Table 1. Evidence for G×G in human host–pathogen systems.
| Pathogen | Human gene | Pathogen gene/protein | Type of study | G x G test (host–pathogen) | Trait | Effect of host polymorphism for each pathogen genotype | GFG/MA1 | Signature of balancing selection? | Cost of resistance? |
|---|---|---|---|---|---|---|---|---|---|
| Plasmodium falciparum | HLA-B | csp | Naturally infected (cases) | Candidate gene x candidate gene | Clinical malaria | HLA-B*35 associated with higher frequency of 2 P. falciparum genotypes and lower frequency of 2 genotypes [53]. | ? | Yes [5,54]. | Yes; the haplotype involved in G x G (HLA-B x 35) is associated with a number of diseases, with OR both >1 and <1 [40]. |
| Plasmodium falciparum | HBB | msp-1 | Naturally infected (cases) | Candidate gene x candidate gene | P. falciparum infection | Frequencies of msp-1 genotypes differ between HBB AS and AA individuals [55]. | ? | Yes; recent positive/balancing selection [47]. | Yes; HbS leads to anemia and associated diseases [56]. |
| Plasmodium falciparum | HBB | 3 genes on 2 chromosomes; all in strong LD | Case-control | Candidate genes x genome wide | Severe malaria | HbS protective against severe malaria if infected with parasite having major alleles at all 3 genes (OR≈0.02) but not when infected with parasite having minor alleles (OR≈1) [23]. | GFG | Yes; recent positive/balancing selection [47]. | Yes; HbS leads to anemia and associated diseases [56]. |
| Streptococcus pneumoniae | SASH1, FARP1 | pspC | Case-control | Genome x candidate genes | Pneumococcal meningitis | SASH1: OR for 1 pspC allele >1; OR for 2 other pspC alleles not ≠ 1. FARP1: OR for 1 pspC allele >1; OR for 2 other pspC alleles not ≠ 1 [20]. | GFG | No | No known. |
| Streptococcus pneumoniae | STK32C | Case-control | Genome x pathogen strain (”serotype”) | Pneumococcal meningitis | OR for 1 serotype > 1; OR for 6 other serotypes not ≠ 1 [20]. | GFG | No | No known. | |
| Mycobacterium tuberculosis | HLA-DRB1, HLA-DQB1 | Case-control | Candidate gene x pathogen strain | Tuberculosis | DRB1*09:01 OR and DQB1*03:03 OR for “modern strains” >1; OR for other strains not ≠ 1 [26] | GFG | Yes [5,54] | Yes [40]. | |
| Mycobacterium tuberculosis | HLA-A, HLA-B, HLA-C | Naturally infected (cases) | Candidate gene x pathogen strain | Pulmonary tuberculosis | Specific HLA class I alleles associated with specific M. mycobacterium strains [57]. | ? | Yes [5,54] | Yes [40]. | |
| Mycobacterium tuberculosis | TLR2 | Case-control | Candidate gene x pathogen strain (“lineage”) | Tuberculous meningitis | OR for Beijing lineage >1; OR for other lineages not ≠ 1 [24]. | GFG | No | No known. | |
| Mycobacterium tuberculosis | CD209 | Naturally infected (cases) | Candidate gene x pathogen strain | Mortality from pulmonary tuberculosis | Infection with Beijing vs. other strains in patients who died from TB associated with SNP in CD209 [58]. | ? | No | No known. | |
| Mycobacterium tuberculosis | SLC11A1 | Naturally infected (cases) | Candidate gene x pathogen strain | Pulmonary tuberculosis | Infection with Beijing vs. other strains associated with 2 SNPs in SLC11A1 [59]. | ? | (Yes) [60]. | Yes; see below. | |
| Mycobacterium tuberculosis | SLC11A1 | Naturally infected (cases) | Candidate genes x pathogen strain (“lineage”) | Tuberculosis severity | Opposite effects of SLC11A1 alleles in individuals with lineage “L4.6 Uganda” vs. other lineages [35]. | MA | (Yes) [60]. | Yes; high expressing allele associated with autoimmune diseases while low expressing allele associated with infectious diseases [44]. | |
| Mycobacterium tuberculosis | PPIAP22 | Naturally infected (cases) | Genome x pathogen strain (“lineage”) | Tuberculosis severity | Opposite effects of PPIAP22 alleles on disease severity in individuals with lineage “L4.6 Uganda” vs. other lineages [21]. | MA | No | No known. | |
| Helicobacter pylori | ABO | babA | In vitro functional analysis | Candidate gene x pathogen strain (“isolate”) | Binding of host receptor | Most strains bind both A and H antigen (generalists), but significant fraction of strains in S America are specialists and bind only H (i.e., blood group O) [29]. | Mainly GFG | Yes [46]. | Yes; blood group A, B, and AB increase susceptibility to severe malaria [27,61]. |
| Vibrio cholerae | ABO | ctxAB | Case-control | Candidate gene x pathogen strain (“serogroup”/“biotype”) | Disease severity | Blood group O confers higher risk of severe disease than blood group A or B, but only when infected with O1 El Tor or O139 strains; in contrast, no association between disease severity and ABO group for classical O1 strains [27]. | GFG | Yes [46]. | Yes; blood group A, B, and AB increase susceptibility to severe malaria [27,61]. |
| HIV | HLA-A, HLA-B, HLA-C | 48 amino acid residues throughout HIV proteome | Naturally infected (cases) | Genome x genome | Immune escape mutations | 48 HIV-1 amino acid variants associated with SNPs in HLA-A, B, or C [17]. See also [30]. | ? | Yes [5,54]. | Yes [40]. |
| HIV | HLA-A, HLA-B, HLA-C | Protease/RT, Nef, Vpr | Naturally infected (cases) | Candidate genes x candidate genes | Immune escape mutations | Different HLA-A, HLA-B, and HLA-C alleles often select for different amino acid escape mutations at a given HIV residue (at 23/57 residues where polymorphism is associated with specific HLA alleles) [31]. | MA for at least 23/57 HIV residues involved in GxG. | Yes [5,54]. | Yes [40]. |
| HIV | HLA-A, HLA-B, HLA-C | 31 amino acid residues throughout HIV proteome | Naturally infected (cases) | Candidate genes x genome | Immune escape mutations and viral load | Specific HLA-A, HLA-B, and HLA-C alleles select for specific amino acid escape mutations at 31 HIV residues (but no case where different HLA alleles select for different amino acids at a given residue, as in [31]). In addition, effect of HLA-A/B allele x immune escape mutation on viral load at 3 of the 31 residues, such that viral load is reduced in individuals carrying specific HLA alleles if HIV has not acquired escape mutation [32]. | Immune escape:?; viral load: GFG. | Yes [5,54]. | |
| HIV | KIR2DL2 | Vpu, Env | In vitro infection assay | Candidate gene x candidate gene | Viral replication | NK cells with ≥1 copy of KIR2DL2 inhibit replication of HIV Vpu-Env(WT/WT) but not Vpu-Env (V/V), but NK cells w/o KIR2DL2 inhibit neither HIV genotype [34]. | GFG | Yes [62]. | Yes; haplotype w KIR2DL2 associated w several AID, including T1D and UC [43]. |
| Hepatitis C virus | HLA-A, HLA-B, HLA-C, DQA1, DRB1 | NS3, NS4B | Naturally infected (cases) | Genome x genome | Immune escape mutations | Specific HLA alleles select for specific amino acid escape mutations at certain residues, but no case where different HLA alleles select for different amino acids at a given residue [18]. See also [33]. | ? | Yes [5,54]. | Yes [40]. |
| Hepatitis C virus | IFNL4 | NS5B | Naturally infected (cases) | Genome x genome | Immune escape mutations and viral load | Immune escape: IFNL4 rs12979860 associated with amino acid variants at 11 HCV residues. Viral load: IFNL4 rs12979860 C>T associated with reduced viral load, but only if HCV has serine at site 2414 [18]. See also [63]. | Immune escape:?; Viral load: GFG | No (but population-specific positive selection; [64]). | No known. |
| Hepatitis C virus | KIR genes | Naturally infected (cases) | Candidate gene x pathogen strain (HCV genotype) | Risk of hepatocellular carcinoma (HCC) | For HCV genotype 1, 2, and 3, risk of HCC decreases with number of activating KIR genes; for HCV genotype 4, risk is low regardless of number of activating KIR genes [36]. | GFG | Yes [62]. | Yes, several activating KIR genes associated with autoimmune diseases [43]. | |
| Norovirus | FUT2 | Challenge studies/case-control/prospective cohort studies | Candidate gene x pathogen strain | Acute gastroenteritis | Challenge studies: Nonfunctional FUT2 protects against strain GI.1 and GII.4, but not against GII.2. Case-control studies: Nonfunctional FUT2 protects against GII.4, but not against GI.3. Prospective cohort studies: Nonfunctional FUT2 protects against GII.3 and GII.4, but not against GI.3, GI.6, GII.1, GII.2, and GII.7 [25]. |
GFG | Yes [65,66]. | Yes; nonfunctional FUT2 associated w susceptibility to Crohn’s disease [41] and other diseases [42]. | |
| Epstein-Barr virus (EBV) | UNC5D, LINC01830, non-coding region on chr 7 | BALF5, BBRF1, BRLF1 | Naturally infected (cases) | Genome x genome | Immune escape mutations | 25 host SNPs in 3 genomic regions (17, 1, and 7 SNPs, respectively) associated with variants in 3 EBV genes [19]. | ? | Yes, UNC5D [67]. | Possibly for UNC5D; SNP in this gene associated with adolescent idiopathic scoliosis. |
| Epstein-Barr virus (EBV) | HLA-B | EBNA-1 | In vitro functional analysis | Candidate gene x pathogen strain | In vitro immune response | EBV peptide variant HPVG most immunogenic on HLA-B*35:01, but peptide variant HPVG-D5 most immunogenic on HLA-B*35:08 [37]. | MA | Yes [5,54]. | Yes [40]. |
| Human papillomavirus (HPV) | HLA-DR/DQ | Case-control | Candidate gene x pathogen strain | Cervical cancer | Different HLA-DR-DQ haplotypes affect resistance/susceptibility to cancer caused by different HPV types [28]. | MA | Yes [5,54]. | Yes [40]. | |
| Human papillomavirus (HPV) | HLA-DR/DQ | E6 | Naturally infected (cases; i.e., cervical cancer patients) | Candidate gene x candidate gene | Infection (or immune escape mutation?) | HLA-DR04-DQ03 haplotype associated with HPV E6 L83V variant in cervical cancer patients [68,69]. | ? | Yes [5,54]. | Yes [40]. |
1 GFG = gene-for-gene type G×G, MA = matching allele type G×G,? = type of G×G could not be inferred from published data. For case-control studies of binary traits (presence/absence of infection or disease), GFG was inferred when the OR for ≥1 pathogen genotype was different from 1 while the OR for ≥1 other pathogen genotype was equal to 1, whereas MA was inferred when the OR for different pathogen genotypes where in opposite directions. For analyses of continuous traits (pathogen load, disease severity), GFG was inferred when there was an effect of host genotype on the trait for ≥1 pathogen genotype but no effect of host genotype for other pathogen genotypes, whereas MA was inferred when host genotype had opposite effects on the trait for different pathogen genotypes. For escape mutations, MA was inferred when alternative alleles at a HLA gene were associated with different escape mutations at a given pathogen residue (see main text for more detailed explanations).