Shedding Light on the African Enigma: In Vitro Testing of Homo sapiens-Helicobacter pylori Coevolution

The continuous characterization of genome-wide diversity in population and case–cohort samples, allied to the development of new algorithms, are shedding light on host ancestry impact and selection events on various infectious diseases. Especially interesting are the long-standing associations between humans and certain bacteria, such as the case of Helicobacter pylori, which could have been strong drivers of adaptation leading to coevolution. Some evidence on admixed gastric cancer cohorts have been suggested as supporting Homo-Helicobacter coevolution, but reliable experimental data that control both the bacterium and the host ancestries are lacking. Here, we conducted the first in vitro coinfection assays with dual human- and bacterium-matched and -mismatched ancestries, in African and European backgrounds, to evaluate the genome wide gene expression host response to H. pylori. Our results showed that: (1) the host response to H. pylori infection was greatly shaped by the human ancestry, with variability on innate immune system and metabolism; (2) African human ancestry showed signs of coevolution with H. pylori while European ancestry appeared to be maladapted; and (3) mismatched ancestry did not seem to be an important differentiator of gene expression at the initial stages of infection as assayed here.


Introduction
Coevolution is a biological term coined in 1964 by Ehrlich and Raven [1] to describe relationships between two entities where selective pressures are exerted on each other's evolution. Coevolution occurs in many forms of mutualism, host-parasite, and predatorprey relationships, as well as competition within or between species. The most extreme examples of exquisite adaptation, displaying evidence of tightly coevolved morphology, physiology, and behavior, are the symbiotic integrations of mitochondria and chloroplasts in eukaryotic cells [2]. Improvements in mathematical models [3] have been showing that the expected stable adaptive peaks attainable through coevolution are rarely maintained, as the selective landscapes are under continual change through reciprocal selection on the two communities, whereas the H. pylori virulence factor CagA did not. Thus, these findings were consistent with the idea that neither human nor H. pylori genetic variation can confer susceptibility or virulence per se, making it necessary to consider these together [21].
Despite the likely etiological importance of human-pathogen coevolution, attempts at laboratory confirmation have been rare. In the context of H. pylori infection, only one in vitro study has been conducted using strains of European and African ancestry to infect the Caucasian-derived AGS human cell line [22]. The authors showed that European strains promoted significantly higher host cell IL8 expression than African strains, whereas African strains promoted host cell apoptosis. Although these findings support the influence of H. pylori ancestry in promoting gastric disease, they did not precisely inform about the molecular mechanisms regulating coevolution.
In this work, we propose to shed light on the coevolution theory by investigating which human genes and pathways are involved in specific interactions between host and pathogen, under dual matched and mismatched ancestry conditions. Given the theoretical expectation of stronger adaptation for the African host and pathogen organisms, we focused on comparing the dual African and European ancestries. For that, and after selecting human gastric cell lines and H. pylori strains of African and European ancestries, we conducted matched and mismatched in vitro coinfection assays, evaluated complete human gene expression profiles and functionally assessed the cytotoxicity, viability, apoptosis, oxidative stress and lactate production in those settings.

Ancestry Inference of the Gastric Cell Lines
Ancestry was inferred for 41 gastric cell lines whole exome sequenced as part of the Cancer Cell Line Encyclopedia (CCLE; [23]). The whole genome sequence information for the worldwide populations from the 1000 Genomes Project [24] was used as reference for ancestry inference. GATK HaplotypeCaller (version 3.7) was used for variant calling of these two sets and after merging common variants (amounting to 173,128 single nucleotide polymorphisms; SNPs), these were pruned for pairwise linkage disequilibrium (LD) in PLINK [25], by removing any SNP that had an r 2 > 0.2 with another SNP, within a 50-SNPs sliding window with a step of 10 SNPs (final count of 58,537 SNPs). ADMIXTURE [26] was used to infer genetic structure of the pruned merged dataset in K = 3 ancestry components, representing the main population groups from Africa, Europe and Asia.

Infection of Gastric Cells
Gastric cancer cell lines were grown in antibiotic-free medium at approximately 80% confluence in 75 cm 2 tissue culture flasks (VWR, Radnor, PA, USA). Medium changes were carried out every other day and immediately before infection. For infection experiments, bacteria grown for 48 h were collected in phosphate buffer saline (PBS; pH 7.4) and added to gastric cell monolayers, at a multiplicity of infection (MOI) of 100 bacteria per cell. Coinfection of the selected gastric cell lines with the different H. pylori strains is schematized in Supplementary Materials, Figure S1. Co-cultures were maintained for 24 h at 37 • C, under a 5% CO 2 humidified atmosphere. Uninfected control cell cultures were processed similarly, with the addition of PBS without bacteria. Three biological replicates were performed per infection and per gastric cancer cell line.

Human RNA Processing and AmpliSeq Expression Profiling
mRNA was extracted from the samples with Trizol (Life Technologies, Carlsbad, CA, USA) according to the manufacturer's protocol and quantified by spectrophotometry on an Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA). Quality of samples was checked through Qubit 3.0 fluorometer and Qubit RNA HS Assay kit (ThermoFisher Scientific, Waltham, MA, USA) in the Agilent 2100 Bioanalyzer (Agilent, Santa Clara, CA, USA). Reverse transcription of RNA was done using the SuperScript ® VILO™ cDNA Synthesis Kit (ThermoFisher Scientific, Waltham, MA, USA). Target transcriptome sequencing was performed with the Ion AmpliSeqTM Transcriptome Human Gene Expression Kit (ThermoFisher Scientific, Waltham, MA, USA), which contains a 150 bp amplicon for each of 20,802 human genes, and quality control was checked in an Agilent 2200 TapeStation (Agilent, Santa Clara, CA, USA). The template preparation was done in an Ion Chef TM System with the Ion 550™ Chip Kit, and sequencing was performed in an Ion S5TM XL System (ThermoFisher Scientific, Waltham, MA, USA). FASTQ files were generated and aligned using the Torrent Suite™ Software against the Ion AmpliSeq TM Transcriptome target region (GRCh37 human reference) to infer the human expression profiles.

Cell Viability, Cytotoxicity and Apoptosis Assays
Gastric cancer cells were seeded in 96 well plates (BD Biosciences, San Jose, CA, USA) for two days to guarantee a density of 20,000 cells per well. The coinfection of these cells was carried out following the specifications previously described, maintaining the 24 h as period of infection. ApoTox-Glo TM triplex assay (Promega, Madison, WI, USA) was then used to measure apoptosis, cell viability and cytotoxicity. An amount of 20 µL of viability/cytotoxicity reagent containing both GF-AFC and bis-AAF-R110 substrates was added to all wells, and briefly mixed. GF-AFC, glycyl-phenyl-alanylaminofluorocoumarin, enters in intact cells where it is cleaved by the live-cell protease activity to generate a fluorescent signal proportional to the number of living cells. Bis-AAF-R110, bis-alanylalanyl-phenylalanyl-rhodamine 110 is not cell-permeant, hence, no signal from this substrate is generated by intact, viable cells, being instead transformed by dead-cell proteases which were released from cells that have lost membrane integrity. After incubation for 1 h at 37 • C, the live-and dead-cell proteases produce AFC and R110 products, respectively, which have different excitation and emission spectra, allowing their simultaneous detection in a Synergy microplate reader (BioTek, Winooski, VT, USA): cell viability was measured by fluorescence at 400 Ex /505 Em , while cytotoxicity was estimated at 485 Ex /520 Em . For analysis of apoptosis, 100 µL of Caspase-Glo ® 3/7 reagent was then added to all wells and incubated for 1 h at room temperature. Addition of this reagent results in cell lysis, followed by caspase cleavage of the substrate and generation of a luminescent signal produced by luciferase that was measured in the same reader. Three biological replicates were performed per infection and per gastric cancer cell line. As positive controls of viability/cytotoxicity and apoptosis, respectively, 100 µM ionomycin (Sigma-Aldrich, St. Louis, MO, USA) which is toxic to the cells and 10 µM of staurosporine (Sigma-Aldrich) that causes apoptosis, were incubated for 6 h before addition of the assay compounds.

L-Lactate Concentration and Reactive Oxygen Species (ROS) Production Assays
Gastric cancer cell lines were grown in antibiotic-free medium at approximately 500,000 cells in 6 wells plates for two days. Infection was carried out for 24 h following the specification previously described. After this period, culture medium was removed, filtered by a 0.22 µm cellulose filter (Frilabo, Maia, Portugal), and stored at −80 • C after deproteinization with perchloric acid and neutralization by KOH, for further measurement of L-Lactate concentrations, using the Lactate Assay Kit (MAK064; Sigma-Aldrich, St. Louis, MO, USA). Briefly, 40 µL of the Master Mix were added to each well with 40 µL of deproteinized culture medium (either uninfected or infected) and incubated for 30 min at room temperature, protected from light. This enzymatic assay resulted in a colorimetric product that was measured in a Synergy microplate reader at 570 nm (BioTek, Winooski, VT, USA), whose intensity is proportional to the L-Lactate concentration in the sample. A calibration curve was constructed with 0, 2, 4, 6, 8 and 10 nmol/well of the standard solution. Concentration values were normalized for the amount of cells per sample. Three biological replicates were used per infection and per gastric cancer cell line.
New culture medium was added to the adherent cells (post-infection) which were used to estimate mitochondrial superoxide (the most abundant ROS) production, through MitoSOX TM Red (Invitrogen, Carlsbad, CA, USA). MitoSOX™ Red reagent is a fluorogenic dye specifically targeted to mitochondria in live cells, and its oxidation by superoxide produces red fluorescence. Briefly, 2.5 µM MitoSOX were added to each well, and incubated for 30 min at 37 • C, protected from light. Culture medium was then removed, and adherent cells were treated with 0.25% trypsin-EDTA (Invitrogen 25,200, Carlsbad, CA, USA) and collected separately into a microcentrifuge tube. Cells were then washed once with PBS and centrifuged at 300× g. MitoSOX fluorescence was measured by flow cytometry at 488 nm (excitation) and 575 nm (emission) using the BD Accuri C6 Plus (BD Biosciences, San Jose, CA, USA). Results were analysed by FlowJo v10.0.7 (Tree Star, Inc., Ashland, OR, USA) using the mean fluorescence intensity values for the FL2 channel. Three biological replicates were used per infection and per gastric cancer cell line. Cell controls without MitoSOX probe and with MitoSOX with 50 µM antimycin (Sigma-Aldrich, St. Louis, Missouri, USA; that causes oxidative stress) treatment were used as negative and positive controls, respectively.

Algorithms and Statistics Applied to Gene Expression and Functional Data
All the statistical analysis and graphical representations presented in this work were performed in R version 3.6 [27]. Quality control checks of expression profiles of triplicates were investigated through Principal Component Analysis (PCA) and the rooted neighborjoining tree, based on a matrix of Euclidean genetic distance, was obtained from the pruned whole exome variability from the 41 gastric cancer cell lines.
Differentially expressed genes between paired tests were identified by DESeq2 package [28], which applies a negative binomial distribution test (adjusted p-value threshold below 0.05 was considered). Clustering and heatmap representations of these significantly expressed genes, between uninfected and infected tests, were obtained using heatmap.3 package.
Pre-ranked pairwise gene-set enrichment analyses (GSEA) were conducted in GSEA-InContext [29] for the GO-biological process ontology. Results were ordered by the normalized enrichment score (NES) and the false discovery rate (FDR). Specific significantly enriched pathways were further explored based on information contained in the publicly available database of Ingenuity (https://targetexplorer.ingenuity.com/index.htm; Qiagen, Hilden, Germany). The list of intervening genes and chemical reactions was collected and used in drawing schematic representations containing information of significant fold-changes in expression levels when comparing the infected versus non-infected experimental settings.
In order to get a focused insight on the innate immune response to the infection, the InnateDB [30] was used, amounting to 951 protein-coding genes having a role in the innate immune response.
Welch's t-tests were applied in R to compare the mean values of cell viability, cytotoxicity, apoptosis, L-Lactate concentration and ROS production, and values below 0.05 were considered significant. The Welch's t-test adjusts the number of degrees of freedom when the variances are unequal.

Ancestry of the Gastric Cell Lines and H. pylori strains
Ancestry information was not available for most of the stomach cancer cell lines at the time we began this work. We selected the 41 cancer cell lines exome sequenced as part of the Cancer Cell Line Encyclopedia (CCLE, [23]) and merged these samples with the populations from the 1000 genomes project [24]. Admixture analysis ( Figure 1A, Supplementary Table S1) revealed that the great majority of the cell lines (35 out of 41) derive from an East Asian (EAS) background. Only two cell lines (Hs746T and 23132-87) were isolated from individuals of European ancestry. One cell line harbored 75% African and 25% European backgrounds (NCI-N87), probably derived from an admixed African-American individual (according to the mean 27% and 22% European input in the African-American cohorts of northern and southern USA, respectively [31]). The remaining three cell lines had varying degrees of admixture between the three ancestries. Based on these results, we selected the only option for HsAFR (NCI-N87), and between the two available options for the HsEUR, we selected the one with fewer mutations (Hs746T with 608 mutations according to CCLE website, instead of 23132-87 with 3150 mutations; for comparison, NCI-N87 has 477, and MKN74 has 738 mutations). Additionally, we selected the widely used HsEAS (MKN74) for infection assays with European (close mismatch) and African (distant mismatch) H. pylori.  Meanwhile, Dutil et al. [32] published their ancestry inference in the entire CCLE panel, based on another technology, the Affymetrix SNP6.0 chip. Their results, also using ADMIXTURE but ascertaining information for seven ancestry components (African, Native American, North and Southeast Asian, South Asian, and North and South European) inferred the following proportions of the cell lines used here: HsEUR (Hs746T)-96.36% European, 3.48% Asian (East and South), and 0.16% Native American; HsAFR (NCI-N87)-61.23% African; 36.60% European and 2.17% Asian; and HsEAS (MKN74)-98.71% Asian and 1.29 European. These values are largely concordant with our inference for HsEUR and HsEAS. However, the differences are higher (by 14%) for the proportion of African ancestry inferred for the HsAFR. We downloaded their data and confirmed their proportions when running an ADMIXTURE for K = 3, which leads us to suggest the high European-prone ascertainment biases introduced by the chip used in Dutil et al. [32] as the best cause for this discrepancy when compared to the non-ancestry biased whole exome sequencing. Despite these incongruities, the HsAFR cell line has a dominant African ancestry.
Regarding the H. pylori strain selection, we followed published phylogeographic data [33] and selected H. pylori J99 of West African ancestry and H. pylori 26695 of European ancestry, henceforth abbreviated as HpAFR and HpEUR, respectively. These strains are both CagA-positive and have the toxic form VacA.

Global Expression Profile Alterations
The PCA for the overall expression profile ( Figure 1B) shows, as expected, that the cell line background is the main clustering factor. The three cell lines are almost equidistant in terms of their expression profiles, despite the closeness between HsEUR and HsEAS in terms of exome diversity (Supplementary Figure S2). Another important information from the PCA is the clear distinction between the expression profiles of uninfected versus infected (either matched or mismatched) status for HsEAS and HsAFR, and less so for HsEUR.
Differentially expressed genes between uninfected and infected conditions showed statistically significant (adjusted p < 0.05) alterations in gene expression (Supplementary  Tables S2-S7): in HsEUR, 515 and 209 were up-regulated, and 364 and 31 were downregulated when infected by HpEUR and HpAFR, respectively; in HsAFR, those numbers were 2335 and 1554, and 2279 and 1745; and in HsEAS, those numbers were 2554 and 2214, and 2464 and 2159.
Twenty-three genes were up-regulated (log2-fold change > 1) in common to all infection settings, mainly related to I-kappaB kinase/NF-kappaB signaling (adjusted p = 0.0015; Figure 1C). The MYCBP gene, which controls the transcriptional activity of the protooncogene MYC and that plays a role in cell cycle progression, apoptosis and cellular transformation, was the only gene down-regulated in common to all infection settings ( Figure 1D).
In summary, this global expression pattern shows that H. pylori infection, irrespectively of matched or mismatched ancestries, activates immune-related I-kappaB kinase/NF-kappaB signaling, and down-regulates the cell cycle related pathways of the human gastric cells. Additional cell line specific pathways are activated upon infection, and their involvement in a differential human ancestry cellular response to infection was further explored.

Detailed Analysis of Altered Molecular Pathways
Molecular pathways are highly complex, with multiple redundant control systems. Therefore, the additive effect of small changes in expression of several genes can better describe the complex biological systems (and have higher statistical power) than an extreme fold change in expression of one or two genes. This is the rationale behind gene set enrichment analysis (GSEA), which we applied to the pairwise comparisons for each experimental setting. Figure 1E summarizes the top pathways (Supplementary Tables S8-S16 report all  results). For the HsEUR, the comparison between uninfected vs. infected (both H. pylori strain ancestries) revealed top enrichment of immune-related pathways (and at the first position, the response to type I interferon) and top down-regulation of cell cycle and DNA repair pathways. These enrichment hits remained the main significant result when comparing matched (HsEUR × HpEUR) versus mismatched (HsEUR × HpAFR) ancestry sets. In HsAFR, infection in general increased especially lipid metabolism, and down-regulated cell cycle, while the matched-mismatched conditions led to top enrichment of response to oxidative stress in the mismatched (HsAFR × HpEUR) and cilium morphology and protein transport in the matched setting (HsAFR × HpAFR), respectively. Concerning the HsEAS, the pairwise comparison between uninfected and infected status revealed a major enrichment of metabolic pathways related to the production of energy, and down-regulation of cell cycle pathways. Infection with HpEUR in HsEAS (close-mismatch) led to enrichment of immune-related pathways whereas infection with HpAFR (distant-mismatch) led to enrichment of pathways related to blood vessels and wound healing.
The GSEA confirmed the up-regulation of NF-kappaB related pathways in all infection sets, especially significant in HsEUR (FDR = 0.0027 for HpEUR; 0.0015 for HpAFR) than in HsEAS (0.071 for HpEUR; 0.65 for HpAFR) and HsAFR (0.053 for HpEUR; and not statistically significant, 0.16, for HpAFR). Statistically significant up-regulation of response to type I interferon was observed in HsEUR (0 for infections with HpEUR and HpAFR).

Exploring the Enriched Pathways-Innate Immune System
Considering that our experimental infection model only includes gastric epithelial cells we focused on innate immune response genes, by referring to the InnateDB database [30], which includes 951 curated protein-coding genes involved in this response. Figure Figure S3). Genes in block 4 are mainly up-regulated by the infection in HsAFR, and are associated with endocytosis (p = 2.5 × 10 −2 ) and response to cytokine (p = 2.1 × 10 −2 ). Genes in block 5 are mainly unaffected by the infection in HsAFR, and these genes are related to IL-17 signaling pathway (p = 2.0 × 10 −10 ), NF-kappa B signaling pathway (p = 4.4 × 10 −10 ), NOD-like receptor signaling pathway (p = 7.2 × 10 −10 ) and TNF signaling pathway (p = 1.1 × 10 −9 ). No major differences in expression levels of these innate immune genes were observed in the host cell lines of different ancestries upon infection with bacteria of matched and mismatched ancestries ( Figure 2B).
The chemokine IL8/CXCL8 (in block 5) deserves a more careful inspection, as some of its SNPs have been associated, although disputably, to risk of gastric cancer [34]. This gene is up-regulated upon infection in the European and East Asian cell lines, but it is consistently expressed at low levels in both the uninfected and infected settings of the African cell line. We compared the expression of this gene between European (Great Britain) and African (Yoruba) in the RNASeq data available in the 1000 Genomes consortium website and confirmed that the gene is significantly (p = 2.0 × 10 −7 ) less expressed in the latter than in the former population ( Figure 2C). When attending to the genomes of the associated SNP, rs4073 (formally designated as −251 T/A), the expression of the gene ( Figure 2D) decreases with the dose of allele A (adjusted p-values: AA vs. AT-0.0335; AA vs. TT-0.0023; AT vs. TT-0.1541). This SNP presents ( Figure 2E) a high heterogeneity between population groups. The low-expressing IL8 AA genotype is predominant in African populations (~70% frequency) and has a low frequency in European and East Asian populations (<20%). The genotypes for rs4073 of the cell lines used in this study are AA for both HsAFR and HsEAS, and TT for HsEUR.
Asian populations). Lower plot is a zoom of the CCLE panel. (B)-PCA plot (PC1 explaining 54% of variation vs. PC2 explaining 31% of variation) of the human transcriptomic profile of the three gastric cell lines (triplicates for each condition; blue colors for the European; green for the Asian and pink for the African) without and after infection with HpAFR and HpEUR H. pylori strains. (C)-Venn diagrams for up-regulated and (D)-down-regulated genes in all experimental H. pylori infected settings compared with the uninfected status (indicated by Ø ). (E)-Top significantly enriched pathways in pairwise comparisons between infected (positive side) and uninfected (negative side) conditions in the three cell lines. The different color represents H. pylori strains (pink for African and blue for European ancestries). The scale reports the normalized enrichment score (NES) values.

Exploring the Enriched Pathways-Lactate Metabolism
Few studies addressed the metabolic effects of H. pylori infection on gastric epithelial cells [35][36][37]. H. pylori use L-lactate released by gastric epithelial cells to obtain growth benefits [37]. The bacterial genes encoding the L-lactate dehydrogenase were recently identified [35]. Our data showed that infection is associated with enrichment in gene expression of LDHA (production of L-lactate in the host cell) and SLC16A4/MCT4 (export of L-lactate from the host cell) in HsEAS and HsAFR ( Figure 3A,B and Supplementary  Figures S4-S6), and with a significant decrease of L-lactate concentration in the extracellular culture medium of the infected settings ( Figure 3C). These results endorse the idea that H. pylori stimulate the secretion of L-lactate by gastric cells for its benefit. Regarding the HsEUR cell line, no significant differences were observed for the L-lactate concentrations between the uninfected and the infected settings. A possible explanation for this observation is the a priori high concentration of L-lactate in the HsEUR cell line compared with the other two (Supplementary Figure S7, based on metabolome data from [38]), allowing a higher amount of L-lactate released to the extracellular medium readily available to H. pylori. In terms of lactate metabolism stimulation by the bacteria, there were no major differences between matched and mismatched infections, except in HsAFR infected by HpEUR.  Further evaluation with available data on human and bacteria gene expression profiles (PRJNA378649; no associated publication yet) from co-infection experiments with the same cell line as our HsAFR and HpEUR bacterial strain allowed us to ascertain the significant increase in expression of two out of the three lactate-related bacterial genes (HP0137-1.6 log2 fold change; and HP0138-0.2 log2 fold change).

Exploring the Enriched Pathways-Mitochondrial Oxidative Stress
H. pylori infection in humans is associated with enhanced levels of reactive oxygen species (ROS), increased oxidative DNA damage, and diminished glutathione in the mucosa [39]. Based on this, we examined whether HpAFR and HpEUR elicited different ROS production in the different host cells and whether these results were linked to specific cellular responses associated with oxidative stress. ROS levels ( Figure 4A-D) were significantly increased by infection with HpEUR in all cell lines compared to the uninfected controls, with the HsEUR cell line being the most affected (250% increase). In contrast, infection with HpAFR had irregular effects in the different cell lines (no differences in HsEUR, increase in HsAFR and reduction in HsEAS).   Analysis of the differentially expressed genes associated with oxidative stress revealed that infection in all cell lines significantly activated genes related to antioxidants, most probably as a means to reduce the oxidative damage ( Figure 4E,F and Supplementary  Figures S8-S10). Interestingly, while in HsEUR, only that part of the pathway was showing the expression changes, in both HsAFR and HsEAS there were also alterations in chaperon and ubiquitination proteins related with repair and removal of damaged proteins (mostly up-regulation, in particular in HsAFR), and in the detoxifying and metabolizing enzymes associated with cell survival (specially down-regulation when HsAFR was infected by HpEUR). These results might indicate that HsAFR and HsEAS, have a more efficient response to ROS than HsEUR, which might explain the higher concentration of ROS displayed by this cell line upon infection with HpEUR.
Concerning the well-established relationship between oxidative stress and cellular damage, we investigated the impact of the different ancestries in cellular viability, cytotoxicity, and apoptosis in each cell line infected for 24 h (Figure 5; t-test results are listed in Supplementary Table S17). The infection decreased cell viability of the HsEUR, HsAFR and HsEAS cell lines, although in the latter the decrease was not statistically significant. Cytotoxicity was negatively associated with viability. Quantification of caspase 3/7 luminesce for the presence of apoptosis revealed that H. pylori did not induce any significant differences in the level of apoptosis.

Discussion
By applying a genome-wide approach to the host transcriptome, supported by some functional confirmations, in controlled double-ancestry host-pathogen in vitro settings, we were able to provide experimental evidence to address three main questions: (1) do the host responses to H. pylori infection, in terms of whole-genome expression profile, differ between human ancestries; (2) are coevolution model expectations fulfilled, with higher adaptation of the African ancestry to H. pylori infection, and maladaptation of European and Asian populations groups; and (3) do infections caused by a mismatched H. pylori ancestry differ from a matched setting?
Our results clearly show that a higher number of genes were significantly changed upon infection (up-and down-regulated) in the African and Asian cell lines than in the European (≈5 to 10 times), indicating a more complex response (possibly meaning better adaptation) to the infection stimulus. Within each cell line, numbers were of the same order of magnitude for both HpEUR and HpAFR infections, although slightly lower for the latter. More interesting, the top activated pathways differed between human ancestries: African up-regulated lipid metabolism, while Asian activated energy production, contrary to European which relied on immune-related pathways as the most up-regulated. All ancestries displayed down-regulation of pathways related to the cell cycle. These results seem to indicate that African and Asian ancestries had a broader, and probably better, adapted molecular response to infection by H. pylori, while European ancestry relied substantially on the immune system (and the signal is stronger when this cell line is infected by HpEUR than HpAFR). In agreement with the lipid metabolism activation in the African ancestry upon infection with H. pylori, we recently showed that lipid metabolism plays a major role in naturally protecting the African human ancestry from the worse phenotypes of dengue virus disease [40]. Differences in the cholesterol efflux regulatory protein (encoded by ABCA1 gene) were already observed in H. pylori infection, which depletes cholesterol in gastric glands to prevent interferon-gamma signaling and to escape the inflammatory response [41].
Mucosal epithelial cells are a defense barrier against invading pathogens such as bacteria. They express pattern recognition receptors that play important roles in the initiation, maintenance, and regulation of both innate and adaptive immune responses [42], although in the context of H. pylori infection the elicited immune response does not lead to bacterial clearance [43]. H. pylori infection of the gastric mucosa induces T helper type 1 (Th1) and Th17 adaptive immune responses, as well as innate immune responses that have been less studied. The main rule of thumb was human ancestries being associated with very diverse innate immune responses. In a host with European ancestry, H. pylori infection distinctly activated type I IFNs. This type of response was traditionally associated with a defense against viruses but has recently been shown to be involved in bacterial infections, having protective or detrimental roles depending on the bacterium [44]. In a mouse model with impairment of ISGF3 signaling, H. pylori led to decreased CXCL10 responses and increased susceptibility to infection [45]. This type I IFN up-regulation was also consistent with the up-regulation of several interferon-stimulated genes, in the HsEUR cell line, such as MX1, MX2, IFIT1, IFI6, and OAS3. In contrast, some genes important in IL-17, NF-kappa B, NOD-like receptor and TNF signaling pathways were not up-regulated in the host cell with African ancestry upon infection, which preferentially stimulated genes involved in phagocytosis and endocytosis. IL17 is an essential cytokine for host defense against bacteria, promoting pro-inflammatory cascades [46]. Cells triggered by microbes secrete IL17A, which is recognized by an IL17 receptor [47] activating NF-kB and MAPK/AP-1 inflammatory pathways. Activation of these pathways leads to the production of pro-inflammatory cytokines, chemokines and antimicrobial peptides, which induce inflammation required for host defense [48,49]. This evidence seems to point to a decreased pro-inflammatory response in the African human epithelial cells. We also confirmed that Africans had a baseline lower IL-8 expression than Europeans and that IL-8 did not play a role in the immune response to H. pylori in the human African ancestry, suggesting a lower bacterial-induced inflammation in this ancestry background. Our results are instrumental in contributing information that will help to clarify contradictory published evidence.
The response to oxidative stress induced by ROS seemed to be more effective in HsAFR (and in HsEAS) than in HsEUR, as several genes with parallel actions to antioxidants/reduction of oxidative damage were also activated in the former and not in the latter. The HpEUR strain appeared to be especially inductive of ROS accumulation, reaching a massive value when infecting with the matched ancestry HsEUR cell line. In the HsAFR and HsEAS cell lines, both strains were able to induce the production and export of lactate by the host, allowing an increased input of lactate into the bacteria. In contrast, the concentration of lactate was already high in the uninfected HsEUR, which might explain the lack of changes in this metabolism upon infection.
We can thus conclude, in response to the raised questions: (1) yes, the host response to H. pylori was greatly molded by human ancestry; (2) the African human ancestry showed clear signs of coevolution with H. pylori while the European human ancestry was maladapted, with the Asian ancestry in between (but closer to the coevolved African); and (3) the mismatched host-bacterium ancestry did not appear to be an important differentiator of gene expression, at least at the initial stages of infection, as we analyzed here. HpEUR × HsEUR induced similar gene expression profiles as HpAFR × HsEUR, and the same within the African cell line. This observation does not exclude the possibility of a worse phenotype being conferred by a mismatched rather than matched in vivo infection, in more advanced stages of the process. This possibility would render our results compatible with the observations collected in Colombian admixed cohorts [21]. Future validation of these results, when other gastric cancer cell lines of Africa and European ancestries become available, is recommended. In fact, this work exemplifies the current limitations in the basic tools available to conduct oncobiology research in the African ancestry, due to the scarcity of appropriate cancer cell lines, mostly of which are derived from admixed African-Americans (mean of 25% of European ancestry). Our results preliminarily point towards the following predictions in terms of disease risk: mismatched HpAFR × HsEUR will present a higher risk when compared with HpEUR × HsEUR, while both HpEUR × HsAFR and HpAFR × HsAFR situations will present a low risk because the African cell line seems to be better adapted to H. pylori in general.
Supplementary Materials: The following are available online at https://www.mdpi.com/2076-260 7/9/2/240/s1, Table S1: Admixture proportions (%; K ancestral groups = 3) for each gastric cell line from the CCLE in relation with the 1000 Genomes reference superpopulations.  Table S17-T-test results for the analysis of apoptosis, cellular viability and cytotoxicity. Figure S1: "Schematic representation of the co-infection experimental design. Infection was performed in three cancer cell lines, each representative of the three main human population groups (AFR-African, EUR-European and EAS-East Asian), by two H. pylori strains (European and African origin). Admixture percentages for each human ancestry per cell line are specified in the left-hand side. Each infection was done in triplicate. Figure S2: SNP based phylogenetic tree. Phylogenetic tree depicting the SNP distance between the 41 CCLE gastric cancer cell lines. Figure S3: Log2 changes in gene expression between HpAFR and HpEUR infection sets for the Type I interferon genes and Interferon stimulated genes. Figure S4: Metabolism of lactate in the HsEUR cell line. Statistically significant fold changes (up-regulation in pink and down-regulation in blue) in gene expression upon infection with A-HpEUR and B-HpAFR strains against the uninfected set in glycolysis, Krebs cycle, and fatty acids synthesis. Figure S5: Metabolism of lactate in the HsEAS cell line. Statistically significant fold changes (up-regulation in pink and down-regulation in blue) in gene expression upon infection with A-HpEUR and B-HpAFR strains against the uninfected set in glycolysis, Krebs cycle, and fatty acids synthesis. Figure S6: Metabolism of lactate in the HsAFR cell line. Statistically significant fold changes (up-regulation in pink and down-regulation in blue) in gene expression upon infection with A-HpEUR and B-HpAFR strains against the uninfected set in glycolysis, Krebs cycle, and fatty acids synthesis. Figure S7: Lactate cellular concentration (normalized mass-spectrometry values; see original paper for description) amongst 37 CCLE stomach cell lines, based on metabolomic data from [38]. HsEUR in blue, HsEAS in green and HsAFR in red. Figure S8: Illustration of the oxidative stress response in the HsEUR cell line. Statistically significant fold changes (up-regulation in pink and down-regulation in blue) in gene expression upon infection with A-HpEUR and B-HpAFR strains against the uninfected set. Figure S9: Illustration of the oxidative stress response in the HsEAS cell line. Statistically significant fold changes (up-regulation in pink and down-regulation in blue) in gene expression upon infection with A-HpEUR and B-HpAFR strains against the uninfected set. Figure S10: Illustration of the oxidative stress response in the HsAFR cell line. Statistically significant fold changes (up-regulation in pink and down-regulation in blue) in gene expression upon infection with A-HpEUR and B-HpAFR strains against the uninfected set.