Research Paper
Crohn’s disease-associated ATG16L1 T300A genotype is associated with
improved survival in gastric cancer
Changqing Maa,
*, Chad E. Storerb
, Uma Chandranc
, William A. LaFramboised,1
,
Patricia Petroskod,1
, Madison Franka
, Douglas J. Hartmana
, Liron Pantanowitza,2
,
Talin Haritunianse
, Richard D. Headb
, Ta-Chiang Liuf,
*
a
Department of Pathology, University of Pittsburgh School of Medicine, 200 Lothrop Street, A-610, Pittsburgh, PA 15213, United States
b
Department of Genetics, Washington University School of Medicine, Saint Louis, MO 63110, United States
c
Department of Biomedical Informatics, University of Pittsburgh School of Medicine, Pittsburgh, PA 15213, United States
d
UPMC Hillman Cancer Center, Cancer Genomics Facility, Pittsburgh, PA 15232, United States
e
F. Widjaja Family Foundation Inflammatory Bowel and Immunobiology Research Institute, Cedars-Sinai Medical Center, Los Angeles, CA, United States
f
Departments of Pathology and Immunology, Washington University School of Medicine, 660 South Euclid Avenue, Campus Box 8118, Saint Louis, MO 63110,
United States
A R T I C L E I N F O
Article History:
Received 24 November 2020
Revised 30 March 2021
Accepted 31 March 2021
Available online xxx
A B S T R A C T
Background: A non-synonymous single nucleotide polymorphism of the ATG16L1 gene, T300A, is a major
Crohn’s disease (CD) susceptibility allele, and is known to be associated with increased apoptosis induction
in the small intestinal crypt base in CD subjects and mouse models. We hypothesized that ATG16L1 T300A
genotype also correlates with increased tumor apoptosis and therefore could lead to superior clinical outcome
in cancer subjects.
Methods: T300A genotyping by Taqman assay was performed for gastric carcinoma subjects who underwent
resection from two academic medical centers. Transcriptomic analysis was performed by RNA-seq on formalin-fixed
paraffin-embedded cancerous tissue. Tumor apoptosis and autophagy were determined by cleaved
caspase-3 and p62 immunohistochemistry, respectively. The subjects' genotypes were correlated with demographics,
various histopathologic features, transcriptome, and clinical outcome.
Findings: Of the 220 genotyped subjects, 163 (74%) subjects carried the T300A allele(s), including 55 (25%)
homozygous and 108 (49%) heterozygous subjects. The T300A/T300A subjects had superior overall survival
than the other groups. Their tumors were associated with increased CD-like lymphoid aggregates and
increased tumor apoptosis without concurrent increase in tumor mitosis or defective autophagy. Transcriptomic
analysis showed upregulation of WNT/b-catenin signaling and downregulation of PPAR, EGFR, and
inflammatory chemokine pathways in tumors of T300A/T300A subjects.
Interpretation: Gastric carcinoma of subjects with the T300A/T300A genotype is associated with repressed
EGFR and PPAR pathways, increased tumor apoptosis, and improved overall survival. Genotyping gastric cancer
subjects may provide additional insight for clinical stratification.
© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/)
Keywords:
Autophagy
Apoptosis
Genotyping
Transcriptomics
Stratification
1. Introduction
Our knowledge of cancer genetics has centered on oncogenes and
tumor suppressor genes [1,2]. Recent studies have demonstrated that
single nucleotide polymorphisms (SNPs) are also associated with
susceptibility to sporadic cancers [3À5]. To date, genome-wide association
studies (GWAS) have identified nearly a thousand predisposition
variants significantly associated with over thirty malignancies
[5À12]. These alterations have provided insights into new pathways
for tumorigenesis and new targets for therapeutics [5]. Although
most variants demonstrate a modest increase in risk [5], the combined
effects of multiple SNPs are potentially useful in populationbased
cancer risk stratification and prevention programs [13,14].
Likewise, GWAS studies have also identified SNPs associated with
response to chemotherapy [15], treatment-related toxicity [16], and
clinical outcomes in cancer patients [17À19]. Therefore, identifying
SNPs and other genetic alterations that correlate with clinical
* Corresponding authors.
E-mail addresses: mac2@upmc.edu (C. Ma), tliu27@wustl.edu (T.-C. Liu).
1
Present address: Pathology and Laboratory Medicine, Allegheny Health Network,
320 East North Avenue, Pittsburgh, PA 15212.
2
Present address: Department of Pathology & Clinical Labs, University of Michigan,
Ann Arbor, MI 48109.
https://doi.org/10.1016/j.ebiom.2021.103347
2352-3964/© 2021 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
EBioMedicine 67 (2021) 103347
Contents lists available at ScienceDirect
EBioMedicine
journal homepage: www.elsevier.com/locate/ebiom
outcome may provide novel insight into the pathobiology and management
of cancer.
One area of which GWAS has provided insight into disease pathogenesis
which resulted in mechanistic discoveries is immune-associated
disorders [20]. Common SNPs/variants that increase the
susceptibility for inflammatory bowel disease (IBD), asthma, rheumatoid
arthritis, and systemic lupus erythematosus have been identified
[11,20À25], and some have also been associated with increased cancer
risk [26]. Lessons from these informative studies have fostered
functional validations of selected SNPs. Most importantly, since these
SNPs may exert effects in a wide spectrum of tissue and cell types
[27], exploration of the role of these SNPs in other diseases such as
malignancy may yield new hypotheses and/or insights into tumorigenesis
through gene-specific, tissue-specific, or environment-specific
mechanisms that merit testing.
Notably, some of the SNPs identified in immune-related disorders
are also common in non-disease control subjects, suggesting that
evolutionarily these SNPs may have acquired important roles in
maintaining tissue homeostasis [27À29]. One such example is the
ATG16L1 Thr300Ala (T300A) SNP, the most significant susceptibility
SNP in the development of Crohn’s disease (CD) [30]. While originally
described as a key execution member of the autophagy pathway [31],
ATG16L1 is also known to affect multiple cell death pathways, including
apoptosis [32] and necroptosis [33]. We and others have shown
that Atg16l1 T300A mice showed defects in secretory cell lineages in
the gut, resulting in altered immunity that affects pathogen clearance
[29,34,35]. Notably, CD subjects harboring ATG16L1 T300A have more
aggressive disease course after resection [32,36,37]. Interestingly,
ATG16L1 T300A also confers protection from Escherichia coli infection
[28], but promotes Helicobacter pylori infection [38]. Thus, it is likely
that the effect of ATG16L1 T300A is organ- and disease-specific.
We previously showed that CD subjects and mice with ATG16L1
T300A possessed increased apoptosis (through repression of the
PPAR-g pathway) in the small intestinal crypt base [32]. Given the
role of apoptosis as a major cell death pathway employed to target
cancers [39], we hypothesized that ATG16L1 T300A may also confer
increased tumor apoptosis, which could lead to improved survival in
cancer. Herein we show that in gastric cancer, the T300A genotype is
associated with unique histologic features, increased cancer
apoptosis without concomitant increase in cancer proliferation or
autophagy deficiency, and importantly, superior overall survival.
2. Methods
2.1. Gastric cancer cases
Gastric cancer cases resected between 2008 and 2012 were retrospectively
collected by searching the surgical pathology archives of
the University of Pittsburgh Medical Center (UPMC). Also included
were gastric cancer cases resected between 1999 and 2013 from the
Barnes-Jewish Hospital/Washington University. These cases were
first described by Olsen et al. [40]. The University of Pittsburgh Medical
Center and the Barnes-Jewish Hospital/Washington University are
both large academic medical centers well-known in the United
States. Cases from both institutions were collected in the aforementioned
time frame to ensure the post-surgical follow-up period is
long enough (at least 5 years) for survival analysis. For each subject /
case from the UPMC, the electronic medical records were reviewed
and the following demographic, clinical, and pathologic information
were retrieved: gender, race, age at operation, pathologic and clinical
stage, lymphovascular invasion, perineural invasion, status of surgical
resection margins, pre- and post-surgery treatments if received, history
of Helicobacter pylori infection, and survival. For cases from the
Barnes-Jewish Hospital, the aforementioned demographic, clinical,
and pathologic information were updated and only cases with up-todate
clinical follow-up information at the time of data collection
were included in this study.
2.2. Ethics statement
This study was approved by the Institutional Review Boards of the
University of Pittsburgh (STUDY20050115) and the Washington University
(201301164). Informed consent from all participants was
waived by the IRB committees.
2.3. Histologic evaluation
Tissue sections of the carcinoma for each case were reviewed
without knowledge of the T300A genotype of each subject. Histologic
subtyping for each tumor was performed using criteria of the Lauren
classification system [41]. Specifically, a tumor is classified as intestinal-type
if the carcinoma demonstrates intestinal differentiation
with glandular / tubular structure formation. A tumor is classified as
diffuse-type if the carcinoma cells demonstrate lack of cohesiveness
(in other words lack of adhesion with each other) and infiltrating the
gastric wall as single cells or small clusters of tumor cells. A tumor is
classified as mixed-type if histologic features of both intestinal-type
and diffuse-type carcinomas are observed in the same tumor.
Histologic grading of each case was performed according to criteria
in the eighth edition of the American Joint Committee on Cancer
recommendation on cancer staging [42]. This histologic grading system
is based on the extent of glandular differentiation. In particular,
a well-differentiated adenocarcinoma (G1) is defined as a tumor with
greater than 95% of the tumor composed of glands. A moderately-differentiated
adenocarcinoma (G2) is a tumor with 50% to 95% of the
tumor composed of glands. A poorly-differentiated adenocarcinoma
(G3) is a tumor with 49% or less of the tumor composed of glands. In
this study, a tumor with mixed differentiation such as a mixed adenoneuroendocrine
carcinoma was classified as histologic grade X (Gx).
The presence of a CD-like lymphoid reaction was evaluated as follows.
For each case, all tumor sections were evaluated at low magnification
to find one section with the most intra- and peri-tumoral
lymphoid aggregates. In this tumor section, the number of intra- and
peri-tumoral lymphoid aggregates were counted in 3 consecutive
microscopic fields using 10£ objective lens and 10£ eye piece (the
Research in context
Evidence before this study
Single nucleotide polymorphisms can be associated with cancer
susceptibility and/or prognosis. We previously showed that the
ATG16L1 T300A variant, a major Crohn’s disease susceptibility
allele, induces apoptosis (through repression of the PPAR-g
pathway) in the small intestinal crypt base in CD subjects. Notably,
CD subjects with the T300A/T300A genotype have more
aggressive disease course after surgical resection, whereas
T300A/T300A subjects with colorectal carcinoma demonstrate
superior prognosis.
Added value of this study
Our results demonstrate that the T300A/T300A genotype is
associated with superior overall survival, increased tumor apoptosis
and repressed EGFR and PPAR pathways in gastric
cancer.
Implications of all the available evidence
Genotyping of genes and variants that are not involved in oncogenic
or tumor suppressor pathways may provide insight in
stratifying gastric cancer subjects.
2 C. Ma et al. / EBioMedicine 67 (2021) 103347
use of both in combination yields a 100£ field of examination, which
corresponds to a surface area of 3 ¢ 8 mm2
using an Olympus BX46
microscope). When on average the number of lymphoid aggregates
was one or more (mean number of lymphoid aggregates per
100 £ field  1), the tumor was scored as positive for CD-like lymphoid
reaction.
Mitotic activity of each tumor was evaluated by counting the
number of mitotic figures in the most mitotically active area in the
tumor. Briefly, for each case, all tumor sections were first evaluated
at low magnification to find a “hotspot” area with the most mitotic
activity. In the “hotspot” area, the number of mitotic figures was then
counted in 3 consecutive high-power fields (one high-power
field = 400 £ [40 £ objective lens and 10£ eye piece], which correspond
to a surface area of 0 ¢ 24 mm2
using an Olympus BX46 microscope).
The mean number of mitotic figures in one high-power field
(400£) was calculated for each tumor.
2.4. Genotyping
Genomic DNA was extracted from formalin-fixed paraffin-embedded
(FFPE) tissue using the QIAamp DNA FFPE tissue kit (Qiagen Inc.,
Valencia, CA, catalog number: 56404) and subsequently used for
T300A genotyping with TaqManÒ
SNP Genotyping Assay (Thermo
Fisher Scientific, catalog number: 4351379) following the manufacturer’s
instructions.
2.5. Immunohistochemistry and in-situ hybridization
For cleaved caspase-3 immunohistochemistry, unstained slides
were deparaffinized, followed by antigen retrieval using Trilogy (Cell
Marque, catalog number: 920P-09). The slides were quenched with
10% H2O2 in methanol for 10 min, before being blocked with 1%
bovine serum albumin and 0.1% Trizol in phosphate buffer saline for
1 h. Thereafter, primary antibody (1:200; Cell Signaling, RRID:
AB_2069869) was applied overnight. After overnight incubation, goat
anti-rabbit biotinylated secondary antibody (1:200; ThermoFisher
Scientific, catalog number: 31820) was applied for 45 min. Signal
detection was then performed by using Vectastain Elite ABC kit (Vector,
catalog number: PK-6100) and DAB (Vector, RRID: AB_2336520).
Immunohistochemistry stains with CD4 (clone SP35, Ventana, catalog
number: 790-4423), CD8 (clone CD8/144B, DAKO, catalog number:
M7103), CD56 (clone 123C3.D5, Ventana, RRID: AB_2335987),
EGFR (clone 3C6, Ventana, RRID: AB_2335974), c-MYC (clone Y69,
Ventana, catalog number: 790-4628), p62 (clone 3/P62 LCK ligand,
BD Biosciences, RRID:AB_398151), MLH1 (clone M1, Ventana, catalog
number: 790-4535), MSH2 (clone G219-1129, Ventana, catalog number:
760-4265), MSH6 (clone 44, Ventana, catalog number: 790-
4455), and PMS2 (clone EPR3947, Cell Marque, catalog number:
288M-18) antibodies were performed on unstained 4 mm tissue sections
using the Ventana XT automated stainer (Ventana Medical Systems)
according to the manufacturer’s recommendations. All cases
were tested for Epstein À Barr virus (EBV) by in-situ hybridization
using pre-diluted probes targeting EBV encoded early RNA (EBER,
Ventana, catalog number: 760-1209) using the Ventana XT automated
stainer. All cases tested for EBER were also tested for the integrity
of total RNA using RNA control probe (Leica, catalog number:
RS7760) and all showed adequate total RNA preservation.
2.6. Quantitative digital image analysis of CD8-positive or CD4-positive
T cell density and CD56-positive NK cell density
The CD4, CD8, and CD56 stained slides were digitized using an
Aperio AT2 scanner (Leica Biosystems, Buffalo Grove, IL) at 40x magnification.
The CD4 immunostained slide and CD8 stained slide for
each case were manually annotated to outline the areas of invasive
adenocarcinoma. Non-neoplastic gastric tissue and areas of necrosis
were specifically excluded from the area of analysis. The Aperio
nuclear v9 algorithm, a component of the commercially available
Leica/Aperio image analysis platform, was used to count the CD4- or
CD8-stained cells as previously described [43]. The CD4-positive or
CD8-positive T cell density was determined by dividing the number
of CD4-positive or CD8-positive cells by the area (measured in mm2
)
examined. Densities (number of cells / mm2
) of CD56-positive, CD4positive,
or CD8-positive cells in peri- and intratumoral lymphoid
aggregates were additionally determined using the same algorithm.
2.7. Transcriptomic analysis
For each tumor included in transcriptomic analysis, a representative
FFPE tumor tissue block was selected. The tumor area desired for
transcriptomic analysis was manually circled on the hematoxylin and
eosin stained tissue section. Manual microdissection of the desired
tumor area was then performed on 6 to 10 unstained tumor tissue sections
at 10 mm thickness for each tumor to ensure that only area of
viable tumor but not necrosis or uninvolved tissue was included in
transcriptomic analysis (Supplementary Fig. 1). Total RNA was
extracted from manually dissected tumor tissue using the Qiagen miRNeasy
mini kit (Qiagen Inc., catalog number: 74104) according to the
manufacturer’s recommendations. RNA library was prepared using the
TruSeq RNA ACCESS library preparation kit (Illumina, San Diego, CA,
catalog number: 15049525). Transcriptomic analysis was performed
on the NextSeq 500 system according to the manufacturer’s recommendations
(please see Supplementary Method for details).
Quality of the RNAseq results of all 47 cases were assessed using
FastQC (v0.11.5). All 47 cases fulfilled the quality control parameters.
Samples were then aligned to the GRCh38 genome with HISAT2
(v2.1.0) [44] and to the human transcriptome with Salmon (0.11.0)
[45]. Alignments were quality assessed with QoRTs (v1.1.8) [46] and
RNASeQC (v1.1.8) [47]. Gene expression was quantified with HTSeq
(v0.9.0) [48]. Data normalization with the weighted trimmed mean
of M-values (TMM) method and differential gene expression between
T300A genotypes were performed with the edgeR R package [49].
BenjaminiÀHochberg method was used for correction of multiple
testing for differential gene expression.
Differentially expressed genes with P < 0 ¢ 01 were further analyzed
by a novel knowledge engine called ‘COmprehensive Multiomics
Platform for Biological InterpretatiOn’ (CompBio: https://
www.percayai.com/CompBio) [50]. CompBio performs a literature
analysis to identify relevant biological processes and pathways represented
by the differentially expressed entities (genes, proteins, miRNA’s,
or metabolites). This is accomplished with an automated
Biological Knowledge Generation Engine that extracts all abstracts
from PubMed that reference entities of interest (or their synonyms),
using contextual language processing and a biological language dictionary
that is not restricted to fixed pathway and ontology knowledge
bases. Conditional probability analysis calculates the statistical
enrichment of biological concepts (processes/pathways) over those
that occur by random sampling. Related concepts built from the list
of differentially expressed entities are further clustered into higherlevel
themes (e.g., biological pathways/processes, cell types and
structures, etc.). The dataset was then cross-referenced with the
T300A/T300A-specific transcriptomic themes identified in mouse
ileum (ArrayExpress database E-MTAB-5707) [32].
2.8. Statistics
Statistical analysis was performed with GraphPad Prism 7 for
Windows Version 7.00 (GraphPad Software, La Jolla, CA) and IBM
SPSS (Release 23.0.0.0). GraphPad Prism 7 was used to perform both
paired and unpaired Student t-tests, MannÀWhitney test, and KruskalÀWallis
test, which were used for comparison of continuous variables.
IBM SPSS was used to perform survival analysis, Two-sided
C. Ma et al. / EBioMedicine 67 (2021) 103347 3
Fisher’s exact test, and Chi-Square test. The latter two tests were used
for the comparison of categorical data. HardyÀWeinberg equilibrium
testing was performed by comparing observed genotype frequencies
with those expected in Caucasian subjects and in African American
subjects by Chi-Square test in R using the package named HardyÀWeinberg
(https://cran.r-project.org/web/packages/HardyWein
berg/HardyWeinberg.pdf). All tests were two-tailed and a P value less
than 0 ¢ 05 (P < 0 ¢ 05) was considered significant. For survival analysis,
overall survival was defined as the duration between surgical
resection and either death or the latest clinical follow-up time. Death
occurring within one month of the initial operation was attributed to
operative mortality and was not included in the survival analysis. Subjects
who were clinical stage IV at the time of surgery, i.e., with distant
metastasis at the time of surgery, were considered never disease free
and hence excluded from the analysis for disease-free survival. Survival
analysis was performed using Cox proportional hazard model
and Kaplan-ÀMeier analysis with log-rank statistics. Potential confounding
factors in univariate and/or multivariate analysis included
race, histologic grade, histologic subtype by Lauren classification, clinical
stage, surgical resection margin, lymphovascular invasion, perineural
invasion, Helicobacter pylori infection, and chemotherapy.
Sequencing files are deposited at Gene Expression Omnibus (GEO)
under accession number GSE152415.
2.9. Role of funding source
The funder (NIH) did not have any role in study design, data collection,
data analyses, interpretation, or writing of report.
3. Results
3.1. Characteristics of the gastric cancer cohort
We retrospectively collected tissue from 220 consecutively
resected gastric cancer subjects from two academic medical centers
for genotyping and histopathology analysis. Among them, 163 (74%)
subjects carried the T300A allele(s), including 55 (25%) homozygous
T300A/T300A and 108 (49%) heterozygous T300A/wild-type (WT)
subjects. Homozygous WT/WT genotype was detected in 57 (26%)
subjects. The flow diagram of the gastric cancer cases analyzed in this
study is shown in Fig. 1, and the demographics and clinicopathologic
characteristics of the 220 genotyped subjects are listed in Table 1.
The T300A allele frequency was 0 ¢ 56 and the WT allele frequency
was 0 ¢ 44 in Caucasian subjects in our cohort. In African American
subjects the allele frequency was 0 ¢ 23 for the T300A allele and 0 ¢ 77
for the WT allele. Allele frequencies of the T300A SNP in our cohorts
were comparable to those reported in public databases [30]. In addition,
genotype frequencies in Caucasian subjects and African American
subjects did not deviate from HardyÀWeinberg equilibrium
(P > 0 ¢ 05 [Chi-Square test]). Since the T300A allele is more frequent
in Caucasians than in African American subjects, homozygous T300A/
T300A genotype was more frequently identified in Caucasians than
other genotypes (P < 0 ¢ 01 [Fisher’s exact test]; Table 1). Only 33
(15%) subjects had either clinical history or active/concurrent Helicobacter
pylori infection. When stratified by genotype, Helicobacter
pylori infection was more often seen in homozygous WT/WT subjects
than in subjects with other genotypes (14/57, 25% versus 11/108, 10%
versus 8/55, 15%). However, this difference was not statistically significant
(P = 0 ¢ 07 [Fisher’s exact test]; Table 1). There was also no
significant difference in the histologic degree of tumor differentiation,
histologic type by the Lauren classification [41], clinical stage,
receipt of pre- and post- surgical treatment, or other histopathologic
parameters between different genotype groups (P > 0 ¢ 05 for all
[Fisher’s exact test]; Table 1).
3.2. Gastric cancer subjects with ATG16L1 T300A/T300A showed
superior prognosis after surgery
We next assessed if the ATG16L1 T300A genotype correlated with
clinical outcome. Among all 220 subjects with confirmed genotypes,
those with the T300A/T300A genotype showed superior overall survival
compared to the other two groups combined (WT/T300A and
WT/WT) by Kaplan-ÀMeier survival analysis with log-rank statistics
(mean, standard error, and 95% confidence interval [CI] of overall survival:
107 months, 14 months, [80À135] versus 64 months, 9
months, [52À76], P = 0 ¢ 035) (Fig. 2a). Using Cox proportional hazards
modeling, however, the T300A/T300A genotype was not an
independent predictor of overall survival for the entire cohort in multivariate
analysis (multivariate analysis hazard ratio and 95% CI:
0 ¢ 77, 0 ¢ 47 À 1 ¢ 26, P = 0 ¢ 31) (Supplementary Table 1).
In order to exclude potential confounding effects on overall survival
introduced by pre- and post-surgical chemotherapy, we next
performed survival analysis in the subset of 126 subjects who
received only surgical treatment. Among them, subjects with T300A/
T300A genotype had significantly longer overall survival than subjects
with WT/T300A and WT/WT genotypes combined using KaplanÀMeier
survival analysis with log-rank statistics (mean, standard
error, and 95% CI: 164 months, 16 months, [132À196] versus 60
months, 5 months, [51À69], P = 0 ¢ 019) (Fig. 2b). Using Cox proportional
hazards modeling, the T300A/T300A genotype was associated
with a significant increase in overall survival by both univariate and
multivariate analysis (multivariate analysis hazard ratio and 95% CI:
0 ¢ 31, 0 ¢ 11À0 ¢ 88, P = 0 ¢ 03) (Table 2).
Approximately one-third (45, 36%) of the 126 subjects with surgery
as the only treatment died during the follow-up period. WT/WT subjects
had significantly higher mortality than subjects with other T300A
genotypes (WT/WT: 19/35, 54% versus WT/T300A: 20/57, 35% versus
T300A/T300A: 6/30, 20%, P = 0 ¢ 02 [Fisher’s exact test]). The percentages
of patients who died of disease progression were similar
between the three groups: 7/19 (37%) in WT/WT, 7/20 (35%) in WT/
T300A, and 2/6 (33%) in T300A/T300A groups (P = 0 ¢ 73 [Fisher’s exact
test]). Likewise, other comorbidities and surgical complications
accounted for death in 12/19 (63%) WT/WT, 13/20 (65%) in WT/T300A,
and 4/6 (67%) in T300A/T300A groups (Supplementary Table 2).
3.3. T300A/T300A genotype was associated with unique histological
features
We and others have previously found that T300A/T300A genotype
was associated with histomorphologic changes in CD subjects and
mouse models [34,51,52], including increased apoptosis induction in
Fig. 1. Flow chart illustrating type of analysis and number of cases. Flow diagram of the
number of subjects genotyped and the number of cases analyzed for histopathologic
features, clinical outcome, and transcriptomes.
4 C. Ma et al. / EBioMedicine 67 (2021) 103347
small intestinal crypt base where Paneth cells and intestinal stem
cells (both are CD-relevant cell types) reside [53]. We therefore
hypothesized that T300A/T300A genotype was also associated with
unique histologic features in gastric cancer. To exclude the possibility
that neoadjuvant therapy may affect histologic features, we performed
histologic assessments of the tumors from subjects who did
not receive neoadjuvant therapies, including subjects who received
no additional treatment either before or after surgery and those who
received only adjuvant treatment. Cases with tumor blocks available
(n = 119) were included in histologic evaluation (Fig. 1). As shown in
Supplementary Table 3, there was no significant difference in Helicobacter
pylori infection, the histologic degree of tumor differentiation,
histologic subtype, clinical stage, or other histologic features between
different genotype groups (P > 0 ¢ 05 for all [Fisher’s exact test]).
In contrast, we found that the T300A/T300A tumors were more
likely to contain intra- and peri-tumoral CD-like lymphoid aggregates
than homozygous WT/WT tumors (25/28, 89% versus 21/32, 66%,
P = 0 ¢ 037 [Fisher’s exact test]) (Fig. 3a, b, and Table 3). We further
analyzed the density of lymphocytes not associated with lymphoid
aggregates, using an unbiased automated image analysis system [43].
There was no difference in the density of CD8-positive or CD4-positive
T cells at the tumor invasive front (tumor-stroma interface)
(P = 0 ¢ 60 and 0 ¢ 89, respectively, [Kruskal-ÀWallis test]) or within
the tumors (P = 0 ¢ 72 and 0 ¢ 61, respectively, [Kruskal-ÀWallis test])
(Supplementary Table 4 and Supplementary Fig. 2). We additionally
analyzed the density of CD4-positive T cells, CD8-positive T cells, and
CD56-positive NK cells in peri- and intratumoral lymphoid aggregates
using the same automatic image analysis system [43]. These
lymphoid aggregates contained similar proportions of CD4-positive T
cells and CD8-positive T cells (mean and standard deviation: CD4positive
T cell density: 1587 § 618 versus CD8-positive T cell density:
1556 § 611; P = 0 ¢ 78 [paired t-test]). The ratio of CD4-positive cell
density and CD8-positive cell density was not associated with tumor
T300A genotypes (mean and standard deviation of CD4 cell density /
CD8 cell density: WT/WT: 1 ¢ 06 § 0 ¢ 38 versus T300A/T300A:
1 ¢ 11 § 0 ¢ 76; P = 0 ¢ 63 [Mann-ÀWhitney test]) (Supplementray
Figs. 3a, c, and d). In contrast CD56-positive NK cell was not a significant
component in peri- and intratumoral lymphoid aggregates (Supplementray
Figs. 3b and e).
3.4. T300A/T300A genotype was associated with increased tumor
apoptosis
Given the role of ATG16L1 T300A in mediating various cell death
pathways, predominantly autophagy and apoptosis, we next determined
if proliferation, autophagy, and apoptosis were differentially
correlated with subject genotypes. We found that there was no difference
in the mitotic activities in the tumors of the T300A/T300A
group compared to the others (P = 0 ¢ 64 [Kruskal-ÀWallis test]) (Supplementary
Figs. 4a, b, and Supplementary Table 4). Likewise, as a
Table 1
Demographics and clinicopathologic characteristics stratified by T300A genotype.
All cases(N = 220) WT/WT(N = 57) WT/T300A(N = 108) T300A/T300A(N = 55) P
Age (year; median, [IQR]) 73 (19) 75 (15) 71 (20) 72 (22) 0.40
Gender (n, %)
Female 96 (44) 28 (49) 48 (44) 20 (36) 0.38
Male 124 (56) 29 (51) 60 (56) 35 (64)
Race (n, %) N = 167 N = 41 N = 82 N = 44
Caucasian 135 (81) 24 (59) 71 (87) 40 (91) < 0.01
African American 20 (12) 13 (32) 5 (6) 2 (5)
Others 12 (7) 4 (10) 6 (7) 2 (5)
Histologic type (n, %)
Intestinal 104 (47) 31 (54) 45 (42) 28 (51) 0.48
Diffuse 84 (40) 18 (32) 49 (45) 20 (36)
Mixed and other types 29 (13) 8 (14) 14 (13) 7 (13)
Histologic grade (n, %)
G1 14 (6) 2 (4) 7 (6) 5 (9) 0.14
G2 57 (26) 21 (37) 20 (19) 16 (29)
G3 146 (66) 33 (58) 79 (73) 34 (62)
Gx 3 (1) 1 (2) 2 (2) 0
Clinical stage (n, %)
I 72 (33) 17 (30) 31 (29) 24 (44) 0.10
II 58 (26) 20 (35) 25 (23) 13 (24)
III 62 (28) 16 (28) 32 (30) 14 (25)
IV 26 (12) 3 (5) 19 (18) 4 (7)
Unknown 2 (1) 1 (2) 1 (1) 0
Positive resection margin (n, %) 34 (15) 8 (14) 20 (19) 6 (11) 0.44
Positive lymphovascular invasion (n, %) 135 (61) 33 (58) 71 (66) 31 (56) 0.31
Positive perineural invasion (n, %) 93 (42) 21 (37) 52 (48) 20 (36) 0.29
Helicobacter pylori infection (n, %) 33 (15) 14 (25) 11 (10) 8 (15) 0.07
Vital status (at the time of data collection) (n, %)
Alive 93 (43) 21 (37) 43 (40) 29 (53) 0.24
Death 121 (55) 34 (60) 61 (56) 26 (47)
Lost to follow-up 6 (3) 2 (4) 4 (4) 0
Overall follow-up time (month; median, [IQR]) 25 (46) 21 (51) 20 (41) 37 (53) 0.02
Recurrence (at the time of data collection) (n, %) N = 178 N = 49 N = 81 N = 48
No 116 (65) 30 (61) 56 (69) 31 (65) 0.73
Yes 62 (35) 19 (39) 26 (32) 17 (35)
Disease free time (month; median, [IQR]) 23 (46) 18 (45) 22 (48) 25 (44) 0.37
Chemotherapy (n, %)
None 126 (57) 36 (63) 60 (56) 30 (55) 0.79
Neoadjuvant only 13 (6) 2 (4) 7 (6) 4 (7)
Adjuvant only 46 (21) 13 (23) 23 (21) 10 (18)
Both 35 (16) 6 (11) 18 (17) 11 (20)
Abbreviations: WT: wild-type; IQR: interquartile range; G: grade.
Kruskal-ÀWallis test was used for comparison of continuous variables. Fisher’s Exact Test was used for the comparison of categorical data.
C. Ma et al. / EBioMedicine 67 (2021) 103347 5
readout for autophagy activation, the prevalence of p62-positive
immunohistochemical staining was similar between all groups
(P = 0 ¢ 27 [Chi-Square test]) (Supplementary Figs 4c, d, and Supplementary
Table 4). In contrast, the T300A/T300A tumors showed significantly
increased intratumoral apoptosis compared with tumors of
other genotypes (P = 0 ¢ 04 [Kruskal-ÀWallis test]) by cleaved-caspase
3 immunohistochemistry (Figs. 3c, d, e, f, and Table 3). Therefore, the
T300A/T300A genotype was associated with increased intratumoral
apoptosis without concurrent changes in autophagy activation or
mitosis.
3.5. Gastric cancer from T300A/T300A subjects showed unique
transcriptomic signatures
To further characterize the pathways that may be involved in
mediating the apoptosis induction in the T300A/T300A tumors, we
performed global transcriptomics on a subset of 47 tumors that did
not receive neoadjuvant therapies (including 13 WT/WT,
17 WT/T300A, and 17 T300A/T300A cases) (Fig. 1). Of note, since gastric
carcinoma with DNA mismatch repair (MMR) protein deficiency
Fig. 2. Gastric cancer subjects with T300A/T300A genotype had superior overall survival. Kaplan-ÀMeier survival analysis using log-rank statistics comparing the overall survival of
subjects with gastric carcinoma stratified by T300A genotype in the entire cohort (n = 220, P = 0 ¢ 035 [log-rank]) (a) and in the subset of subjects without additional pre- or post-surgical
treatment (n = 126, P = 0 ¢ 019 [log-rank]) (b).
Table 2
Univariate and multivariate analysis of overall survival in subjects without additional
treatment (N = 126) by Cox regression.
No additional
treatment cases
(N = 126)
Univariate analysis Multivariate analysis
P HR 95% CI P HR 95% CI
T300A/T300A 0.02 0.37 0.16À0.88 0.03 0.31 0.11À0.88
African American 0.31 0.61 0.24À1.58
High histologic
grade
0.71 0.89 0.49À1.64
Diffuse type
histology
0.59 0.82 0.40À1.68
Stage III & IV < 0.001 6.0 2.98À12.24 0.02 3.0 1.16À7.74
Positive surgical
resection margin
0.01 2.85 1.24À6.56 0.96 0.97 0.32À2.94
Lymphovascular
invasion
< 0.001 5.45 2.57À11.57 < 0.001 4.11 1.86À9.07
Perineural invasion 0.04 1.96 1.04À3.72 0.71 1.17 0.52À2.61
Helicobacter pylori
infection
0.51 0.74 0.31À1.78
Abbreviations: HR: hazard ratio; CI: confidence interval.
6 C. Ma et al. / EBioMedicine 67 (2021) 103347
Fig. 3. Tumors from T300A/T300A subjects showed more lymphoid aggregates and apoptosis. (a) T300A/T300A tumors were more likely to have intra- and peritumoral lymphoid aggregates
than WT/WT tumors (P = 0 ¢ 037 [Fisher’s exact test]) (WT/WT n = 32, WT/T300A n = 55, T300A/T300A n = 28). (b) Representative photomicrograph of a gastric adenocarcinoma
with intra- and peritumoral lymphoid aggregates (arrows). (c) Representative photomicrograph of a WT/WT tumor with 6% of cleaved-caspase 3-positive (CC3+) tumor cells
(arrows) / 400£ field on average in 5 representative fields. (d) Representative photomicrograph of a WT/T300A tumor with 6% of CC3+ tumor cells (arrows) / 400£ field on average
in 5 representative fields. (e) Representative photomicrograph of a T300A/T300A tumor with 10% of CC3+ tumor cells (arrows) / 400£ field on average in 5 representative fields. (f)
T300A/T300A tumors had significantly increased CC3+ tumor cells compared with tumors of the other two genotypes (P = 0 ¢ 04 by ANOVA) (WT/WT n = 30, WT/T300A n = 30,
T300A/T300A n = 26). Error bars represent mean with 95% confidence interval. Scale bars: b-200 mm; c, d, e-50 mm.
Table 3
Tumor immune microenviroment and apoptosis by cleaved-caspase3 immunoistochemistry stratified by T300A genotype.
WT/WT WT/T300A T300A/T300A P1
P2
P3
Peri- and intra-tumoral lymphoid aggregates (n, %) N = 32 N = 55 N = 28
Negative (< 1 / 100 £ field) 11 (34) 10 (18) 3 (11) 0.06 0.18 0.037
Positive ( 1 / 100 £ field) 21 (66) 45 (82) 25 (89)
No. of cases stained for CC3 N = 30 N = 30 N = 26
% of CC3-positive tumor cells / 400£ field (median, [IQR]) 4.7 (5.1) 5.2 (3.5) 6.8 (5.5) 0.04 0.01 0.04
Kruskal-ÀWallis test and Mann-ÀWhitney test were used for comparison of continuous variables. Fisher’s exact test was used for the
comparison of categorical data.
Abbreviations: WT: wild-type; IQR: interquartile range; CC3: cleaved-caspase 3.
1
P-value for the comparison between all three groups.
2
P-value for the comparison between WT/WT and WT/T300A combined and T300A/T300A (WT/WT & WT/T300A vs. T300A/
T300A).
3
P-value for the comparison between WT/WT and T300A/T300A.
C. Ma et al. / EBioMedicine 67 (2021) 103347 7
/microsatellite instability (MSI) or Epstein - Barr virus (EBV) infection
have distinct molecular features [54], tumors that were MMR deficient/MSI
or EBV-positive were also excluded from transcriptomic
analysis. Manual microdissection was also performed to ensure that
only area of viable tumor and not necrosis or uninvolved tissue was
submitted for bulk transcriptomic analysis for each tumor to minimize
potential transcriptomic variations caused by differences in cell
types (Supplementary Fig. 1). Transcriptomic analysis yielded 487
significantly expressed genes associated with the T300A/T300A genotype,
including 296 upregulated genes and 191 downregulated genes
(P < 0 ¢ 001) (Fig. 4a and Supplementary Table 5). These were used
subsequently in pathway analysis.
Using a novel biological knowledge generation engine (CompBio)
[50], we identified biological processes that were associated with
cancers from subjects with different genotypes. The T300A/T300A
tumors showed upregulation of genes involved in the following
themes: embryonic stem cell reprogramming, immunoglobulin,
WNT/b-catenin signaling, and microvessel density (Fig. 4b and Supplementary
Fig. 5). In addition, these tumors also showed repression
of expression of genes involved in the following transcriptomic
themes: PPAR pathway, EGFR signaling, and inflammatory chemokines
(CCL2, CXCL2, CCL13, CCL3l3, IL-6), among others (Fig. 4c and
Supplementary Fig. 6). Previous studies demonstrated that the T300A
mutation is associated with type I interferon (IFN) activation in colon
cancer cells in vitro [55] and in mouse model [29]. A type I IFN mRNA
expression signature was not detected in T300A/T300A tumors in our
cohort (Supplementary Fig. 7 and Supplementary Table 6).
Immunohistochemistry study of selected genes with up regulated
(c-MYC) and down regulated (EGFR) RNA expression in T300A/T300A
tumors demonstrated strong c-MYC staining but a lack of EGFR staining
in T300A/T300A tumors (Supplementary Figs. 8 and 9). In contrast,
WT/WT or WT/T300A tumors showed strong EGFR staining
without c-MYC staining. These findings support that gene expression
alterations observed from transcriptomic analysis are associated with
tumor T300A genotypes.
Finally, to understand how much the transcriptomic themes identified
are generalizable between different tissue types, we also compared
the data to our previously published dataset where fullthickness
ileal tissue between T300A/T300A and WT/WT mice were
compared [32]. Using CompBio, we identified several common transcriptomic
themes between the two datasets. Most of the transcriptomic
themes identified as upregulated in the T300A/T300A gastric
cancers were also seen in the T300A/T300A mouse ileum (Supplementary
Fig. 10a) whereas themes associated with mesenchymal
stem cells, retinoic acid signaling, chemokines, and PPAR signaling
were common downregulated transcriptomic themes between the
T300A/T300A gastric cancers and T300A/T300A mouse ileum (Supplementary
Fig. 10b). Overlapping genes in association with the
T300A/T300A between the two datasets are shown in Supplementary
Tables 7 and 8. Therefore, despite the species and organ differences,
we identified common transcriptomic themes associated with the
T300A/T300A genotype.
4. Discussion
Recent advances in cancer genetics have focused on genes whose
functions are beyond the roles of oncogenes and tumor suppressor
genes. In particular, genetic studies from immune-associated diseases
have shown that common SNPs can be associated with disease-relevant
pathophysiologic changes. In the current study, we hypothesized
that a well-studied SNP associated with CD susceptibility,
ATG16L1 T300A, may be associated with survival difference in cancer
subjects. We showed that the T300A/T300A genotype was associated
with superior overall survival in gastric cancer subjects. Histologically,
the T300A/T300A genotype was associated with CD-like lymphoid
aggregates and prominent cancer cell apoptosis. Finally, the
molecular signatures associated with these tumors provided possible
mechanisms by which the T300A allele functions.
In contrast to its survival benefit in gastric cancer demonstrated in
our study, the T300A/T300A genotype is associated with more
aggressive clinical phenotype and disease course in CD [32,56]. The
dichotomy may be due to the biologic processes involved in different
disease contexts. In CD, increased crypt base apoptosis leads to Paneth
cell defect, resulting in diminished innate immunity and dysbiosis,
whereas, in gastric cancer, increased tumor apoptosis without
concomitant increase in tumor proliferation leads to improved survival.
This is supported by a recent study demonstrating the T300A
phenotype differences are stimulation (environmental factors)- and
cell type-dependent [57]. On the other hand, survival benefit of the
T300A/T300A genotype has also been shown in colorectal and nonFig.
4. Transcriptomic analysis identified gene signatures associated with T300A/T300A
genotype in gastric cancer. (a) Intensity plot of normalized gene expression level for
487 significantly expressed genes (P < 0 ¢ 01) associated with the T300A/T300A genotype
(WT/WT & WT/T300A n = 30, T300A/T300A n = 17). (b) Transcriptomic themes
from genes upregulated in the T300A/T300A tumors performed by CompBio analysis.
(c) Themes from genes downregulated in the T300A/T300A tumors performed by
CompBio analysis.
8 C. Ma et al. / EBioMedicine 67 (2021) 103347
small cell lung cancers [55,58]. Our study suggests that the protective
effect of the T300A may be common to multiple cancer types,
although the mechanisms by which T300A confers improved survival
may be organ-specific. In particular, the T300A/T300A genotype was
not associated with a decrease in clinical stage IV/metastatic gastric
carcinoma in our cohort. More importantly, the main molecular pathways
associated with T300A in gastric carcinoma may be different
from pathways identified in colon cancer. Type I IFN activation was
detected in human colon cancer cell line with T300A mutation [55]
but was not detected in gastric carcinoma with T300A/T300A geno-
type.
Notably, the ATG16L1 T300A variant has been associated with susceptibility
of gastric cancer. In the study by Burada et al., the risk of
developing gastric carcinoma was significantly reduced (odds ratio:
0 ¢ 52, 95% confidence interval 0 ¢ 13À0 ¢ 88, P = 0 ¢ 013) in subjects carrying
the T300A allele compared with those with WT/WT genotype in
a cohort of 350 Romanian subjects [59]. In contrast, the T300A allele
was associated with an increased risk of gastric carcinoma (adjusted
odds ratio: 2 ¢ 38, 95% confidence interval 1 ¢ 34À4 ¢ 24, P = 0 ¢ 003) in
304 ethnic Chinese subjects [60]. These findings suggest that the
ATG16L1 T300A allele may be associated with susceptibility of gastric
cancer but its impact may be population specific. Additionally, studies
performed in Crohn's disease have shown that the genetics underlying
disease prognosis may be independent of the genetics underlying
disease susceptibility [61]. Further validation of the association
between T300A genotype and gastric cancer susceptibility will ideally
require studying large numbers of subjects, as has been performed in
other disease types [3À5,30,61À64]. However, our findings together
with results in colon and non-small cell lung cancers may imply a
survival benefit of the T300A allele against carcinoma. In addition, we
feel, this protective effect may be generalizable to wider populations
given the relatively large number of subjects from different geographic
areas included in our study (St. Louis and Pittsburgh, U.S.)
and studies of colorectal carcinoma (n = 460, Chicago, U.S.) [55] and
non-small cell lung cancer (n = 323, China) [58].
A unique feature of our study is the focus on transcriptomics of
tumors that are not associated with known outcome confounders
such as MSI or EBV infection, as these are also known to have distinct
pathogenic pathways [54,65,66]. In addition, we have also focused
our analysis on subjects who did not receive pre-surgery treatment,
such that the histopathologic and transcriptomic readouts were not
affected by other therapeutic modalities.
Our transcriptomic study provides insight into the etiopathogenesis
by which SNP variants of autophagy genes may modulate
tumor biology. Importantly, despite the difference in species and
organs, we were able to validate key transcriptomic themes in
the gastric cancer dataset specific to T300A/T300A genotype using
our previously published mouse model [32]. The correlation
between the T300A/T300A genotype and repressed PPAR signaling
is also in keeping with our previous study in CD patients and
mouse model which showed that hosts with T300A were prone
to have repressed PPAR-g signaling [32]. It further validates our
previous mouse model by which repressed PPAR signaling
directly links to increased apoptosis [32]. Our data is also consistent
with a previous study which showed that activated PPAR signaling
negatively correlated with prognosis in gastric cancer
patients, possibly through regulating lipid and fatty acid metabolism
[67]. Therefore, one possible mechanism by which T300A
confers survival benefit in gastric cancer is through repression of
PPAR pathway. As PPAR agonists are being developed for gastric
cancer patients [67,68], our study provides a unique insight into
the patient population that may benefit most from this therapy.
Potential limitations of our study include the retrospective study
design, and the lack of in vitro and/or in vivo studies to further define
the role of the identified molecular pathways. Prospective studies
with a larger cohort with more complete clinical data and in vitro/in
vivo studies for mechanisms are needed to validate the findings
described in our study.
In summary, we showed that CD susceptibility allele ATG16L1
T300A was associated with improved survival in gastric cancer subjects,
in contrast to its prognostic implication in CD. As the T300A
genotype is also prevalent in non-IBD subjects [30], our data suggests
a possible evolutionary benefit in the setting of non-IBD. Furthermore,
genotyping of genes that are not involved in oncogenic or
tumor suppressor pathways may provide insight in stratifying sub-
jects.
Contributors
CM and TCL designed the study. CM, WAL, PP, MF, DJH, and LP
acquired data. CM, CS, UC, RH, TH, and TCL analyzed data. CM and
TCL verified the data. CM and TCL drafted the manuscript.
All authors reviewed and approved the manuscript.
Data sharing statement
The RNA-seq data generated in this manuscript are available at
NCBI’s Gene Expression Omnibus and publicly accessible through
GEO Series accession number GSE152415 (https://www.ncbi.nlm.nih.
gov/geo/query/acc.cgi?acc=GSE152415).
Declaration of Competing Interest
R.D. Head and C. E. Storer may receive royalty income based on
the CompBio technology developed by R.D. Head and licensed by
Washington University to PercayAI.
L. Pantanowitz is a member of the medical advisory board for Ibex
and reports personal fees from Ibex.
D.J. Hartman reports personal fees from Genae (adjudication
work) and Up-To-Date (royalty for educational content).
The other authors have declared that no conflict of interest exists.
Acknowledgments
The authors wish to thank members of the In Situ Hybridization
and Developmental Laboratory of the Department of Pathology, University
of Pittsburgh for excellent technical support.
The authors wish to thank members of the Image Analysis Lab of
the Department of Pathology, University of Pittsburgh, especially Ms.
Lindsey Seigh, Mr. Matthew O’Leary, and Mr. Jon Duboy, for excellent
technical support.
This study utilized the University of Pittsburgh Hillman Cancer
Center shared resource facility (Cancer Genomics Facility) supported
in part by award P30CA047904 (Dr. LaFramboise and Dr. R Ferris).
This study used the University of Pittsburgh Health Sciences Core
Research Facilities (HSCRF) Genomics Research Core services.
T-C. L. was funded by NIH grants R01 DK125296 and R01
DK124274.
Supplementary materials
Supplementary material associated with this article can be found
in the online version at doi:10.1016/j.ebiom.2021.103347.
References
[1] Tomasetti C, Marchionni L, Nowak MA, Parmigiani G, Vogelstein B. Only three
driver gene mutations are required for the development of lung and colorectal
cancers. Proc Natl Acad Sci USA 2015;112:118–23.
[2] Bailey MH, Tokheim C, Porta-Pardo E, Sengupta S, Bertrand D, Weerasinghe A,
et al. Comprehensive characterization of cancer driver genes and mutations. Cell
2018;173:371–85 e18.
C. Ma et al. / EBioMedicine 67 (2021) 103347 9
[3] Lichtenstein P, Holm NV, Verkasalo PK, Iliadou A, Kaprio J, Koskenvuo M, et al.
Environmental and heritable factors in the causation of cancerÀanalyses of
cohorts of twins from Sweden, Denmark, and Finland. N Engl J Med
2000;343:78–85.
[4] Fletcher O, Houlston RS. Architecture of inherited susceptibility to common cancer.
Nat Rev Cancer 2010;10:353–61.
[5] Sud A, Kinnersley B, Houlston RS. Genome-wide association studies of cancer:
current insights and future perspectives. Nat Rev Cancer 2017;17:692–704.
[6] Abnet CC, Freedman ND, Hu N, Wang Z, Yu K, Shu XO, et al. A shared susceptibility
locus in PLCE1 at 10q23 for gastric adenocarcinoma and esophageal squamous
cell carcinoma. Nat Genet 2010;42:764–8.
[7] Helgason H, Rafnar T, Olafsdottir HS, Jonasson JG, Sigurdsson A, Stacey SN, et al.
Loss-of-function variants in ATM confer risk of gastric cancer. Nat Genet
2015;47:906–10.
[8] Shi Y, Hu Z, Wu C, Dai J, Li H, Dong J, et al. A genome-wide association study identifies
new susceptibility loci for non-cardia gastric cancer at 3q13.31 and 5p13.1.
Nat Genet 2011;43:1215–8.
[9] Mocellin S, Verdi D, Pooley KA, Nitti D. Genetic variation and gastric cancer risk: a
field synopsis and meta-analysis. Gut 2015;64:1209–19.
[10] Milne RL, Kuchenbaecker KB, Michailidou K, Beesley J, Kar S, Lindstr€om S, et al.
Identification of ten variants associated with risk of estrogen-receptor-negative
breast cancer. Nat Genet 2017;49:1767–78.
[11] Noguchi E, Sakamoto H, Hirota T, Ochiai K, Imoto Y, Sakashita M, et al. Genomewide
association study identifies HLA-DP as a susceptibility gene for pediatric
asthma in Asian populations. PLoS Genet 2011;7:e1002170.
[12] Huang KL, Mashl RJ, Wu Y, Ritter DI, Wang J, Oh C, et al. Pathogenic germline variants
in 10,389 adult cancers. Cell 2018;173:355–70.
[13] Lecarpentier J, Silvestri V, Kuchenbaecker KB, Barrowdale D, Dennis J, McGuffog L,
et al. Prediction of breast and prostate cancer risks in male BRCA1 and BRCA2
mutation carriers using polygenic risk scores. J Clin Oncol 2017;35:2240–50.
[14] Shi Z, Yu H, Wu Y, Lin X, Bao Q, Jia H, et al. Systematic evaluation of cancer-specific
genetic risk score for 11 types of cancer in the cancer genome atlas and electronic
medical records and genomics cohorts. Cancer Med 2019;8:3196–205.
[15] Johnson N, De Ieso P, Migliorini G, Orr N, Broderick P, Catovsky D, et al. Cytochrome
P450 allele CYP3A7*1C associates with adverse outcomes in chronic lymphocytic
leukemia, breast, and lung cancer. Cancer Res 2016;76:1485–93.
[16] Aminkeng F, Bhavsar AP, Visscher H, Rassekh SR, Li Y, Lee JW, et al. A coding variant
in RARG confers susceptibility to anthracycline-induced cardiotoxicity in
childhood cancer. Nat Genet 2015;47:1079–84.
[17] Fachal L, Gomez-Caama~no A, Barnett GC, Peleteiro P, Carballo AM, Calvo-Crespo P,
et al. A three-stage genome-wide association study identifies a susceptibility
locus for late radiotherapy toxicity at 2q24.1. Nat Genet 2014;46:891–4.
[18] McGranahan N, Rosenthal R, Hiley CT, Rowan AJ, Watkins TBK, Wilson GA, et al.
Allele-specific HLA loss and immune escape in lung cancer evolution. Cell
2017;171:1259–71 e11.
[19] Chatrath A, Przanowska R, Kiran S, Su Z, Saha S, Wilson B, et al. The pan-cancer
landscape of prognostic germline variants in 10,582 patients. Genome Med
2020;12:15.
[20] Jostins L, Ripke S, Weersma RK, Duerr RH, McGovern DP, Hui KY, et al. Hostmicrobe
interactions have shaped the genetic architecture of inflammatory bowel
disease. Nature 2012;491:119–24.
[21] Ferreira MAR, Matheson MC, Duffy DL, Marks GB, Hui J, Le Sou€ef P, et al. Identification
of IL6R and chromosome 11q13.5 as risk loci for asthma. Lancet
2011;378:1006–14 (London, England).
[22] Himes BE, Jiang X, Hu R, Wu AC, Lasky-Su JA, Klanderman BJ, et al. Genome-wide
association analysis in asthma subjects identifies SPATS2L as a novel bronchodilator
response gene. PLoS Genet 2012;8:e1002824.
[23] Raychaudhuri S, Thomson BP, Remmers EF, Eyre S, Hinks A, Guiducci C, et al.
Genetic variants at CD28, PRDM1 and CD2/CD58 are associated with rheumatoid
arthritis risk. Nat Genet 2009;41:1313–8.
[24] Bønnelykke K, Matheson MC, Pers TH, Granell R, Strachan DP, Alves AC, et al.
Meta-analysis of genome-wide association studies identifies ten loci influencing
allergic sensitization. Nat Genet 2013;45:902–6.
[25] Han JW, Zheng HF, Cui Y, Sun LD, Ye DQ, Hu Z, et al. Genome-wide association
study in a Chinese Han population identifies nine new susceptibility loci for systemic
lupus erythematosus. Nat Genet 2009;41:1234–7.
[26] Orlando G, Law PJ, Palin K, Tuupanen S, Gylfe A, Hanninen UA, et al. Variation at
2q35 (PNKD and TMBIM1) influences colorectal cancer risk and identifies a pleiotropic
effect with inflammatory bowel disease. Hum Mol Genet 2016;25:2349–
59.
[27] Liu TC, Stappenbeck TS. Genetics and pathogenesis of inflammatory bowel disease.
Annu Rev Pathol 2016;11:127–48.
[28] Wang C, Mendonsa GR, Symington JW, Zhang Q, Cadwell K, Virgin HW, et al.
Atg16L1 deficiency confers protection from uropathogenic Escherichia coli infection
in vivo. Proc Natl Acad Sci USA 2012;109:11008–13.
[29] Marchiando AM, Ramanan D, Ding Y, Gomez LE, Hubbard-Lucey VM, Maurer K,
et al. A deficiency in the autophagy gene Atg16L1 enhances resistance to enteric
bacterial infection. Cell Host Microbe 2013;14:216–24.
[30] Rioux JD, Xavier RJ, Taylor KD, Silverberg MS, Goyette P, Huett A, et al. Genomewide
association study identifies new susceptibility loci for Crohn disease and
implicates autophagy in disease pathogenesis. Nat Genet 2007;39:596–604.
[31] Choi AM, Ryter SW, Levine B. Autophagy in human health and disease. N Engl J
Med 2013;368:651–62.
[32] Liu T-C, Kern JT, VanDussen KL, Xiong S, Kaiko GE, Wilen CB, et al. Interaction
between smoking and ATG16L1T300A triggers Paneth cell defects in Crohn’s disease.
J Clin Invest 2018;128:5110–22.
[33] Matsuzawa-Ishimoto Y, Shono Y, Gomez LE, Hubbard-Lucey VM, Cammer M, Neil
J, et al. Autophagy protein ATG16L1 prevents necroptosis in the intestinal epithelium.
J Exp Med 2017;214:3687–705.
[34] Lassen KG, Kuballa P, Conway KL, Patel KK, Becker CE, Peloquin JM, et al.
Atg16L1 T300A variant decreases selective autophagy resulting in altered cytokine
signaling and decreased antibacterial defense. Proc Natl Acad Sci USA
2014;111:7741–6.
[35] Zhang H, Zheng L, McGovern DPB, Hamill AM, Ichikawa R, Kanazawa Y, et al. Myeloid
ATG16L1 facilitates host-bacteria interactions in maintaining intestinal
homeostasis. J Immunol 2017;198:2133–46.
[36] VanDussen KL, Liu TC, Li D, Towfic F, Modiano N, Winter R, et al. Genetic variants
synthesize to produce paneth cell phenotypes that define subtypes of Crohn’s disease.
Gastroenterology 2014;146:200–9.
[37] Liu T-C, Naito T, Liu Z, VanDussen KL, Haritunians T, Li D, et al. LRRK2 but not
ATG16L1 is associated with Paneth cell defect in Japanese Crohn’s disease
patients. JCI Insight 2017;2:e91917.
[38] Raju D, Hussey S, Ang M, Terebiznik MR, Sibony M, Galindo-Mata E, et al.
Vacuolating cytotoxin and variants in Atg16L1 that disrupt autophagy promote
Helicobacter pylori infection in humans. Gastroenterology 2012;142:
1160–71.
[39] Ashkenazi A. Targeting the extrinsic apoptotic pathway in cancer: lessons learned
and future directions. J Clin Invest 2015;125:487–9.
[40] Olsen S, Jin L, Fields RC, Yan Y, Ilk Nalbantoglu. Tumor budding in intestinal-type
gastric adenocarcinoma is associated with nodal metastasis and recurrence. Hum
Pathol 2017;68:26–33.
[41] Lauren P. The two histological main types of gastric carcinoma: diffuse and socalled
intestinal-type carcinoma. an attempt at a histo-clinical classification. Acta
Pathol Microbiol Scand 1965;64:31–49.
[42] Shi C, Berlin J, Branton PA, Fitzgibbons PL, Frankel WL, Hofstetter WL, et al. Protocol
for the examination of specimens from patients with carcinoma of the stomach.
2020. https://documents.cap.org/protocols/cp-giupper-stomach-20-4100.
pdf (accessed on December 24, 2020).
[43] Hartman DJ, Ahmad F, Ferris RL, Rimm DL, Pantanowitz L. Utility of CD8 score by
automated quantitative image analysis in head and neck squamous cell carcinoma.
Oral Oncol 2018;86:278–87.
[44] Kim D, Langmead B, Salzberg SL. HISAT: a fast spliced aligner with low memory
requirements. Nat Methods 2015;12:357–60.
[45] Patro R, Duggal G, Love MI, Irizarry RA, Kingsford C. Salmon provides fast
and bias-aware quantification of transcript expression. Nat Methods
2017;14:417–9.
[46] Hartley SW, Mullikin JC. QoRTs: a comprehensive toolset for quality control and
data processing of RNA-Seq experiments. BMC Bioinform 2015;16:224.
[47] DeLuca DS, Levin JZ, Sivachenko A, Fennell T, Nazaire MD, Williams C, et al. RNASeQC:
RNA-seq metrics for quality control and process optimization. Bioinformatics
2012;28:1530–2.
[48] Anders S, Pyl PT, Huber W. HTSeq—a python framework to work with highthroughput
sequencing data. Bioinformatics 2015;31:166–9.
[49] Robinson MD, McCarthy DJ, Smyth GK. edgeR: a bioconductor package for differential
expression analysis of digital gene expression data. Bioinformatics
2010;26:139–40.
[50] Cowardin CA, Ahern PP, Kung VL, Hibberd MC, Cheng J, Guruge JL, et al. Mechanisms
by which sialylated milk oligosaccharides impact bone biology in a gnotobiotic
mouse model of infant undernutrition. Proc Natl Acad Sci USA
2019;116:11988–96.
[51] Cadwell K, Liu JY, Brown SL, Miyoshi H, Loh J, Lennerz JK, et al. A key role for
autophagy and the autophagy gene Atg16l1 in mouse and human intestinal Paneth
cells. Nature 2008;456:259–63.
[52] Patel KK, Miyoshi H, Beatty WL, Head RD, Malvin NP, Cadwell K, et al. Autophagy
proteins control goblet cell function by potentiating reactive oxygen species production.
EMBO J 2013;32:3130–44.
[53] Murthy A, Li Y, Peng I, Reichelt M, Katakam AK, Noubade R, et al. A Crohn’s disease
variant in Atg16l1 enhances its degradation by caspase 3. Nature 2014;506:456–
62.
[54] Bass AJ, Thorsson V, Shmulevich I, Reynolds SM, Miller M, Bernard B, et al. Comprehensive
molecular characterization of gastric adenocarcinoma. Nature
2014;513:202–9.
[55] Grimm WA, Messer JS, Murphy SF, Nero T, Lodolce JP, Weber CR, et al. The
Thr300Ala variant in ATG16L1 is associated with improved survival in human
colorectal cancer and enhanced production of type I interferon. Gut
2016;65:456–64.
[56] Cleynen I, Gonzalez JR, Figueroa C, Franke A, McGovern D, Bortlik M, et al. Genetic
factors conferring an increased susceptibility to develop Crohn’s disease also
influence disease phenotype: results from the IBDchip European Project. Gut
2013;62:1556–65.
[57] Varma M, Kadoki M, Lefkovith A, Conway KL, Gao K, Mohanan V, et al. Cell typeand
stimulation-dependent transcriptional programs regulated by Atg16L1 and
its Crohn’s disease risk variant T300A. J Immunol 2020;205:414–24.
[58] Li Q, Zhou X, Huang T, Tang Y, Liu B, Peng P, et al. The Thr300Ala variant of
ATG16L1 is associated with decreased risk of brain metastasis in patients with
non-small cell lung cancer. Autophagy 2017;13:1053–63.
[59] Burada F, Ciurea ME, Nicoli R, Streata I, Vilcea ID, Rogoveanu I, et al. ATG16L1
T300A polymorphism is correlated with gastric cancer susceptibility. Pathol
Oncol Res 2016;22:317–22.
[60] Castano-Rodriguez N, Kaakoush NO, Goh KL, Fock KM, Mitchell HM. Autophagy in
helicobacter pylori infection and related gastric cancer. Helicobacter
2015;20:353–69.
10 C. Ma et al. / EBioMedicine 67 (2021) 103347
[61] Lee JC, Biasci D, Roberts R, Gearry RB, Mansfield JC, Ahmad T, et al. Genome-wide
association study identifies distinct genetic contributions to prognosis and susceptibility
in Crohn's disease. Nat Genet 2017;49:262–8.
[62] Brant SR, Okou DT, Simpson CL, Cutler DJ, Haritunians T, Bradfield JP, et al.
Genome-wide association study identifies African-specific susceptibility loci in
African Americans with inflammatory bowel disease. Gastroenterology
2017;152:206–17 e2.
[63] Cleynen I, Boucher G, Jostins L, Schumm LP, Zeissig S, Ahmad T, et al. Inherited
determinants of Crohn's disease and ulcerative colitis phenotypes: a genetic association
study. Lancet 2016;387:156–67.
[64] Rivas MA, Avila BE, Koskela J, Huang H, Stevens C, Pirinen M, et al. Insights into
the genetic epidemiology of Crohn's and rare diseases in the Ashkenazi Jewish
population. PLoS Genet 2018;14:e1007329.
[65] Camargo MC, Kim WH, Chiaravalli AM, Kim KM, Corvalan AH, Matsuo K, et al.
Improved survival of gastric cancer with tumour Epstein-Barr virus positivity: an
international pooled analysis. Gut 2014;63:236–43.
[66] Pietrantonio F, Miceli R, Raimondi A, Kim YW, Kang WK, Langley RE, et al. Individual
patient data meta-analysis of the value of microsatellite instability as a biomarker
in gastric cancer. J Clin Oncol 2019;37:3392–400.
[67] Chang WH, Lai AG. The pan-cancer mutational landscape of the PPAR pathway
reveals universal patterns of dysregulated metabolism and interactions with
tumor immunity and hypoxia. Ann NY Acad Sci 2019;1448:65–82.
[68] Chen L, Peng J, Wang Y, Jiang H, Wang W, Dai J, et al. Fenofibrate-induced mitochondrial
dysfunction and metabolic reprogramming reversal: the anti-tumor
effects in gastric carcinoma cells mediated by the PPAR pathway. Am J Transl Res
2020;12:428–46.
C. Ma et al. / EBioMedicine 67 (2021) 103347 11