Diversity Outbred Mice Reveal the Quantitative Trait Locus and Regulatory Cells of HER2 Immunity

Cancer immunotherapy such as active immunization and checkpoint blockade induces or strengthens endogenous tumor immunity to render lasting protection in some patients while providing minimal benefit in others (1). Such disparity in treatment outcome is due, in part, to different levels of neoantigens (2, 3). Although it has not received as much attention, the host genetic background can also dictate immune response to tumor-associated Ags. We reported that ERBB2 (HER2) DNA electroimmunization induced markedly more Ab and IFN-γ–producing T cells in BALB/c (BALB) HER2 transgenic (Tg) mice than those on C57BL/6 (B6) background, with (BALBxB6)F1 HER2 Tg mice producing an intermediate response (4). Similarly, in outbred domestic cats that received feline HER2 electroimmunization, only 30% generated detectable IFN-γ–producing T cells (5). Together, these results indicate genetic regulation in HER2-specific immune responses. Consistent with these observations, HER2 breast cancer patients produced variable vaccine responses in our and others’ clinical trials (6, 7). Our recent trial (NCT01730118; https://ascopubs.org/doi/abs/10.1200/JCO.2019.37.15_suppl.2639) immunized patients who had not received prior HER2-specific therapy with an adenovirus/HER2 (Ad/E2TM)–transduced dendritic cell (DC). Patients with HER2+ metastatic disease received 10–20 × 10⁶ DCs, and clinical benefit (complete response, partial response, or stable disease) was seen in approximately half of the patients (6). Final results including dose expansion cohort (40 × 10⁶ DCs) and in patients who received prior anti-HER2 therapy are under review. The specific nature of individual tumor can influence the vaccine response, but a genetic component of regulation is also likely.

Human genome-wide association studies have been performed extensively to identify genetic variants that influence human traits (8). Minor polymorphic alleles are expressed at very low frequency in humans, requiring a large sample size for the analysis. Although thousands of loci have been confirmed as disease risk factors, very few studies have investigated the mechanisms underlying particular associations (9).

A new Diversity Outbred (DO) mouse system arose after an extensive breeding effort (10, 11). DO mice were created by nonsibling crossing of eight inbred founder strains (A/J, C57BL/6J, 129S1/SvlmJ, NOD/HILtJ, NZO/HILtJ, CAST/EiJ, PWK/PhJ, and WSB/EiJ) to encompass >40 million single-nucleotide polymorphism (SNP), or >90% of all polymorphic alleles in mice. The genomes of the founder strains have been fully sequenced so that the genetic composition of each DO mouse can be fully profiled using the Giga Mouse Universal Genotyping Array (GigaMUGA) that displays 143,259 genetic markers, mostly SNPs, distributed evenly across the 20 chromosomes. This array allows the inference of small chromosomal segments, haplotype blocks, by their inbred strain of origin in each DO mouse by computational analysis. DO mice express polymorphic alleles at high frequencies because each allele comes from one of the eight founders. This model enables robust mapping analysis with a manageable number of mice (12, 13). The ability to control the environment and treatment conditions of experimental mice further improves the accuracy of genetic mapping.

In this study, syngeneic HER2 Tg mice were crossed with DO mice to generate (B6xDO)F1 or (BALBxDO)F1 HER2 Tg mice. These F1 mice carry one chromosome from the syngeneic parent to encompass the human HER2 gene in chromosome (Chr)5 and the DO genes in the other chromosome to provide individually distinct genetic backgrounds (14). Mice immunized with either Ad/E2TM or naked pE2TM developed a wide range of HER2-specific IgG and IFN-γ–producing T cells, consistent with genetic regulation of vaccine response. We analyzed the association between anti-HER2 IgG levels and genetic markers using DO quantitative trait locus (QTL) analysis. We found elevated logarithm of odds (LOD) scores, an adjusted measurement of genomic association significance, with total IgG and IgG2a/b/c (IgG2) on Chr2 and Chr17. We further identified NK cells, which interact with MHC-IB gene products on Chr17, as positive regulators of anti-HER2 Ab response.

DO mice have been used previously to identify genetic loci regulating disease conditions, such as chemical toxicity (15), atherosclerosis (16), and pain (17). In a DO F1 Tg adenocarcinoma of the mouse prostate (TRAMP) model, QTLs regulating prostate cancer development were identified, and RWDD4 and CENPU were implicated as increasing the aggressiveness of cancer cells (18). We show that the DO F1 model is effective for identifying key regulators of immunotherapy response as well as for investigating the mechanism of action and new intervention strategies.

All animal procedures were conducted in accordance with the U.S. Public Health Service Policy on Use of Laboratory Animals and with approval by Wayne State University Institutional Animal Care and Use Committee. C57BL/6 (B6) and BALB female mice were purchased from Charles River Laboratory. DO female mice (J:DO, stock number 009376) were purchased from The Jackson Laboratory. Heterozygous C57BL/6 HER-2 Tg mice (B6 HER-2 Tg) expressing wild-type human HER-2 under the whey acidic protein promoter were generated in our laboratory (19) and available from The Jackson Laboratory [B6.Cg-Pds5bTg(Wap-ERBB2)229Wzw]. BALB/c HER-2 Tg (BALB HER-2 Tg) mice were generated by back-crossing B6 HER-2 Tg mice with wild-type BALB/c mice.

Human breast cancer cell line SKBR3 was obtained from American Type Culture Collection (Manassas, VA) and cultured in the recommended medium. Authentication of cell line by short tandem repeat profiling was carried out with Promega’s Cell ID Systems as described by the supplier before cell stocks were frozen. Authenticated frozen stocks were thawed and maintained for <10 passages for experiments.

The Ad/E2TM was constructed from pE2TM (4) by inserting E2TM sequence in the E1a/E1b region of the adenovirus 5 vector that also has the E1a/E1b and E3 regions deleted and the Ad35 Knob and Fiber substituted for the corresponding Ad5 regions. The product is an Ad5f35 vector expressing the ECD and TM domains of human HER2, as described by Maeng et al. (H.M. Maeng, L.V. Wood, B. Moore, M.H. Bagheri, S. Webb, L. England, G. Martinez, S.M. Steinberg, S. Pack, D. Stroncek, J.C. Morris, M. Terabe, and J.A. Berzofsky, manuscript in preparation). The vector modification changes the receptor tropism of the adenovirus from CAR to CD46 and allows more efficient transduction of human DCs and hematopoietic cells. The adenoviral construct was developed by Dr. Malcolm Brenner at the Baylor Center for Cellular and Gene Therapy under contract with the Berzofsky laboratory, Vaccine Branch, National Cancer Institute. Ad/E2TM was propagated in HEK293 cells, Adenopure column purified (Puresyn), and quantified by QuickTiter immunoassay (Cell Biolabs). Mice were twice immunized 2 wk apart by i.m. injection with 1 × 10⁸ PFUs of Ad/E2TM in 50 μl of PBS.

pE2TM and pEF-Bos/GM-CSF encoding murine GM-CSF were previously described (20). Mice were anesthetized and 50 μg pE2TM in an admix with 20 μg pGM-CSF in 50 μl PBS was injected i.m. in the quadriceps muscle. Square wave electroporation using NEPA21 super electroporator (Nepa Gene) was applied with pulses at 100 V with 50 ms duration delivered eight times in two opposite orientations (4).

To deplete regulatory T cell (Treg), mice were injected once i.p. with 500 μg mAb PC61 directed at CD25 10 d prior to the first Ad/E2TM immunization. To deplete NK cells, (NODxDO)F1 HER2 Tg mice received 1 mg mAb anti-NK1.1 i.p., and (A/JxDO)F1 HER2 Tg mice received both 1 mg of anti-NK1.1 and 50 μl of rabbit anti-asialo GM1 serum on days 0, 2, and 4 before mice were immunized on day 7.

Sera and splenocytes (SC) were collected 2 wk following the last immunization. Anti-HER2 IgG was measured by binding to HER2+ SKOV3 cells using flow cytometry and Ab concentrations calculated by regression analysis using mAb TA-1 as the standard (21). Normal mouse serum or isotype-matched mAb was the control. Results were analyzed by Student t test.

HER2-reactive T cells were enumerated by IFN-γ ELISpot assay (BD Biosciences) (22). PBL or SC were incubated with recombinant HER2 or Neu protein (ecd–Fc fusion; Sino Biological). Results were expressed as spot-forming units per 1 × 10⁶ SC and analyzed using Student t test.

Ab levels were standardized by Ζ = (χ−μ)/δ, where χ is the raw data, μ is the mean, and δ is the SD. The Z scores represent the number of SD above or below the mean that a specific IgG level falls.

The genotype of each test mouse is determined with the GigaMUGA, built on the Illumina Infinium platform (23). The assay is performed centrally at Neogen (https://genomics.neogen.com/en/mouse-universal-genotyping-array).

DO QTL or the later R/qtl2 software package calculates genotype probabilities from GigaMUGA with a hidden Markov model that generates a probabilistic estimate of the diplotype state at each marker locus in each DO animal and evaluates the association between genotype and phenotype (https://rqtl.org/qtl2cran) (11–13, 24). Genotyping data for the DO founders are available at ftp://ftp.jax.org/MUGA. DO mice carry 36 possible diplotypes, but F1 mice have only eight possible haplotypes. With 7 df in the genome scan model, the F1 model has a greater power for detecting QTL peaks than the DO model with 35 df but does not detect recessive alleles.

Support intervals for QTL localization were determined using a 95% Bayesian credible interval (25). The LOD curve is transformed by raising it to the power of 10 and a region covering 95% of the area under the transformed curve defines the support interval. The area under the curve is numerically approximated using trapezoids between the marker loci. Allele association was inferred by the strain influence on the QTL. Alleles were examined using the SNP viewer (www.sanger.ac.uk/sanger/Mouse_SnpViewer) and filtered on the consensus regions.

To verify that HER2 immune response is influenced by genetic background, we analyzed human HER2-expressing B6 HER2 or BALB HER2 Tg mice after they received Ad/E2TM, which is currently being tested in trial NCT01730118. HER2 Tg mice express and are immune tolerant to wild-type human HER2 but do not develop tumors (19). Mice were immunized i.m. with Ad/E2TM twice 2 wk apart. HER2 IgG levels were measured 2 wk after the second immunization (Fig. 1A, 1B) per our established protocol (21). BALB HER2 Tg mice produced 42.2 ± 10.1 μg/ml HER2 IgG (n = 14) (Fig. 1B) with a coefficient of variation (CV) of 23.9%, whereas B6 HER2 Tg mice produced ∼10-fold lower IgG (4.3 ± 1.1 μg/ml; n = 10; CV = 24.9%) (Fig. 1A). These findings are consistent with our previous report that BALB HER2Tg mice are much more responsive than B6 HER2 Tg mice to HER2 DNA electroimmunization (4).

The difference between BALB and B6 suggests a genetic control of the Ab response to self-antigen. To determine the extent and mechanism of Ab variability, we crossed syngeneic HER2 Tg males with DO females to generate (B6xDO)F1 or (BALBxDO)F1 HER2 Tg mice, allowing the pups to display HER2 in individually unique genetic backgrounds. HER2 IgG response induced by 2× Ad/E2TM immunization in (B6xDO)F1 HER2 Tg (n = 84) and (BALBxDO)F1 HER2 Tg (n = 29) varied much more than their syngeneic inbred counterparts (Fig. 1A, 1B). (B6xDO)F1 HER2 Tg mice produced 0–128 μg/ml IgG (22.7 ± 28.6 μg/ml; CV = 125.6%) (Fig. 1A) and (BALBxDO)F1 HER2 Tg mice produced 3–172 μg/ml IgG, with a higher mean value (51.3 ± 49.6 μg/ml; CV = 96.7%) (Fig. 1B), which is consistent with a higher HER2 vaccine response in mice of BALB/c origin. In nontransgenic littermates in which human HER2 was a foreign Ag, Ad/E2TM immunization resulted in consistently higher IgG response: (B6xDO)F1 produced 93.7 ± 37.1 μg/ml (n = 58; CV = 39.5%) and (BALBxDO)F1 produced 67.5 ± 43.8 μg/ml (n = 18; CV = 64.9%) HER2 IgG (Fig. 1C). Taken together, these data indicate that F1 mice developed strong and more consistent HER2 IgG response when HER2 was a foreign Ag, but the response was highly heterogeneous when HER2 is a self-antigen.

We also analyzed HER2-specific IgG subclasses. In (B6xDO)F1 HER2 Tg mice, the IgG2/IgG1 ratio was 1.8 ± 1.0, whereas in (BALBxDO)F1 HER2 Tg mice, the ratio was 0.76 ± 0.29 (p < 0.0001) (Fig. 1D). IgG2 is indicative of Th1 dominance, commonly detected in mice from the B6 background, whereas IgG1 indicates Th2 dominance, which is more prominent in mice from the BALB/c background (26). These differential IgG subclass trends are maintained in the F1 pups even though the variability is greater. Nontransgenic littermates also showed the same Th1 or Th2 polarization when immunized with Ad/E2TM (Fig. 1E).

We further examined the Th1/CD8 response directly by measuring T cell IFN-γ production. T cell response was also variable in (B6xDO)F1 HER2 Tg mice ranging from 0 to 348 spot-forming units per 1 × 10⁶ SC (45.3 ± 84.5 spot-forming units per 1 × 10⁶ SC) (n = 54; CV = 186.5%) (Fig. 2A). Half of the mice did not produce measurable IFN-γ at all. Response in the (BALBxDO)F1 HER2 Tg mice was similar, with just over half of the mice generating highly variable T cell responses (Fig. 2B). These findings further indicate the genetic regulation of Th1/CD8 HER2 immunization response. Nontransgenic (B6xDO)F1 and (BALBxDO)F1 littermates produced more consistent T cell response to foreign HER2 (Fig. 2C). These results point to the conclusion that genetic polymorphism regulates immune response to self-antigen, but less so to foreign Ags.

The percentage of CV for IgG response was 24.9 for syngeneic B6 HER2 Tg mice and 125.6 for all (B6xDO)F1 HER2 Tg mice (Fig. 1A). The percentage of CV for siblings in each of 14 litters of (B6xDO)F1 HER2 Tg mice ranged from 38 to 124, with an average of 80.1. Therefore, DO F1 siblings whose genetic heterozygosity falls between syngeneic mice and nonsibling DO F1 mice showed intermediate levels of variation, indicating a correlation between immune response variation and genetic heterozygosity.

In one batch of (B6xDO) F1 HER2 Tg mice, Pearson correlation coefficient between the levels of IgG and IFN-γ–producing T cells was calculated to show strong to moderate correlation (data not shown), consistent with genetic regulation of both humoral and cellular immunity. It will be necessary to further analyze Th2 cells that produce IL-4 or IL-5 to fully elucidate the relationship between Ab and T cell responses.

For QTL analysis, the immunization of and tissue collection from 84 (B6xDO)F1 HER2 Tg mice were performed in three batches. Tail tissue DNA samples were submitted for haplotype analysis using GigaMUGA, which encompasses 143,259 chromosomal markers (23). Genotype probabilities/reconstructions and kinship matrices were generated using the DO QTL package (https://github.com/rqtl/qtl2) (12). Kinship is expressed in color with a value of 1 (white), representing complete identity or relatedness to self, and 0 (red) denoting no genetic relatedness at all (Fig. 3A). The shades between red and white indicate degrees of relatedness. Of 84 (B6xDO) F1 mice, 69 mice had kinship values <0.4, or <40%, genetic identity with any other mouse in the cohort. These were chosen for further analysis to reduce bias from closely related mice. HER2-binding Ab in immune sera were measured by flow cytometry. Ab levels from these 69 mice were subjected to DO QTL analysis. To minimize batch effects, the Ab levels in each batch were converted to a standardized Z score: Ζ = (χ − μ)/δ, where χ is the IgG level of individual mice, μ is the mean, and δ is the SD. The Z scores of the 69 samples ranged from +3.2 to −1 (Fig. 3B). Similar Z score ranges were observed for IgG1 and IgG2(a/b/c) (IgG2) (Fig. 3C, 3D).

The association between HER2-binding Abs and genetic markers was calculated, and LOD values are presented in the genome scan Manhattan plot (Fig. 3E). The most prominent QTLs were found in Chr2, Chr15, and Chr17 (red and green arrows), suggesting that genes in these regions regulate HER2 IgG response. Genome scans of IgG1 and IgG2 levels are shown in Fig. 3F, 3G, respectively. There are several peaks in each genome scan, consistent with many regulatory mechanisms for Ab production. Because the LOD scores are relatively modest, we have chosen to focus on the regions that are identified in multiple scans and shared among the total IgG, IgG1, and IgG2 responses. Comparing LOD profiles of total IgG (Fig. 3E) and IgG2 (Fig. 3G), the dominant IgG isotype in (B6xDO) HER2 F1 mice, common QTLs emerged in Chr2 and Chr17 (red arrows). Meanwhile, the QTL in Chr15 (green arrows) was shared between the total IgG and IgG1 datasets (Fig. 3F). We focused on the QTL mapping to Chr2 and Chr17 because IgG2 was the most prominent isotype.

Total IgG and IgG2 responses were further analyzed to evaluate founder strain QTL effects at the Chr2 locus (Fig. 4A, 4B, respectively). The eight founder strains are color coded. The founder strain haplotype at each marker is analyzed against Ab response (Fig. 4A, 4B, upper panels). Positive values indicate higher-than-average response. The corresponding genome scan is displayed just below the founder strain effect plot (Fig. 4A, 4B, lower panels). Three regions were noted. For total IgG, Chr2:137,958,080 (LOD = 5.6) and Chr2:153,912,130 (LOD = 5.5) were prominent QTLs, although the LOD scores were modest. In the IgG2 dataset, the LOD scores for Chr2:137,207,420 (LOD = 7.4) and Chr2:153,912,130 (LOD = 7.8) were higher, generating more confidence. Importantly, despite the relatively low LOD scores, these two regions largely overlap in total IgG and IgG2. Comparing the 95% confidence interval of the QTLs in total IgG and IgG2, Chr2:136,565,356–137,954,640 and Chr2:152,765,413–153,917,380 represent the shared regions. Another prominent QTL appeared in Chr2:64,912,590 (LOD = 9.33) for IgG2 and in Chr2:64,912,850 (LOD = 5.1) for total IgG, with overlapping 95% confidence interval (Chr2:64,670,932–65,972,260).

Meanwhile, cancer immunogens can be delivered with viral vectors by DNA electroporation or in other formulations. Each vaccine formulation could induce shared or unique regulatory mechanisms. To identify QTL that regulate immunity to HER2 rather than the vector, we also analyzed HER2 IgG response induced by naked DNA electroimmunization.

(B6xDO)F1 HER2 Tg mice were electroimmunized with naked DNA encoding E2TM, the same recombinant HER2 protein expressed by Ad/E2TM. Sixty-eight mice with kinship <0.4 were analyzed. Immune serum was collected 2 wk after the second immunization, and HER2 Ab was measured. As expected, total IgG (Fig. 5A) and IgG2 levels (Fig. 5B) were highly variable. A genome scan Manhattan plot for pE2TM-induced total IgG (Fig. 5C) identified an LOD peak in Chr17 adjacent to that in Ad/E2TM total IgG plot (Fig. 3F, Supplemental Table I). IgG2, again the most abundant IgG, was also measured and an LOD peak identified (Figs. 3G, 5D) within 1.5 megabase (Supplemental Table I). Based on the 95% confidence interval for LOD peaks, Chr17:34,962,626–36,638,254 was the most stringent denominator shared by all four datasets (Supplemental Table I). The founder strain effects for Ad/E2TM and pE2TM at this QTL (Fig. 5E–H) show A/J (yellow) or NOD (dark blue) haplotypes associating with elevated total IgG and IgG2. B6 (gray) haplotype at this QTL is associated with poor total IgG and IgG2 consistent with low HER2 IgG found in B6 HER2 Tg mice (Fig. 1A).

We tested if anti-HER2 IgG in (B6xDO)F1 HER2 Tg mice is functional by incubating HER2+ breast cancer cell line SKBR3 cells with sera collected before and after immunization (Supplemental Fig. 1). Immune serum was diluted to 3, 1.5, and 0.75 μg/ml based on HER2-binding IgG level measured by flow cytometry. We observed a dosage-dependent inhibition of SKBR3 cell survival by the immune serum, showing antitumor activity of immunization induced anti-HER2 IgG.

We evaluated the polymorphic genes in the shared QTL regions. We coupled that information with the strongest strain inference (Figs. 4, 5) to identify unique polymorphisms in those regions. The genes in the region of Chr2:136,565,356–137,954,640 (region 2) were identified. In this region Ab response was driven by NZO and CAST, and these two strains share several unique missense polymorphisms in Mkks and Slx4ip genes (Supplemental Table II). In region 3, Chr2:152,765,413–153,917,380, the response was driven by PWK/CAST. PWK has unique missense SNPs in Mylk2, Ttll9, and Pofut1. CAST has missense variants in Tpx2, Mylk2, Foxs1, Ttll9, Ccm2l, Tspyl3, Commd7, Bpifb2, and Bpifb6. The genes in these regions are potential regulators of Ab response to self-antigen and their missense SNPs are shown in Supplemental Table II. For the first QTL region in Chr2 (Chr2:64,670,932–65,972,260), the Ab response appeared driven by 129. Strain 129 did not have any unique missense polymorphisms in this region but has one unique structural alteration at 65,284,693–695 because of insertion.

The region of Chr17 (Chr17:34,962,626–36,638,254) is a large region that encompasses many genes (83 polymorphic genes). The major driver strain of elevated total IgG and IgG2 in all four datasets is A/J. Strain NOD contributes in three of four datasets. There are 23 genes with A/J unique substitutions in this region, including the genes Atat1, Cdsn, Ddx39b, Gm8909, Gm9573, Gtf2h4, H2-Bl, H2-D1, H2-M10.1, H2-Q1, H2-Q2, H2-Q5, H2-Q7, H2-Q10, H2-T22, H2-T23, Ly-6g6e, Ly-6g6f, Msh5, Prrc2a, Sapcd1, Vars2, and Vwa7. These are shown in Supplemental Table III.

These 23 genes encompass several important biological and immunological mechanisms. A similar gene cluster is found in human Chr6, suggesting a preservation of immune regulatory genes through evolution. Of note are H2-Q7, H2-Q10, and H2-T23, all encoding proteins that interact with inhibitory receptors of NK cells (Supplemental Table I). Both H2-Q7 and H2-Q10 proteins are ligands of inhibitory receptor Ly-49C (27). H2-T23 (Qa1b), a homolog of human HLA-E, binds another inhibitory receptor CD94/NKG2A (28, 29). Interestingly, A/J and NOD haplotypes share 17 missense mutations that are distinct from all other founder strains within Chr17:36,031,051–36,032,420 of H2-T23 gene (Supplemental Table III). We proceeded to test the hypothesis that NK cells regulate HER2 vaccine response in HER2 Tg mice. We crossed B6 HER2 Tg mice with A/J and NOD mice and tested the effect of depleting NK cells in their HER2 Ab response. Test mice were immunized with Ad/E2TM twice. Before immunization, half of the mice received mAb NK1.1 or NK1.1 plus anti-asialo GM1 to deplete NK cells. HER2 binding total IgG, IgG1, and IgG2 were measured after each immunization. HER2 IgG production was reduced by 5–10-fold in NK-depleted (B6xA/J)F1 (Fig. 6A) and (B6xNOD)F1 (Fig. 6B) HER2 Tg mice, showing NK cells as positive regulators of HER2 Ab response. Similar reductions were observed in IgG1 and IgG2 (p < 0.05). Furthermore, the reduction in IgG1 is near complete in (B6xNOD)F1 HER2 Tg mice. Given that NK cells produce copious amount of IFN-γ, one might expect greater impact on IgG2 production. The reduction of IgG1 may suggest mechanisms beyond IFN-γ. It should be noted that anti-NK1.1 would also deplete a portion of NKT cells, which can also produce IFN-γ, but because their restriction element CD1d is on mouse Chr3, not Chr17, the genetics do not point to this lineage as playing a major role.

Immune reactivity is often regulated by both positive and negative mechanisms. Because NK cells positively regulate HER2 immune response, negative regulation is to be expected. We previously reported enhanced HER2 vaccine response by Treg depletion in syngeneic HER2 Tg mice (4, 30). To test if Treg also control HER2 immune response in outbred mice, (B6xDO)F1 HER2 Tg mice were treated with anti-CD25 mAb 10 d prior to the first Ad/E2TM immunization per our standard protocol (4). HER2 IgG response was elevated and became more uniform (Fig. 7, gray circles). Treg depletion of nontransgenic (B6xDO)F1 littermates had little impact on their response to HER2 as a foreign Ag (Fig. 7, white circles). Therefore, Treg negatively regulate vaccine humoral response to self-HER2 regardless of the genetic background. With these results, we conclude that NK cells and Treg are opposing regulators in HER2 immunization response with strong implications on future planning of immunotherapy.

Our HER2 DNA constructs have advanced to clinical trials including the ongoing phase I autologous AdHER2 DC trial (https://ascopubs.org/doi/abs/10.1200/JCO.2019.37.15_suppl.2639) (6, 7). Preclinical studies indicated direct tumor inhibitory effects of immune sera (6). Clinical benefit has been observed in selected patients suffering from HER2-expressing solid tumors. To expand the benefit of HER2 immunization in heterogeneous human patients, the regulatory mechanisms of HER2 Ab response was sought. We tested HER2 vaccines in (B6xDO)F1 and (BALBxDO)F1 HER2 Tg mice to recapitulate immunization response in outbred populations and to identify candidate regulatory genes. Ad/E2TM and pE2TM immunization were compared side-by-side. Using the DO QTL package in the R environment, we associated genetic markers on GigaMUGA with standardized Ab levels. Chr17:34,962,626–36,638,254, home to the MHC-IB gene family, emerged as a common denominator. H2-Q7, H2-Q10, and H2-T23 in this region interact with inhibitory receptors on NK cells (27, 29). NK cell depletion before immunization reduced HER2 IgG response, demonstrating NK cells as a positive regulator. We further described in DO F1 HER2 Tg mice a negative regulation of HER2 IgG response by Treg. QTL analysis did not pull out Treg-associated genetic loci, likely because Treg suppress tumor immunity regardless of genetic background. Thus, by combining the new DO QTL analysis with conventional tumor immunology experience, we discovered that NK and Treg regulate HER2 Ab response from opposite directions, setting a platform for modulating HER2 immune response by balancing NK and Treg or their functional receptors. We do note that each candidate gene identified through DO QTL still needs to be individually validated to establish its role and mechanism of action. Murine QTL, or the parallel human genome-wide association studies, reveals genetic association, not causation, of the phenotype. Deciphering the functional QTL and causative genes requires some consideration. F1 mice have eight possible haplotypes, or 7 df. If DO mice with 36 possible diplotypes were used, many more mice would be required to reach the necessary power of analysis. Furthermore, we had the opportunity to cross-reference genome scan results from two different immunogens encoding the same Ag, as well as cross-referencing two independent Ig measurements: total IgG and IgG2 in the same serum sample. This allowed us the opportunity to identify a QTL in Chr17 without an excessive number of mice or exceptional LOD value. Interestingly, MHC is the single region most often associated with autoimmunity (31, 32), reinforcing the validity of the Chr17 QTL in self-HER2 immunity. It was somewhat surprising to map the QTL to a region rich in nonconventional MHC-IB and to have the subsequent confirmation that NK cells regulated by MHC-IB contribute directly to HER2 vaccine response. NK cells are recognized for their prompt secretion of IFN-γ when encountering immune stimulation. We reported that NK cells were important in Ad/E2-induced, Ab-dependent, cell-mediated cytotoxicity against mammary tumor growth (33). However, this activity was at the effector, not priming, phase. NK cell or IFN have been associated with both positive or negative impact on viral infection or autoimmunity (20, 34–36). Treatment with NKG2A inhibitory mAb in cancer therapy may shed light on whether blocking this checkpoint receptor will be beneficial (37).

NKT cells, which recognize lipid Ags in the context of CD1d, can also be depleted by anti-NK1.1 mAb (38, 39). Inhibitory NK cell receptors CD94/NKG2A and Ly-49c, which interact with H2-T23 and H2-Q7/Q10, respectively, can also regulate NKT cell activation (40, 41). It is possible that depletion of NK cells with mAb NK1.1 also depleted a portion of NKT cells, and these are also known for producing IFN-γ and could thus promote IgG2 Ab responses as well as activate DCs. However, CD1d, the restriction element for NKT cells, is on mouse Chr3, not Chr17, with other class IB genes, so the genetic mapping does not point to NKT cells. More in-depth analysis could elucidate if NKT cells contribute to HER2 immunity even though HER2 is not a lipid Ag. Still, modulation of inhibitory receptors shared by NK and NKT cells may enhance immune response if NKT cells also contribute to the regulation of HER2 immunity.

Ag delivery formulations have been developed empirically, and they evoke different regulatory mechanisms. The same HER2 ECD and TM domains were expressed by both Ad/E2TM and pE2TM, but adenoviral vector introduces foreign viral components that may tilt HER2 response. The uptake and processing of viral particles involves more cellular mechanisms than direct delivery of naked DNA by electroporation. QTL analysis indicates common and unique regulatory mechanisms in the Ad/E2TM and pE2TM vaccine delivery system, providing additional insights to immunotherapy considerations.

Additional QTL shared by selected datasets merit attention (e.g., Chr2:136,565,356–137,954,640 and Chr2:152,765,413–153,917,380 appeared in both total IgG and IgG2 following Ad/E2TM immunization). Genes encompassed in these regions may lead to improved understanding of Ad/E2TM immune response.

One QTL in Chr15 was found in Ad/E2TM-induced total IgG and IgG1, but not IgG2, datasets. The LOD peak for total IgG (chr15: 26,932,660; LOD = 6.6) and IgG1 (chr15:27,255,750; LOD = 7.0) overlap in their 95% confidence intervals. March11 and FBXL7, both involved in protein ubiquitination, are found in this region. Further verification will be required to determine if ubiquitination contributes to Ad/E2TM-induced IgG1 production.

In addition to assessing genetic regulation of vaccine response, HER2 DO mice are a powerful platform for testing novel cancer immunotherapies. Reduction in total IgG, IgG1, and IgG2 was observed in Ad/E2TM-immunized, NK-depleted DO F1 mice. On the contrary, a striking elevation and more consistent production of HER2 IgG was observed when mice received anti-CD25 mAb prior to Ad/E2TM immunization. This is consistent with our report that Treg suppress HER2 immunity and autoimmunity in HER2 Tg mice of BALB, B6, and HLA-DR3 background (42, 43). Interestingly, a direct antagonism of Treg and NK cells in mammary tumor lung metastases has been previously reported (44). It will be of interest and importance to determine if NK and Treg interact directly in HER2 immunization. The current finding supports the likely benefit of modulating Treg and NK cells during immunotherapy in genetically diverse human patients.

Because anti-CD25 mAb can deplete regulatory as well as activated effector T cells, mAb PC61 was given to mice 10 d before the first immunization and not after any immunization. If Treg depletion should be required after the initiation of immunization, the study could be conducted in mice of DEREG background (45), in which Treg can be removed specifically by administration of diphtheria toxin.

Previous work demonstrated that genetic background can influence Ab production in an MHC-independent manner, with A/J mice producing more Abs but with lower affinity compared with B10 mice (46). Interestingly, we also observed that A/J background led to greater Ab production compared with B6. Similarly, although HER2 IgG levels are highly variable in DO F1 HER2 Tg mice, the average Ab level is higher in mice with a BALB/c parent, consistent with greater HER2 Ab response in BALB HER2 than B6 HER2 Tg mice (4). This is also true when the ratios of IgG2/IgG1 were measured to reflect the impact of Th1 and Th2. Introduction of DO genetic composition did not fully override Th1 versus Th2 preference.

In nontransgenic mice, the response to human HER2 as a foreign Ag is relatively uniform. Regardless of the parental strains or genetic polymorphisms, high levels of HER2 IgG and T cell responses were induced, showing effective immunity against foreign Ag even though their T and B cell and Treg composition in the PBL demonstrated a high degree of diversity (data not shown), suggesting the adaptation of genetically diverse individuals to survive environmental assaults.

Studies using the eight founder strains of DO mice showed strain-driven differences in enteric microbial structure, indicating genetic influence of intestinal microbiome (47). When nearly 400 DO mice were subjected to DO QTL analysis, slc10a2 was identified as a possible regulator of both a gut microbe and a bile acid (48). In other studies, vaccine response to infectious pathogens was associated with HLA polymorphism (47, 49). One could ask whether the same genetic elements could regulate both microbiome and vaccine response and could be identified through DO QTL. Although our DO QTL results did not indicate a direct association between microbiota and vaccine response, the DO and DO F1 platforms would be most suitable for addressing this and many other unanswered biological questions.

One important question in any DO mouse study is whether QTL identified in mice can be translated to humans because not every mouse gene has a human homolog, and the regulation mechanisms of mouse and human phenotypes are not always the same. We show that identification of a QTL rich in MHC-IB gene combined with conventional tumor immunology knowledge led to the discovery that NK cells positively regulate HER2 humoral immunity. With extensive conventional effort, blockade of NKG2A checkpoint is being tested in clinical trials in combination with cetuximab (37). Our findings argue that DO mice are a strong representation of outbred humans and a powerful tool for discovering and validating clinically important genetic variables. The richness of information from DO F1 QTL analysis will withstand repeated query and provide new leads for scientific discoveries and disease control.

This article is prepared in proud memory of Richard F. Jones, who dedicated his professional life to cancer research. We thank Daniel M. Gatti and Pen-Yuan Hsing for invaluable guidance in DO QTL analysis. We thank Claire McCarthy and Soumith Inturi for assistance in performing the analysis. We thank Malcolm Brenner and other members of the Center for Cell and Gene Therapy, Baylor College of Medicine, Texas Children’s Hospital, and Houston Methodist Hospital, Houston, Texas for help in Ad/E2TM construction. We also recognize Karmanos Cancer Institute Microscopy, Imaging and Cytometry Shared Resources for flow cytometry support.

The authors have no financial conflicts of interest.