This investigation was based on 67 index cases with attenuated or profuse adenomatous polyposis (34 from Finland, 21 from Chile, and 12 from Argentina) in which known genetic causes of polyposis had been excluded (APC, POLE, POLD1, PTEN and biallelic MUTYH; Figure 1 ). The cases were ascertained through the national polyposis research registries and local hospitals as described below. Most cases (47/66, 71%) exhibited attenuated polyposis. Detailed clinical data are available in Table S1 . Patient DNA was extracted from blood or EBV-transformed lymphoblasts as described by Renkonen et al. (15) DNA from formalin-fixed paraffin-embedded (FFPE) samples was extracted as described by Isola et al. (16) Patient RNA was extracted from lymphoblastoid cells using the NucleoSpin RNA extraction kit (Macherey-Nagel, Düren, Germany).

ES was performed at the Institute for Molecular Medicine Finland (FIMM, Helsinki, Finland) on Illumina HiSeq 2000 platform. The sequencing coverage and quality statistics for each sample are summarized in Table S2 . Reads were aligned to the human reference genome hg19 using the Burrows-Wheeler Aligner version 0.6.2. Quality control and primary and secondary analysis were carried out as described by Sulonen et al. (19) Tertiary analysis was carried out using VarSeq^® software (Golden Helix). Variants with allele frequency <0.003, nonsynonymous (frameshift, stop gained/lost, missense, disrupting donor/acceptor site variants) and predicted pathogenic with at least five of six programs assessing protein function in silico (for missense changes) were selected. All variants in DNA glycosylase genes were confirmed by Sanger sequencing with primers listed in Table S3 .

CNV analysis on ES data was carried out using the R package ExomeDepth (v1.1.10) (20). The patient ES data was run against appropriate patient samples with known pathogenic changes using default settings and annotated using common CNV data from the DECIPHER database (https://www.deciphergenomics.org/). All samples had a correlation score >0.99. Only CNVs with a BF score of 10 or above were considered as candidate CNVs.

To confirm that the two coding variants detected in NEIL1 in the index case of FAP104 affected different alleles (i.e., were in trans), cDNA was amplified with primers NEIL1_G83D_gcDNA_F and NEIL1_G83D_cDNA_R2 ( Table S3 ) and cloned using the TOPO^® TA Cloning^® Kit for Subcloning (Thermo Fisher) according to manufacturer’s instructions. Transformed E. coli were then grown on selective plates (100 µg Ampicillin) overnight and white colonies were grown in LB (100 µg Ampicillin) overnight. Plasmids were extracted with GenElute™ Plasmid Miniprep Kit (Sigma-Aldrich) according to manufacturer’s instructions and Sanger sequenced using aforementioned primers as well as primers from Sjöblom et al. (21).

To evaluate allele-specific mRNA expression (ASE) in the lymphoblastoid cells from the index individual of FAP104, a Single Nucleotide Primer Extension (SNuPE) reaction was designed based on the heterozygous NEIL1 c.506G>A variant. PCR products specific for cDNA (generated with primers NEIL1_G83D_gcDNA_F + NEIL1_G83D_cDNA_R2, Table S3 ) and gDNA (NEIL1_G83D_gcDNA_F + NEIL1_G83D_gDNA_R) served as templates for primer extensions with NEIL1_G83D_SNuPE_ext as the extension primer and ddA as the stopping nucleotide. The expected extension products were 34 bp (wild-type allele) and 24 bp (variant allele). Allele peak area ratios R<0.6 or R>1.67 indicated ASE (22).

NEIL1 mRNA expression in the lymphoblastoid cells from the index individual of FAP104 and healthy controls was evaluated by quantitative reverse transcription PCR (qRT-PCR) with TaqMan^® Gene Expression Assay (Applied Biosystems) for NEIL1 (Hs00908563_m1) and with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as an endogenous reference. The NEIL1 reaction targeted exons 5 – 6 and covered the two main isoforms. The reactions were normalized against the NEIL1 expression of healthy controls and the relative quantities were calculated using the ΔΔCT analysis.

To evaluate the stability of NEIL1 protein, lymphoblastoid cells from the index of FAP104 and unrelated healthy controls were treated with MG132 (Selleck Chemicals, Houston, Texas, USA). MG132 is a cell-permeable, proteasome inhibitor which reduces degradation of ubiquitin-conjugated proteins. Briefly, 0.5 x 10⁶ cells were incubated on 6 well plates for 8 hours and treated with 10, 30 or 50 µM MG132. Proteins from the cells were extracted in LAEMMLI extraction buffer.

NEIL1 protein expression in the treated and untreated lymphoblastoid cells from the index individual of FAP104 and healthy controls was assessed by Western blotting with the primary NEIL1 rabbit polyclonal antibody (12145-1-AP, RRID:AB_2251228; Proteintech, Rosemont, IL) targeting the NEIL1 short isoform (amino acids 1 – 390). The housekeeping protein glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as a loading control (ab128915, RRID:AB_11143050; Abcam, Cambridge, UK). P53 (#9282 RRID:AB_331476; Cell Signaling Technology, Danvers, Massachusetts, USA) was used as a technical control for MG132 experiments.

A custom assay utilizing Methylation-Specific Multiplex Ligation-Dependent Probe Amplification (MS-MLPA) was designed to evaluate NEIL1 promoter methylation in constitutional and tumor tissues. The NEIL1 promoter region was investigated with four MS-MLPA probe pairs ( Table S3 ), of which NEIL1_1 is located just upstream of the area found to be the most informative for methylation by Chaisaingmongkol et al. (23, 24).

VarScan2 variant detection algorithm version 2.3.2 was applied to tumor-normal pairs to identify non-synonymous somatic variants from ES data. Annotation of the variants was done using SnpEff version 4.0 with the Ensembl v68 annotation database (https://www.ensembl.org). Variants with a somatic p-value less than 0.01 were selected for somatic mutational signature analysis, which was carried out using the R package MutationalPatterns (25). The signatures were mapped against the 30 single-base substitution (SBS) and 18 insertion-and-deletion (ID) signatures recognized by the COSMIC database (v2 for SBS and v3.1 for ID, respectively, cancer.sanger.ac.uk).

A colorectal tumor from ARG046 was investigated for MMR protein expression by standard immunohistochemical procedures (26). Primary antibodies used were as follows (Roche Ventana, Indiana, USA): Anti-MLH1 (M1; 790-4535, RRID:AB_2336022), anti-MSH6 (44; 790-4455, RRID:AB_2336020), anti-MSH2 (G219-1129; 760-4265, RRID:AB_2336002), and anti-PMS2 (EPR3947; 760-4531, RRID:AB_2336010). DNA from the same tumor was evaluated for MLH1 promoter methylation by MS-MLPA using the SALSA MLPA ME011-B3 probemix (MRC-Holland, Amsterdam, the Netherlands).

Blood DNAs from the index individuals from our polyposis cohorts were evaluated for large rearrangements in MMR genes and MUTYH by multiplex ligation-dependent probe amplification (MLPA) according to the manufacturer’s (MRC-Holland, Amsterdam, the Netherlands) instructions. SALSA MLPA P003-D1 and SALSA MLPA P072-D1 were used for MLH1/MSH2 and MSH6/MUTYH, respectively, whereas PMS2 was investigated by SALSA MLPA P008-C1. The results from fragment analysis were analyzed by Coffalyser™ (MRC-Holland, Amsterdam, the Netherlands).

Methylation ratios in sporadic tumors vs. matching normal tissues obtained from MS-MLPA analyses ( Table S4 ) were compared using the Wilcoxon matched pairs test. IBM^® SPSS^® software (IBM SPSS Statistics 27, Armonk, NY: IBM Corp) was used for the analysis.

The European pathogenic founder variant NTHL1 c.268C>T, p.(Gln90Ter) (14) was detected in a homozygous state in the index individual from FAP1015 ( Figure 1 ). This individual had attenuated polyposis and was the only member with colorectal tumor manifestations in the family ( Table S1 ). The patient was additionally diagnosed with carcinomas of multiple organs characteristic of the tumor spectrum of NTHL1-associated polyposis (27).

NEIL1 variants were identified in FAP104 and FAP1021. The index of FAP104 with profuse polyposis (>200 polyps at 54 years of age) had two NEIL1 variants ( Figure 1 , Figure S1 ); a rare missense variant c.506G>A, p.(Gly169Asp), and a very rare frameshift variant c.821delT, p.(Ile274Thrfs*23), absent in the Finnish population. A subsequent cloning assay revealed that the variants affected different alleles. All three individuals with the NEIL1 c.506G>A variant had colorectal disease (cancer or polyps) and the same applied to the two individuals with the c.821delT variant ( Figure 2 ).

Conflicting evidence exists regarding the pathogenic significance of NEIL1 c.506G>A ( Figure 1 , Table S5 ). The variant allele frequency in the (global) population (0.001228) is higher than expected for a dominantly inherited disorder when comparing against NEIL1 variants reported through diagnostics. However, allele frequency of this variant in Finns (0.0001368) is almost ten times lower. Furthermore, previous functional studies conducted on NEIL1 c.506G>A consistently suggest pathogenicity (see Discussion). NEIL1 c.821delT is pathogenic according to the ACMG/AMP criteria ( Figure 1 , Table S5 ). Suitable biological specimens were available from the index of FAP104 to explore the consequences of the NEIL1 variants on RNA and protein level. We evaluated the relative mRNA expression from the two NEIL1 alleles by SNuPE and found that the frameshift variant containing transcripts were approximately twice less abundant than the missense variant containing transcripts in lymphoblastoid cells from the index individual of FAP104 ( Figure 3A ). By qRT-PCR, the total NEIL1 mRNA expression was essentially lower than in healthy controls studied for comparison ( Figure 3B ), suggesting that the ASE seen by SNuPE was more likely to reflect decreased expression from the frameshift allele than increased expression from the missense allele. Interestingly, Western blot analysis revealed a markedly elevated amount of normalized full-length NEIL1 protein compared to healthy controls, and no truncated protein was visible ( Figure 3C ). The abundant full-length protein likely originated from the missense allele, and no stable protein was apparently generated from the frameshift variant containing allele. In the absence of increased NEIL1 mRNA expression ( Figure 3B ), elevated NEIL1 protein in the Western blot was more likely to reflect aberrant protein stabilization than overexpression. The MG132 experiments (see Materials and Methods) did not reveal increased NEIL1 staining after treatment, indicating that regulation of NEIL1 protein expression is not MG132 mediated.

The index of FAP1021 with attenuated polyposis (30 polyps at 72 years of age) had a splice donor variant c.692+2T>C ( Figure 1 , Figure S1 ). In the literature, conflicting interpretations of pathogenicity for this splice variant exist (e.g., Dallosso AR et al. (28); Boldinova EO et al. (29)). In the absence of RNA, we were unable to experimentally verify splicing consequences of the variant. Based on available data, the ACMG/AMP classification is likely benign or VUS ( Table S5 ). Available in silico software evaluated the splice donor variant highly likely to affect splicing (0.9918, 0.6039, 0.96, and 0.99683 for ADA, RF, SpliceAI, and SPiCE, respectively).

A patient from the Argentine family ARG046 with attenuated mixed polyposis and colorectal carcinoma at the age of 60 years was found to be heterozygous for the previously described NEIL1 c.506G>A variant ( Figure 2 ). No other possibly pathogenic variants in NEIL1 were observed in the South American series.

Two families revealed likely deleterious OGG1 variants. The index of family PAF29 with attenuated polyposis had a heterozygous missense variant of OGG1, c.137G>A, p.(Arg46Gln). In the literature, the same variant was described in a patient with synchronous colorectal cancer at 36 years and adenomas (30). It was shown that the G to A change which affects the last nucleotide of exon 1 disrupts a splice donor sequence, resulting in extinct expression from the variant allele in cDNA from the patient (30). The authors classified the OGG1 c.137G>A variant pathogenic. Considering all available information, the ACMG/AMP criteria for likely pathogenic are fulfilled ( Figure 1 , Table S5 ). A heterozygous c.364G>T, p.(Glu122Ter) nonsense variant in the OGG1 gene, likely pathogenic by the ACMG/AMP criteria ( Figure 1 , Table S5 ), was detected in family 91. The variant was present in the index patient (ID 606) with attenuated polyposis but absent in the index patient’s brother (ID 657) with late-onset colorectal carcinoma ( Table S1 ).

Three heterozygous MUTYH variants were observed in the South American series ( Figure 1 , Figure S2 ). By ES, a frameshift variant of MUTYH, c.1101dupC, p.(Arg368Glnfs*164), was present in three individuals (ID 47, 534, and 535) out of four with colorectal adenomas or carcinoma from the Chilean family PAF20 and affected two generations. MUTYH c.1147delC, p.(Ala385Profs*23) was observed in the index individual HI003 (no other affected members were known to exist in this family). Both MUTYH variants described above are pathogenic by the ACMG/AMP criteria ( Table S5 ), with biallelic involvement linked to MAP. A missense variant, MUTYH c.869G>A, p.(Cys290Tyr), classified as likely pathogenic ( Table S5 ), was found in the index individual of family PAF43 (carrier statuses of the remaining family members were unknown). This family showed features of MAP (over 100 polyps in the index individual and an apparent recessive transmission pattern, Figure S2 ), raising the possibility that the MUTYH allele currently considered wildtype might harbor a defect that had escaped detection. However, manual IGV analysis of the gene and MLPA (with MSH6-MUTYH and APC MLPA kits) for large genomic rearrangements provided no support for biallelic MUTYH involvement.

DNA was available from a colorectal tubular adenoma from the index of FAP104 (compound heterozygous NEIL1; c.506G>A and c.821delT), two desmoid tumors from the paternal aunt of the index of FAP104 (heterozygous NEIL1 c.506G>A), and a colorectal carcinoma from the index of ARG046 (heterozygous NEIL1 c.506G>A) for somatic mutational profiling. We first determined the total mutational loads, since elevated numbers of somatic variants may point to defects in DNA replication or repair (10, 15). The total numbers of somatic nonsynonymous variants were 281 (adenoma), 45 and 57 (desmoids), and 1146 (carcinoma) by VarScan2 analysis. Based on the commonly used threshold of 10 variants/Mb, only the carcinoma of ARG046 was hypermutated (35 somatic variants/Mb).

All somatic variants meeting our selection criteria (VarScan2 p<0.01) are listed in Table S7 . No somatic variant or loss of heterozygosity of NEIL1 was observed in any sample. Thus, there was no evidence of a somatic “second hit” to the remaining wildtype allele in the monoallelic NEIL1 variant carriers. The adenoma from the index of FAP104 showed a truncating APC variant (c.4666dupA, p.Thr1556fs; variant allele frequency (VAF) 23%) and KRAS c. 35G>C, p.Gly12Ala (VAF 25%), both representing alterations typical of colorectal tumorigenesis. The two desmoid tumors revealed extensive sharing of somatic variants, suggesting a common origin for the tumors.

As the patterns of somatic variants can offer insights to the underlying biological processes, a mutational signature analysis was conducted on the tumors ( Figures 4A, B ). VarScan2-based somatic variants were included in this analysis. COSMIC (31) SBS signature 3 (defective homologous recombination) was prominent in all three tumors from FAP104 ( Figure 4A ). Desmoid tumors from individual II.4 additionally revealed SBS7 (ultraviolet radiation exposure) and a discernible SBS24 linked to aflatoxin-associated mutagenesis (32). Interestingly, the hypermutable colorectal carcinoma from ARG046 showed prominent MMR deficiency-associated signatures SBS6 and SBS26, together with SBS12 (unknown etiology). The ID signature 6 supported defective homologous recombination in tumors from FAP104, whereas ID7 was compatible with deficient MMR in the colorectal carcinoma from ARG046 ( Figure 4B ).

To resolve the MMR-deficient pattern of somatic alterations in the colorectal carcinoma from ARG046, the tumor was tested for MLH1 promoter methylation, but no hypermethylation was present. However, immunohistochemical analysis revealed selective absence of PMS2 protein ( Figure 4C ). Subsequent MLPA analysis of blood DNA showed a heterozygous deletion of PMS2 exons 2 - 11 (NM_000535.5:c.(23 + 1_24-1)-(2006 + 1_2007-1)del; Figure 4D ). No additional cases with large rearrangements of MMR genes were detected when our entire polyposis series was evaluated by MLPA (and no small sequence alterations with possible pathogenicity existed in MMR genes by ES).

As NEIL1 is commonly hypermethylated in cancer (23, 24), we designed a MS-MLPA kit to determine constitutional and somatic methylation status of our patient samples. Of the four MS-MLPA probe pairs, NEIL1_1 interrogated a region previously shown to be informative for methylation (23, 24) and showed the best discrimination between normal and tumor tissues ( Table S4 ). Blood and normal colonic mucosae even from reference individuals revealed considerable methylation, and examination of blood DNAs from our polyposis cases (with or without NEIL1 variants) raised no suspicion of constitutional NEIL1 epimutation in any case. Compared to paired normal tissues, tumors from individuals with NEIL1 variants occasionally displayed higher methylation dosage ratios, but no significant somatic hypermethylation of the promoter region was evident. Comparing paired tumor and normal tissues from sporadic cases with MSS or MSI carcinomas or adenomas revealed no significant difference by Wilcoxon matched pairs test (Z=-1.03, p=0.133 for MSI carcinomas vs matching normal tissues, and Z=-0.217, p=0.828 for MSS carcinomas vs matching normal tissues by NEIL1 I probe, respectively).