Abstract

Background

   Gastric cancer (GC) is the fifth-leading cause of cancer-related
   mortality, with a 5-year survival rate less than 20%. It develops from
   preneoplastic lesions to adenocarcinoma, but these early genetic
   alterations remain poorly understood. Therefore, we aimed to identify
   early genetic drivers underlying the development of preneoplastic
   lesions and the initiation of gastric carcinogenesis.

Methods

   We characterized preneoplastic and early gastric adenocarcinoma using
   48 samples from 16 Guatemalan patients, a country with a high incidence
   of GC. We sequenced a panel of 127 genes to identify early genetic
   drivers and possible actionable targets.

Results

   We identified extensive genetic heterogeneity, including single
   nucleotide and copy number variations. After comparing our data with
   other studies, we identified TP53 and APC as the most mutated genes in
   preneoplastic lesions and early GC. Our mean tumor mutational burden
   was higher in diffuse (0.017 mutations/Mb) and intestinal
   adenocarcinomas (0.015) than in chronic gastritis (0.005), and analysis
   of the mutational signatures revealed several processes acting at
   different stages of the disease. Signatures S15 (DNA mismatch repair
   deficiency) and S03 (homologous recombination deficiency) were more
   frequent in early adenocarcinoma than in chronic gastritis, intestinal
   metaplasia, necrosis, tubular adenoma, and atrophy. Notably, 10 of 16
   patients (62.5%) had at least one actionable mutation in their
   preneoplastic lesions or gastric adenocarcinomas.

Conclusions

   We show that at the preneoplastic and earliest stages, GC is
   genetically heterogeneous and presents key cancer-driving mutations
   that may participate in neoplastic transformation and progression, with
   62.5% of lesions having the potential for treatment. This study expands
   the limited research on early GC and highlights key opportunities for
   precision medicine in populations with high GC incidence.

   Keywords: early cancer drivers, preneoplastic lesions, chronic
   gastritis, gastric adenocarcinoma, somatic variants, germline variants

Introduction

   Gastric cancer (GC) is the fifth-leading cause of cancer-related
   mortality worldwide [[52]1], and adenocarcinoma accounts for 90% of
   cases, which are classified as either intestinal or diffuse type
   according to the Lauren classification system [[53]2]. The disease has
   four molecular subtypes: tumors positive for Epstein-Barr virus,
   microsatellite unstable tumors, genomically stable tumors, and
   chromosomal instable lesions, which are used for patient stratification
   and clinical trials for targeted therapies [[54]3]. Despite significant
   advances in clinical and pharmacological research, the overall 5-year
   survival rate is still less than 20%, with even lower rates in
   developing countries [[55]4].

   GC is associated with a variety of environmental and genetic factors.
   The majority of cases are due to infectious agents such as Helicobacter
   pylori (H. pylori) and Epstein-Barr virus [[56]5]. A small percentage
   of less than 3% are hereditary, caused mainly by inactivating germline
   mutations in E-cadherin (CHD1) [[57]6]. Other models, such as Correa’s
   cascade, describe the development of GC from precursor lesions,
   starting with chronic gastritis. The progression continues with
   atrophic gastritis, followed by intestinal metaplasia, and finally ends
   in adenocarcinoma [[58]7]. In locally advanced GC, mutations accumulate
   progressively in genes such as TP53, APC, PIK3CA, and KRAS, and
   amplification events have been detected in JAK2, CD274 (PD-L1), and
   PDCD1LG2 (PD-L2), among others [[59]3].

   In Central and South America, the mortality rate of GC is one of the
   highest in the world [[60]8]. Guatemala is among the nations with a
   notably high incidence of GC (12.2 age-standardized rate per 100,000)
   and a high GC mortality (10.7 age-standardized rate per 100,000),
   ranking fifth and fourth in the world, respectively [[61]1]. GC is
   frequently detected at an advanced stage with different prevalence
   among ethnic groups, histological subtypes, and treatment response. For
   instance, diffuse GC is more prevalent in intestinal in the Mayan than
   that in the Mestizo population [[62]9]. Guatemalan patients with
   locally advanced GC treated with chemotherapy have shown a median
   disease-free survival of 5.2 months and a median of overall survival of
   15.3 months [[63]10]. This has fueled a government effort in Guatemala
   to implement policies for prevention and treatment strategies for
   different types of cancer through early detection and the
   implementation in collaboration with research efforts [[64]11], a
   decision that highlights the critical need for enhanced prevention
   strategies, diagnosis, and disease monitoring with early molecular
   markers to reduce the impact of GC in this population.

   Most GC studies have focused on describing the molecular events of the
   disease in its advanced stages with limited information on early
   molecular events, defined as the first genetic alterations conducting
   the development of cancer, including single nucleotide variations
   (SNVs), indels, and copy number variations (CNVs). To address this
   research gap, we aimed to identify early genetic drivers in
   preneoplastic lesions and early GC in a high incidence population on a
   panel of 127 genes significantly mutated in cancer using next
   generation sequencing. We assessed mutations and CNVs, tumor mutational
   burden (TMB), and mutational signatures and identified potential
   actionable mutations in 16 patients with multiple samples recruited at
   the Unit of Gastroenterology at the Department of Internal Medicine at
   Roosevelt Hospital in Guatemala. In addition, to enhance the
   significance of our findings, we compared mutational compositions of
   preneoplastic lesions and early GC between our present study and
   previous studies to provide a broader perspective of its early
   molecular driving events. The evidence reported in this paper on the
   early drivers in preneoplastic lesions and early GC may become critical
   for defining future therapeutic options.

Materials and methods

Sample information

   A total of 48 samples from 16 patients with preneoplastic lesions and
   gastric adenocarcinoma were selected from the Gastroenterology Unit of
   the Internal Medicine Department of Roosevelt Hospital from Guatemala
   ([65]Figure 1). All patients were diagnosed through endoscopic biopsy.
   Tissue samples were formalin-fixed and paraffin-embedded for
   histopathological analysis with hematoxylin and eosin staining by
   pathology specialists, and analytical biopsies for sequencing were
   preserved in RNAlater and fresh frozen. The Sydney classification
   system was used to characterize and grade the histopathological
   features of gastritis and the Lauren classification was used to
   categorize gastric adenocarcinoma from biopsies [[66]2, [67]12]. In
   addition, H. pylori was evaluated by using Giemsa staining. The
   inclusion criteria were as follows: (i) patients over 18 years old and
   diagnosed with preneoplastic lesions and early-stage gastric
   adenocarcinoma; (ii) patients with primary treatment-naïve tumors;
   (iii) tumor tissues with cellularity ≥70%; and (iv) patients with
   complete clinical history. Each patient had at least three samples from
   chronic gastritis, necrosis, atrophy, intestinal metaplasia, tubular
   adenoma, or diffuse or intestinal adenocarcinoma ([68]Supplementary
   Table S1). The study was approved by the Research and Ethics Committees
   of the National Hospital Carlos N. Arana of Chiquimula, Guatemala
   (protocol 8311647) and conducted in accordance with the Declaration of
   Helsinki.

Figure 1.

   [69]Figure 1.
   [70]Open in a new tab

   Experimental design. Panel 1: research design. Patient recruitment and
   approval of the protocol by the ethics committee for the study of
   preneoplastic lesions and gastric adenocarcinoma. Panel 2: sample
   preparation. DNA extraction and library preparation with exome
   enrichment using a commercial kit based on 127 mutated genes. Panel 3:
   sequencing of the libraries and identification of SNVs, CNVs,
   mutational signature, and MEIs. SNVs of preneoplastic lesions and early
   GC from other studies were integrated into our analysis [[71]21–24].
   Statistical analyses and visualization were carried out using packages
   in R and Python. SNVs = single nucleotide variants; CNVs = copy number
   variants; MEIs = mobile element insertions. Created in BioRender.com.

DNA extraction and library preparation

   DNA was extracted from fresh frozen tissues using 50 to 200 mg of
   sample with the DNeasy Blood & Tissue Kit (Qiagen, Hilden, Germany)
   pursuant to the manufacturer’s instructions. DNA quantification was
   performed by fluorometry (Qubit, Invitrogen, USA), DNA integrity was
   assessed on 0.8% agarose gels, and DNA purity was estimated by
   spectrophotometry (Implen, Germany). All samples had an A260/280 ratio
   ≥1.8 and were used for subsequent experiments. The library preparation
   was done with a xGen Pan-Cancer Hybridization Panel kit (IDT,
   Coralville, USA) based on the most 127 significantly mutated genes in
   the most prevalent 12 cancer types, according to The Cancer Genome
   Atlas study [[72]13], a panel that covers 800 kb of protein-coding
   bases. Size distribution of the final libraries was evaluated on the
   Bioanalyzer by using a High Sensitivity DNA kit (Agilent, USA), and
   paired-end sequencing was performed on an Illumina HiSeq 2500 for
   2 × 150 cycles with a mean depth of 119.9X (± 38.4).

Data preprocessing and identification of somatic mutations

   Sequencing reads were aligned to the human genome reference hg19 with
   BWA-MEM [[73]14], and GATK tools were used for data preprocessing
   [[74]15]. Single nucleotide variants were called with Mutect2 [[75]16]
   and annotated with ANNOVAR [[76]17]. Variants were filtered as follows:
   (i) variants that passed the quality filters (“PASS” flag); (ii) Phred
   score ≥ 30; (iii) variants with mapping quality ≥ 60; (iv) variants
   with an allelic frequency less than 0.001 in 1,000 genomes, ESP6500,
   and gnomAD databases; (v) variants with a mutant allelic fraction >
   0.03 were kept; and (vi) variants with evidence in both strands, with a
   minimum of two reads on each strand. Driver mutations were defined
   according to well-known hotspot mutations in COSMIC [[77]18] and
   cBioportal [[78]19], and potential driver mutations if predicted by two
   of three algorithms, SIFT, PolyPhen2, or MutationTaster, as
   deleterious. All driver mutations were manually curated by inspection
   of the BAM files using Integrative Genomics Viewer (IGV) software
   [[79]20]. The number of SNVs and indels in each sample was normalized
   by the size of the analyzed region to evaluate the TMB (mutations/Mb).

   In addition, in order to provide a wider molecular perspective on the
   somatic mutations of early GC, we included additional mutational data
   from previous studies. We conducted a search of the PubMed database
   with the terms “gastric cancer,” “stomach cancer,” “early cancer,”
   “preneoplastic lesion,” and “early adenocarcinoma” for next-generation
   sequencing GC studies and identified four studies with 15, 43, 8, and
   17 samples from USA, Japan, and Korea [[80]21–24]. Samples of
   advanced-stage GC in these studies were eliminated from the analysis.
   We then compared the mutational composition reported in these reports
   with our data, focusing on driver genes associated to SNVs and CNVs,
   described above.

CNV identification

   CNVs were identified by using CNVkit [[81]25] in tumor-only mode under
   the default parameters. Variants were filtered based on depth, the 95%
   confidence interval, and the directionality criteria while considering
   the biological activity of the affected genes as follows: (i) CNVs with
   copy number (CN) = 2, the normal genome state, were excluded; (ii)
   oncogenes with CN ≤ 2 were excluded and CN > 2 were kept; (iii) tumor
   suppressor genes with CN ≥ 2 were excluded and CN < 2 were kept.

Mobile element insertions identification

   Mobile element insertions (MEIs) were analyzed by using MELT [[82]26].
   Variants were filtered as follows: (i) variants marked as “PASS” with
   evidence of MEIs in the 5’ and 3’ regions; (ii) variants with
   ASSESS = 5, evidence of target site duplication (TSD), split reads
   (SRs), discordant reads, breakpoint site, and insertional mutagenesis;
   and (iii) coincident SRs in alternative chromosomic location were kept.
   The selected variants were visualized and manually curated in IGV with
   a bed file of Repeat Masker version hg19 obtained from the UCSC Table
   Browser [[83]27].

Mutational signatures

   The filtered SNVs (described above) with an allelic fraction > 0.03
   were used for the analysis of mutational signatures. Given the low
   number of mutations in individual samples, we consolidated the samples
   into five histopathological groups: chronic gastritis, atrophy,
   intestinal metaplasia, intestinal adenocarcinoma, and diffuse
   adenocarcinoma. Necrosis and tubular adenoma were excluded because each
   consisted of a single sample. Mutational signatures were then evaluated
   in the context of 96 trinucleotides by using the R package
   deconstrictSigs [[84]28]. To evaluate the distribution of the samples
   with mutational signatures, an unsupervised hierarchical clustering
   analysis was performed by calculating Euclidean distances as previously
   described [[85]29].

Pathway enrichment analysis

   Pathway and network enrichment analysis was defined by the David
   Functional Annotation Tool 6.8 [[86]30] using driver genes in the KEGG
   pathway database. Uninformative pathways were eliminated, including
   general cancer and noncancer diseases such as addictions, development
   processes, and metabolic, psychiatric, and parasitic diseases.
   Signaling pathways were then evaluated using Fisher’s exact test with
   Bonferroni corrections for the false discovery rate (FDR). After
   Bonferroni correction, P values less than 0.01 were included in the
   study.

Global actionable alterations

   Treatments approved by the Food and Drug Administration (FDA) for
   actionable genes in all types of cancers were analyzed in the OncoKB
   database [[87]31] and classified into Tiers according to ASCO-CAP
   classification [[88]32]. Only variants that were classified as (i) Tier
   I, FDA-approved therapy or included in professional guidelines or (ii)
   Tier II, FDA-approved therapies for different tumor types or
   investigational therapies were included.

Statistical analysis

   Significant differences in the number of mutations and TMB, as well as
   the mutational signature across disease stages were assessed using the
   Kruskal–Wallis rank sum test, followed by pairwise comparisons with the
   Wilcoxon rank sum test and Bonferroni correction for FDR. P values less
   than 0.05 were considered statistically significant.

Results

Clinical characteristics

   A total of 48 samples were selected from 16 patients with preneoplastic
   lesions and GC at the Gastroenterology Unit of the Internal Medicine
   Department of Roosevelt Hospital in Guatemala. Each patient contributed
   three samples, from chronic gastritis, necrosis, atrophy, intestinal
   metaplasia, tubular adenoma, or diffuse or intestinal adenocarcinoma
   ([89]Supplementary Table S1). Most patients with preneoplastic lesions
   and gastric adenocarcinoma were older than 50 years (56.3%) were male
   (62.5%) and had no family history of cancer (75%). One patient had H.
   pylori infection ([90]Table 1).

Table 1.

   Epidemiological characteristics of patients with preneoplastic lesions
   and early GC
   Characteristic        Mean ± SD  n (%)
   Age
    < 50 years           41.3 ± 4.8 7 (43.7)
    ≥ 50 years           68 ± 9.4   9 (56.3)
   BMI
    ≤18.4 underweight    16.5 ± 0   2 (12.5)
    18.5 < 24.9 normal   22.1 ± 2.3 10 (62.5)
    25 < 29.9 overweight 27.2 ± 1.7 4 (25)
   H. pylori infection
    Yes                  –          1 (6.3)
    No                   –          15 (93.8)
   Sex
    Female               -          6 (37.5)
    Male                 –          10 (62.5)
   FHC
    Yes                  –          4 (25.0)
    No                   –          12 (75.0)
   [91]Open in a new tab

   BMI = body mass index, SD = standard deviation, H. pylori =
   Helicobacter pylori, FHC = family history of cancer.

TMB and mutational signature

   Overall, we identified 477 SNVs with a mean of 9.9 per sample, which
   were categorized into five histopathological groups: diffuse
   adenocarcinoma with a mean of 14.2 SNVs (range, 8–29), intestinal
   adenocarcinoma 12.18 (range, 5–22), intestinal metaplasia 10.6 (range,
   4–21), atrophy 8.35 (range, 3–20), and chronic gastritis 4.71 (range,
   1–13) ([92]Figure 2A). Necrosis and tubular adenoma samples were
   excluded from this analysis because they represented only single cases.

Figure 2.

   [93]Figure 2.
   [94]Open in a new tab

   Tumor mutational burden and mutational signatures in preneoplastic
   lesions and early gastric adenocarcinomas. (A) Number of mutations; (B)
   TMB; and (C) mutational signatures of the preneoplastic lesions and
   early GC segmented by histopathological type. Histopathological types
   with at least three samples were included in the analysis of number of
   mutations in (A). TMB = tumor mutational burden; MMR = mismatch repair;
   HR = homologous recombination.

   The mean TMB was 0.012 (mutations/Mb; range, 0.001–0.036), with
   differences between the histopathological subtypes. Diffuse
   adenocarcinoma had a mean TMB of 0.017 (range, 0.010–0.036), intestinal
   adenocarcinoma of 0.015 (range, 0.006–0.027), intestinal metaplasia of
   0.013 (range, 0.005–0.026), atrophy of 0.010 (range, 0.003–0.025), and
   chronic gastritis of 0.005 (range, 0.001–0.016) ([95]Figure 2B). No
   statistically significant differences were observed between the number
   of mutations and the TMB within the histopathological subtypes.

   An analysis of mutational signatures was performed to identify possible
   molecular etiologies of preneoplastic lesions and gastric
   adenocarcinomas. We computed the mutational signatures by categorizing
   the samples into their histopathological subtypes and identified five
   mutational signatures, four of which (S01, S03, S15, and S17) have been
   previously described in gastric adenocarcinoma ([96]Figure 2C).
   Mutational signature S01, associated with the aging process, was
   predominant in almost all samples except in chronic gastritis.
   Signature S06 (DNA mismatch repair, MMR) was detected in chronic
   gastritis and tubular adenoma, and signature S17 was found in chronic
   gastritis and atrophy. S15, associated with DNA MMR, was found in
   diffuse and intestinal adenocarcinoma, and S03 (homologous
   recombination deficiency, HRD) was found only in diffuse
   adenocarcinoma. An unsupervised hierarchical clustering analysis was
   conducted to identify the mutational signature programs operating in
   the early GC carcinogenesis, and this yielded three different groups of
   mutational signatures patterns Group 1, comprising chronic gastritis,
   was characterized by the signature S06 and S17, and lack of S01; Group
   2, including intestinal metaplasia, necrosis, tubular adenoma, and
   atrophy, was defined by a predominant proportion of S01 signature; and
   Group 3 was characterized by a lower proportion of S01 and the presence
   of S15 in intestinal and diffuse adenocarcinoma, with S03 additionally
   detected in intestinal adenocarcinoma. After the statistical test and
   applying Bonferroni corrections, no significant differences were found
   between the mutational signatures in preneoplastic lesions and early
   GC.

Patients with preneoplastic lesions and GC showed genetic heterogeneity

   An extensive genetic heterogeneity was detected in preneoplastic
   lesions and gastric adenocarcinoma in term of SNVs and CNVs. We
   identified nine driver mutations in four genes and 49 potential driver
   mutations in 29 genes ([97]Figure 3, [98]Supplementary Figure S1 and
   [99]Supplementary Table S2). Driver mutations in SMAD4 were found in
   the atrophy and tubular adenoma of the same patient (p4,
   [100]Supplementary Figure S1), TP53 in intestinal adenocarcinoma (p21,
   p22, p37) and diffuse adenocarcinoma (p23), NRAS in intestinal
   adenocarcinoma (p14), and ATM in atrophy (p6). One patient had a
   potential germline mutation in ATM c.6680G>A p. Arg2227His that was
   validated by Sanger sequencing (p45, [101]Supplementary Figures S1 and
   S2).

Figure 3.

   [102]Figure 3.
   [103]Open in a new tab

   Driver mutation in preneoplastic lesions, early, and advanced GC. The
   distribution and composition of SNVs were identified in this work and
   other studies with early lesions and early GC. Only CNVs were included
   in our study. Legends of type of alteration, tissue sample type, TNM
   grading, sequencing method, and author are shown in the right of the
   figure. *In the early GC [[104]24], we included the 20 genes with the
   highest mutation prevalence. GC = gastric cancer; SNVs = single
   nucleotide variants; CNVs = copy number variants.

   We also compared these results with other data on preneoplastic lesions
   and early GC to identify early driver mutations that may participate in
   cancer progression. In gastric hyperplastic polyp (GHP) with high grade
   dysplasia, TP53 was found in 25.0% (2/8) of cases, whereas PIK3CA was
   found in only 12.5% (1/8) in pyloric-type dysplasia cases
   ([105]Figure 3) [[106]23]. In pyloric gland adenomas (PGA), 10 out 15
   samples with low-grade dysplasia had mutations in APC, KRAS, and GNAS
   genes, whereas 5 samples with high-grade dysplasia exhibited mutations
   in several genes, including APC, CTNNB1, KRAS, GNAS, TP53, CDKN2A,
   PIK3CA, and EPHA5 genes, but did not exhibit mutations in the triad of
   APC, KRAS, or GNAS genes [[107]21]. In the premalignant lesion of
   dysplasia/intraepithelial neoplasia (D/IEN), two genes were frequently
   mutated, APC in 76% (19/25) of cases and TP53 in 48% (11/25) [[108]22].
   In other studies of early GC drivers, mutations were identified in
   TP53, APC, PIK3CA, ARID1A, and KRAS [[109]24]. Overall, the data showed
   that TP53 and APC are the most mutated genes in preneoplastic lesions
   and early GC.

   Additionally, we identified 22 CNVs in 13 genes in the samples from
   this study ([110]Figure 3, [111]Supplementary Figure S1, and
   [112]Supplementary Table S3). Deletion in VHL was the most common
   alteration and was detected in five samples (four patients), followed
   by deletion of ATM and amplification of NOTCH1 in three patients.
   Interestingly, the sample EDM_0009 (p23, diffuse adenocarcinoma) had a
   considerable number of CNVs that consisted of seven amplifications and
   one deletion in VHL. In our study, we classified driver mutations and
   potential drivers into 6 signaling pathways for CNV and 10 for SNVs
   ([113]Supplementary Figure S1). RAS/MAPK and transcription factors
   pathways were the most common in CNVs, and cell cycle, chromatin, and
   transcription were the most prevalent in SNVs

   Furthermore, we evaluated mobile transposable elements as alternative
   molecular pathogenesis mechanisms in early GC lesions. Potential
   candidates were manually visualized in IGV using a bed file of Repeat
   Masker version hg19. However, they were discarded due to lack of
   evidence in TSD, SRs, discordant reads, or insertional mutagenesis.

Possible therapeutic targets for GC

   We performed a specific search in the OncoKB database to identify
   mutations associated with treatment of GC following the ASCO-CAP
   guidelines for mutation classification [[114]32]. The search showed
   that 62.2% (10/16) of patients had at least one actionable somatic
   variant in preneoplastic lesions or gastric adenocarcinoma in six genes
   ([115]Figure 4A and B). ATM, BRAF, and FBXW7 were found to be sensitive
   to olaparib, vemurafenib, and sirolimus, and TP53, and NRAS and SMAD4
   were found to confer resistance to venetoclax and EGFR inhibitors
   ([116]Figure 4C). Three main signaling pathways were detected in these
   genes: cell cycle (50%), RAS/MAPK (33.3%), and transcription factor
   (16.7%) ([117]Figure 4C).

Figure 4.

   [118]Figure 4.
   [119]Open in a new tab

   Patients with potentially actionable alterations in preneoplastic and
   gastric adenocarcinoma. (A) Percentage of patients with actionable
   alterations; (B) percentage of actionable genes; (C) actionable genes
   and possible treatments according to variants. Pathway and drugs are
   shown in the left panel, and preneoplastic lesions and GC in the bottom
   panel.

Discussion

   Most studies have focused on the advanced stages of GC, and there is
   limited research on the earliest genetic alterations that drive the
   transition from preneoplastic lesions to malignancy, particularly in
   populations from high-incidence regions in developing countries. Our
   study provides the first genomic and molecular analysis of
   preneoplastic lesions and early GC in a Guatemalan population, a
   high-incidence region with limited prior research.

   H. pylori was detected in only 6% of our cases. Consistent with our
   findings, one large molecular study showed a very low prevalence of H.
   pylori, but other works have shown higher numbers [[120]3, [121]5]. In
   developing countries, GC often presents at a younger age, with early
   life H. pylori infection and other factors contributing to this
   accelerated onset [[122]33]. In Guatemala, the seroprevalence of H.
   pylori has been reported to reach approximately 70%, although molecular
   studies on GC are lacking [[123]34]. The detection rate in our study
   may be influenced by the patients’ ages at diagnosis, with the majority
   being over 50 years old, as well as the diagnostic methods employed.

   We evaluated the mean TMB to estimate of mutagenic exposure and DNA
   repair defects. Diffuse adenocarcinoma (0.017 mutations/Mb) and
   intestinal adenocarcinoma (0.015 mutations/Mb) had higher rates than
   intestinal metaplasia (0.013 mutations/Mb), atrophy (0.010
   mutations/Mb), and chronic gastritis (0.005 mutations/Mb). Although a
   higher TMB was identified in the advanced stages of GC, no
   statistically significant differences were found between these lesions.
   In precursor lesions of GC, such as chronic gastritis, the mutation
   rate was low, whereas advanced stage showed an increased mutation rate,
   as well as a high TMB, which can be attributed to disease progression
   [[124]35].

   We identified five mutational signatures in preneoplastic and
   adenocarcinoma samples, four of which (S01, S03, S15, S17) have GC
   associations [[125]36]. The aging-related S01 signature, caused by
   5-methylcytosine deamination at CpG sites, dominated most samples
   except chronic gastritis. In these inflammatory lesions, S01
   contribution was attenuated by more active processes, including S17
   (15% median contribution, linked to oxidative damage) [[126]36] and S06
   (85%, MMR deficiency). This likely reflects gastritis inflammatory
   microenvironment, where oxidative stress and epithelial damage promote
   these signatures over S01 clock-like accumulation. Furthermore, the
   progression from chronic inflammation to intestinal metaplasia and
   neoplasia likely involves a shift from exogenous or inflammatory
   mutagenesis to endogenous, age-related processes, which may explain the
   increased prominence of S01 in more advanced lesions.

   The MMR-deficient signature S15 was specifically detected in
   adenocarcinomas, while the oxidative damage-associated S17 appeared in
   atrophy and chronic gastritis. Intriguingly, while MMR deficiency
   signatures (S06 and S15) were present in chronic gastritis, tubular
   adenoma, diffuse and intestinal adenocarcinoma, we found no underlying
   MMR gene mutations. This suggests alternative mechanisms such as MLH1
   promoter methylation (reported in 50% of sporadic and 23% of familial
   GC cases) [[127]37] may drive these mutational patterns. The exclusive
   presence of S06 in chronic gastritis and tubular adenoma, previously
   unreported in GC, highlights DNA MMR dysfunction as a potential early
   event in gastric carcinogenesis, though studies of preneoplastic
   lesions remain limited.

   Notably, we detected HRD-associated S03 in intestinal adenocarcinoma, a
   signature more common in breast and ovarian cancers but previously
   reported in GC [[128]38]. While one atrophy case had a BRCA1 mutation,
   the S03 HRD signature appeared in a separate intestinal adenocarcinoma,
   implying other HR pathway genes (not covered by our panel) or
   epigenetic silencing (e.g. of BRCA1 and RAD51C as in breast cancer
   [[129]39]) could contribute to this signature.

   Unsupervised hierarchical clustering analysis identified three distinct
   sample groups, indicating that mutagenic processes may operate either
   independently or in combination at different stages of disease
   progression. Notably, advanced stages in our study were characterized
   by the presence of the S15 mutational signature, suggesting that it
   could be the driving process of carcinogenesis at this stage (Group 3).

   Our study also revealed extensive genetic heterogeneity, with 31 genes
   harboring driver or potential driver mutations and 13 genes with CNVs,
   especially in SMAD4, TP53, NRAS, and ATM. TP53 and NRAS were found in
   intestinal and diffuse adenocarcinoma. In addition, several studies
   have reported mutation of TP53 in more than 50% of cases of GC [[130]3,
   [131]40, [132]41] and genes involved in the RAS/MAPK signaling pathway
   including KRAS, BRAF, PIK3CA, and NRAS [[133]42–44], consistently with
   our present work in which we found four patients with TP53 mutations
   that may indicate a potential chromosomal instability molecular
   phenotype [[134]3].

   Comparative analysis with other studies of preneoplastic lesions and
   early GC supports our findings. TP53 (25%) and PIK3CA (12.5%) mutations
   were identified in gastric hyperplastic polyps, suggesting their role
   in malignant transformation [[135]23]. In addition, pyloric gland
   adenomas/adenocarcinoma, with low-grade dysplasia presented APC, KRAS,
   and GNAS mutations, while advanced lesions acquired additional
   alterations in TP53, PIK3CA, and other genes [[136]21]. Other studies
   have similarly identified APC and TP53 mutations as early events in
   gastric tumorigenesis, particularly in intestinal-type cancers
   [[137]22, [138]24].

   TP53 and APC emerge as the most consistently mutated genes across the
   gastric carcinogenesis spectrum. As a critical tumor suppressor, TP53
   maintains genomic stability through cell cycle regulation, DNA repair,
   and apoptosis induction. Early TP53 mutations may confer a selective
   advantage in the harsh gastric environment, facilitating proliferation
   despite genotoxic stress, as evidenced by the S17 mutational signature.
   This pattern mirrors observations in other malignancies, where TP53
   defects frequently initiate preneoplastic progression [[139]45].
   Clinically, TP53 mutations correlate with poorer outcomes in GC
   patients [[140]46, [141]47]. While APC mutations are best characterized
   in colorectal cancer (occurring in 80% of cases), they also
   significantly contribute to gastric tumorigenesis through Wnt pathway
   activation [[142]48, [143]49]. These findings underscore the complex
   molecular landscape underlying GC development and progression.

   Hereditary GC represents less than 3% of cases and is primarily driven
   by germline mutations in CDH1 [[144]6]. In our cohort, we identified
   one patient (6.3%) with a likely pathogenic germline mutation in ATM
   (p.R2227H). However, this patient did not have family history of
   cancer. Germline mutations in this gene have been reported in other GC
   studies [[145]50–52], suggesting possible association with the disease,
   but the current evidence of the increased risk is inconclusive
   [[146]53].

   In terms of CNVs, VHL, ATM, and NOTCH1 were the most frequently
   observed genes and were mainly detected in adenocarcinomas. VHL has
   been extensively described in clear cell renal carcinoma and
   hemangioblastomas, associated with hypoxic and angiogenic tumors
   [[147]54], and proposed as a therapeutic target [[148]55]. The affected
   signaling pathways with higher frequency were RAS/MAPK, cell cycle,
   chromatin remodeling, and transcription factor. In line with this work,
   these pathways have been described in several studies of locally
   advanced GC [[149]3]. We evaluated MEI as an additional mechanism of
   mutational pathogenesis, but we did not identify any robust evidence of
   transposon insertion in the target genes covered.

   Our findings carry important therapeutic implications. MMR-deficient GC
   have elevated TMB, which enhances tumor immunogenicity and correlates
   with improved response to immunotherapy and better prognosis [[150]56].
   Similarly, HRD serves as a predictive biomarker for sensitivity to both
   platinum-based chemotherapy and immunotherapy. Metastatic GC patients
   with HRD, including those with diffuse adenocarcinoma, have
   significantly improved overall survival when treated with platinum
   agents [[151]57]. The favorable outcomes in HRD cases may be associated
   to molecular features including high TMB, enhanced immune activity, and
   microsatellite instability [[152]58].

   In addition, we identified at least one potentially actionable genetic
   variant in all the preneoplastic lesions and GC. Variants found in ATM,
   BRAF, and FBXW7 have been associated with sensitivity to actionable
   therapy, and TP53, NRAS, and EGFR with resistance to it. ATM deficiency
   has shown sensitivity to olaparib in some GC cell lines [[153]59], with
   one multicenter study ([154]NCT03829345) in oesophagogastric cancer and
   gastric adenocarcinoma showing promising results [[155]60]. BRAF
   mutations are used to treat melanoma and solid tumors (different than
   GC) with vemurafenib [[156]61, [157]62]. FBXW7 is associated with
   sensitivity to sirolimus in ovarian and endometrial cancer. However,
   TP53 mutations confer poor treatment outcomes and resistance to therapy
   with venetoclax in patients with acute myeloid leukemia [[158]63], and
   NRAS and SMAD4 showed resistance to EGFR inhibitors in lung and
   colorectal cancers [[159]64–66]. Although these findings highlight the
   potential value of these genes in identifying, and developing possible
   targeted treatments in early GC, caution must be taken in the
   interpretation of their potential clinical translation.

   Limitations of this study include but are not limited to the following.
   First there was insufficient sample size to establish statistical
   association with epidemiological variables. Due to the limited number
   of genes analyzed, it was not possible to comprehensively assess the
   number of mutations, TMB, and mutational signature in each patient,
   leading to the grouping of samples by disease histopathological type
   only. Second, the lack of molecular testing to identify H. pylori
   infection and virulence factors prevented a more detailed understanding
   of epidemiological role of this pathogen. CNV events were detected by
   robust bioinformatic methods but were not validated. Other molecular
   mechanisms need to be evaluated, such as epigenetic analysis, which may
   provide alternative pathogenic routes in preneoplastic lesions and
   early GC. Finally, proposed therapy based on mutational data should be
   treated with caution and require further independent replication and
   clinical validation. Despite these limitations, we identified several
   key early genetic alterations acting in preneoplastic lesions and GC
   that may promote cancer progression and may thus be used as possible
   therapeutic targets, as well as for follow-up after surgery, as has
   been shown in liquid biopsy studies [[160]67, [161]68].

Conclusions

   This work provides comprehensive genomic characterization of the
   earliest alterations of preneoplastic and malignant gastric lesions in
   a high-incidence Latin American population. Our findings reveal that at
   the preneoplastic and earliest stages, GC is genetically heterogeneous
   and presents key cancer driving mutations that may participate in
   neoplastic transformation and progression, with 62% of lesions having
   at least one targetable mutation, which presents a significant
   therapeutic opportunity. However, the presence of considerable genetic
   heterogeneity underscores the need for continued molecular
   characterization efforts. Furthermore, our research highlights the
   potential for early detection and disease monitoring using techniques
   such as liquid biopsy analysis. These insights improve our
   understanding of gastric early carcinogenesis and emphasize the
   critical importance of precision medicine approaches in combating GC.

Supplementary Material

   goaf089_Supplementary_Data
   [162]goaf089_supplementary_data.zip^ (5.2MB, zip)

Acknowledgements