasmid (#98290, Addgene), which also packages the Cas9 coding sequence. Each lentiCRISPRv2-POLD1 encoding vector was packaged into lentivirus by using the host cell line HEK293T and the CalPhos Mammalian Transfection kit (TakaraBio, Kusatsu, Japan). HEK293T cells were plated and co-transfected with at a 3:2:1 lentiCRISPRv2:psPAX2:pVSVG2 DNA ratio. Supernatants were harvested at two different time points (24 and 48 h after transfection), pooled, concentrated by centrifugation (15,000xg, 3 h, 4°C) and used for cell transduction in the presence of 8 μg/ml of polybrene. Infected cells were enriched by puromycin selection (1 μg/ml).

Cells with a stable expression of the sgRNA and the Cas9 protein were plated at a ratio of 400,000 cells per well in 12-well plates. Cells were transiently transfected in two consecutive rounds with a mixture of 10 pmol (500 ng) of the HDR template and 125 ng of an episomal vector for p53DD (#25989, Addgene) by using Lipofectamine 3000 reagent (ThermoFisher, Waltham, MA). The dominant negative p53DD inhibits the P53 double-strand break (DSB) response and was used to boost the DNA engineering process (Ihry et al., 2018). Cells were seeded into 96-well plates at a density of 1 cell per well and after 3 weeks, the obtained clones were screened for POLD1 gene editing by Sanger sequencing (Eurofins Genomics).

Cell viability was determined using the colorimetric CellTiter 96^® AQueous One Solution Cell Proliferation kit (Promega, Madison, WI). Either SW837 or SW837-POLD1 ^+/− cells were seeded in 96-well plates at a density of 2500 cells per well, in triplicate. After 4 days, 20 μl of CellTiter reagent was added to each well. Plates were incubated at 37°C for 3 h and absorbance was read at 490 nm wavelength using an Epoch Microplate Spectrophotometer (BioTek, Winooski, VT).

Single-cell suspensions were seeded at low density (400 cells per well into 12-well plates). After 16 days, colonies were fixed in cold methanol for 10 min and stained with a 0.5% crystal violet solution (Sigma Aldrich). After drying, plates were imaged on an EliSpot Reader System (AID GmbH, Strassberg, Germany) and analyzed with ImageJ (National Institutes of Health, Bethesda, MD).

The organoid cultures used in this study were established from normal rectum endoscopic biopsies derived from patient AA3531:III-1 (Figure 1) and a healthy donor, following already published procedures (Dotti et al., 2022).

Organoids were expanded in µ-Slide 8-well ibiTreat chambers (Ibidi, Fitchburg, WI) and treated with 200 nM camptothecin (CPT) for 24 h. As a control, some wells were left untreated. Phosphorylation of the Ser-139 residue of the histone variant H2AX (γH2AX) was assessed by immunofluorescence staining following already published protocols (Mayorgas et al., 2021), with some modifications. After the fixation step, organoids were stored in PBS overnight at 4°C, and the next day the permeabilization and blocking steps were performed as indicated. Samples were then incubated with anti-γH2AX (#ab81299, 1:750, Abcam) and EpCAM (#M0804, 1:150, Dako) primary antibodies overnight at 4°C. The next day, samples were incubated with the secondary antibodies anti-rabbit Alexa Fluor^® 594 and anti-mouse Alexa Fluor^® 488 (ThermoFisher, Waltham, MA), both diluted 1:500. After DAPI nuclear staining, samples were overlaid with Ibidi Mounting medium (#50001, Ibidi, Fitchburg, WI) and stored at 4°C for subsequent fluorescent microscope observation in a Zeiss LSM 880 confocal laser scanning microscope (CCiTUB optical microscopy facility, Universitat de Barcelona, Spain). Positive nuclei (γH2AX–DAPI colocalization) were counted by using the CellCounter plug-in in ImageJ software. The ratio of γH2AX-positive nuclei versus the total number of nuclei per organoid was calculated.

Organoids were expanded in 48-well plates and treated with 200 nM CPT for 24 h. After the genotoxic challenge, organoids were cultured during a 5-day resting period to allow cells to accumulate mutations. Organoids were recovered, and DNA extraction was performed in order to assess changes in their mutational profile, as already mentioned in section 2.4.4. Somatic variant calling was performed using MuTect2 (McKenna et al., 2010) and Strelka2 (Kim et al., 2018), by only considering those variants shared by both computational tools and showing an allelic frequency above 0.20.

Tumor MMRd testing was performed in both tumor samples from the proband. Immunostaining for the four MMR proteins confirmed loss of MLH1 and PMS2 (Figure 2A), and MSI molecular testing revealed that the tumor was indeed MSI-H, due to the alteration of three mononucleotide markers (BAT25, BAT26, Mono-27) (Figure 2B). One tumor also presented MLH1 promoter hypermethylation (25%), a likely cause for the observed MMR deficiency.

Germline DNA (AA3531) and somatic DNA (AA3547) from one MMRd tumor of the proband underwent whole-exome sequencing for the assessment of mutational signatures, tumor substitution mutational burden (TMB), and tumor indel mutational burden (IDB). The sample displayed the COSMIC clock-like signatures SBS1 and SBS5, and a high contribution of the already described MMRd-associated signatures SBS21, SBS26, SBS44, ID2, and ID7 (Figure 2C), in concordance with the molecular MMRd characterization of the tumor. Neither the mutational signatures associated with a defective POLD1 proofreading (SBS10c, SBS10d) nor that associated with concurrent POLD1 mutations and defective DNA mismatch repair (SBS20) were detected. However, the sample appeared to be ultra-hypermutated (TMB = 117.46 mut/Mb, IDB = 100.6 mut/Mb), which is characteristic from combined mismatch-repair deficiency and polymerase alterations.

Additionally, LOH analysis was performed by Sanger sequencing and the detection of microsatellite markers flanking POLD1. The analysis revealed a LOH of the wild-type allele (Figures 2D,E).

The segregation analysis revealed that the POLD1 variant c.1941delG, p.(Lys648fs*46) segregated in six additional members of the family beside the index case: four siblings (three fulfilling SPS 2010’s criteria and one diagnosed with lymphoma at age of 46) and her two healthy daughters (Figure 1; Supplementary Figure S1). The variant is located in the polymerase domain of POLD1 and it causes a frameshift, which changes a Lysine to an Arginine at codon 648, and a premature stop codon is predicted at position 46 of the new reading frame (Supplementary Figure S2A). We evaluated the germline expression of this gene at both RNA and protein levels in the identified variant carriers. We observed a decrease on POLD1 levels (Figures 3A,B), suggesting that the altered mRNA was being degraded by the non-sense mediated decay pathway rather than producing a truncated protein. Nevertheless, although the germline protein expression pattern was markedly reduced in the Western Blot, the nuclear detection of POLD1 in the proband’s tumor revealed that POLD1 expression from the wild-type allele was still evident (Figure 3C).

We planned to establish a POLD1 p.(Lys648fs*46) model in SW837 cells using CRISPR/Cas9 via homologous recombination. At the same time, we also intended to introduce the founder pathogenic mutation POLD1 p.Leu474Pro as a positive control. We tested three different CRISPR strategies: the transient transfection of the sgRNA and Cas9 cloned into a plasmid together with the HDR template; the transient nucleofection of the sgRNA and Cas9 as a ribonucleoprotein complex together with the HDR template; and the generation of a cellular model constitutively expressing the sgRNA and Cas9, in which the HDR template was transiently transfected. The high on-target efficiency of Cas9 and the high mortality observed in all the attempts confirmed POLD1 as an essential gene. After several targeting rounds, we failed to generate a heterozygous model encoding the POLD1 p.(Lys648fs*46) variant. However, one of the attempts with the last strategy randomly produced a clone with a different frameshift mutation and a premature stop codon (Supplementary Figure S2A). Although far from ideal, we pursued to functionally characterize this clone, termed hereafter POLD1 ^+/− for convenience.

POLD1 ^+/− cells showed POLD1 downregulation at both RNA and protein levels, similar to what was observed in POLD1 p.(Lys648fs*46) carriers (Supplementary Figures S2B, 2C). We next assessed whether a reduced amount of wild-type POLD1 could be important for cell viability. Cell proliferation and cell survival were not affected in POLD1 ^+/− cells, indicating that reduced POLD1 levels did not affect cellular growth (Supplementary Figures S2D, 2E). We also challenged POLD1 ^+/− cells with different concentrations of the DNA-damaging agent CPT, but no significant differences were observed between wild-type SW837 cells and POLD1 ^+/− cells, suggesting that a single POLD1 functional copy allele could maintain normal function (Supplementary Figure S2F).

To study the specific functional consequences of POLD1 p.(Lys648fs*46) variant, we generated patient-derived organoids (PDOs) from normal rectum biopsies from the proband, which maintained the genetic background of the original tissue and, therefore, already had the loss-of-function POLD1 variant in their genome (POLD1 ^K648fs). At the same time, PDOs from a control individual were also generated. PDOs were stimulated with CPT to evaluate the effect of POLD1 haplosufficiency in both DNA replication and DNA damage repair (Figure 4A). After the genotoxic stress challenge, we aimed to assess the amount of DNA damage by γH2AX immunofluorescence staining, as it has emerged as a highly specific and sensitive molecular marker for monitoring DNA damage initiation and resolution (Mah et al., 2010). After a 24-h treatment, organoid growth and shape were unaffected (Figure 4B). The increase in γH2AX phosphorylation upon CPT treatment was evident in both control and POLD1 ^K648fs organoids, but no differences in the amount of γH2AX positive nuclei per organoid were observed between them (Figure 4C), indicating that reduced POLD1 expression did not alter the organoid sensitivity to CPT. We next assessed whether the CPT-induced DNA damage repair could be impaired in organoids lacking a functional copy of POLD1. After a 5-day resting period, the organoid growth arrest was evident in both control and POLD1 ^K648fs organoids (Figure 4B). The DNA from CPT-treated organoids and their untreated counterparts was collected, and whole-exome sequencing was performed. The number of substitutions detected in the samples was very low, and the mutational profiles did not indicate any accumulation of drug-induced mutations (Figure 4D).