Lung cancer, the leading cause of cancer-related deaths worldwide, is a heterogeneous disease comprising multiple histologic subtypes that harbor disparate mutational profiles. Immune-based therapies have shown initial promise in the treatment of lung cancer patients but are limited by low overall response rates. We sought to determine whether the host immune response to lung cancer is dictated, at least in part, by histologic and genetic differences, because such correlations would have important clinical ramifications. Using mouse models of lung cancer, we show that small cell lung cancer (SCLC) and lung adenocarcinoma (ADCA) exhibit unique immune cell composition of the tumor microenvironment. The total leukocyte content was markedly reduced in SCLC compared with lung ADCA, which was validated in human lung cancer specimens. We further identified key differences in immune cell content using three models of lung ADCA driven by mutations in Kras, p53, and Egfr. Although Egfr-mutant cancers displayed robust myeloid cell recruitment, they failed to mount a CD8+ immune response. In contrast, Kras-mutant tumors displayed significant expansion of multiple immune cell types, including CD8+ cells, regulatory T cells, IL-17A–producing lymphocytes, and myeloid cells. A human tissue microarray annotated for KRAS and EGFR mutations validated the finding of reduced CD8+ content in human lung ADCA. Taken together, these findings establish a strong foundational knowledge of the immune cell contexture of lung ADCA and SCLC and suggest that molecular and histological traits shape the host immune response to cancer.
Despite decades of research, small cell lung cancer (SCLC) and non–small cell lung cancer (NSCLC) remain among the world’s deadliest diseases (1). SCLC, in which RB1 and TP53 mutations are common (2), accounts for 10–20% of lung cancer diagnoses (3). More than half of NSCLC cases are classified as lung adenocarcinomas (ADCAs), in which KRAS, EGFR, and TP53 mutations are the predominant genetic drivers (4, 5). Although patients with EGFR mutations initially respond to targeted therapies, drug resistance typically develops within the first year (6). SCLC and KRAS-mutant ADCA have proven less tractable because scant progress has been made toward the development of targeted therapeutics for these patients (7). Novel immunotherapeutic strategies have offered new hope for the management of NSCLC (8–10) and SCLC (11), but the clinical success of immunomodulatory agents depends on a strong foundational knowledge of the cells that make up the lung tumor microenvironment (TME) in these molecularly and histologically distinct diseases.
Inflammation is a key attribute of neoplasia (12). The host immune response to cancer is an intricate web of pro- and antitumorigenic signals (13). CD8+ T cells, also known as CTLs, are the body’s main immunological barrier against cancer, because they are capable of recognizing and killing tumor cells. However, CTL activity is curbed by tumor cells expressing immune checkpoint ligands (e.g., PD-L1), as well as an influx of immune-suppressive cells (i.e., regulatory T cells [Tregs], macrophages, monocytes, and neutrophils) into the TME (14, 15). Newly developed immune checkpoint inhibitors (ICIs; e.g., ipilimumab and nivolumab) seek to reverse this suppression and unleash an antitumor response (16).
Although some lung cancer patients have experienced remarkable tumor regression upon commencing ICI therapy, overall response rates have peaked around 20% (9, 10). These disparate outcomes may be explained, in part, by findings that tumor immune cell composition and function vary among anatomical sites and histological origins (17). In lung cancer, for example, squamous cell carcinoma patients exhibited longer progression-free survival with ipilimumab treatment than did ADCA patients (8). However, even within histological subgroups, response rates vary widely, raising the question of what other tumor attributes might predict clinical outcome.
The oncogenic functions of mutant RAS and EGFR in cancer include the production of proinflammatory cytokines, such as IL-8, which help to shape the TME (18–21). TP53 similarly demonstrated non–cell autonomous behaviors during tumorigenesis (22, 23). Nevertheless, the discrete impact of molecular signatures, such as KRAS, EGFR, TP53, and RB1 mutations, on the immune cell composition of lung cancer remains largely undefined. To address this question, we profiled the TME of three genetically engineered mouse (GEM) models of NSCLC (KrasLSL-G12D, KrasLSL-G12D;Trp53Fl/Fl, and EgfrL858R), as well as the Rb1Fl/Fl;Trp53Fl/Fl model of SCLC. In this article, we show that the molecular and histological subtypes of lung cancer predict immune cell composition and may, therefore, demand specific immunotherapeutic regimens.
Materials and Methods
All animal experiments used aged-matched mice and were conducted at the Fred Hutchinson Cancer Research Center using protocols approved by the Institutional Animal Care and Use Committee. TetO-EgfrL858R mice (24) were obtained from the Mouse Models of Human Cancer Consortium on a C57BL/6 background. Ccsp-rtTA mice (25) on an FVB background were provided by Jeff Whitsett (University of Cincinnati). KrasLSL-G12D (Kras) (26), Trp53Fl/Fl (p53) (27), RORγtGFP (28), and Tcrd−/− (29) mice were obtained from the Jackson Laboratory on a C57BL/6 background. TetO-EgfrL858R;Ccsp-rtTA (Egfr), KrasLSL-G12D;Trp53Fl/Fl (Kp53), KrasLSL-G12D;RORγtGFP (K.RORγt), and KrasLSL-G12D;Tcrd−/− (K.Tgd) mice were generated by simple cross-breeding. Rb1Fl/Fl mice (30) were crossed with p53 mice to generate Rb1Fl/Fl;Trp53Fl/Fl (Rbp53) mice on a mixed C57BL/6 × 129 background.
Egfr and single-transgene control mice (Ccsp-rtTA or TetO-EgfrL858R) were fed food impregnated with 200 mg/kg doxycycline (Harlan, Indianapolis, IN). Kras, Kp53, p53, and wild-type (wt) C57BL/6 animals received an intratracheal dose of 2.5 × 107 PFU of adenoviral Cre recombinase (AdCre; University of Iowa Viral Vector Core, Iowa City, IA), as described (31). Each cohort was studied for 6, 10, and 14 wk postinitiation of doxycycline or infection with AdCre. Additional cohorts of K.RORγt and K.Tgd animals were similarly subjected to AdCre infection (2.5 × 107 PFU) and examined 14 wk postinitiation or when moribund. Rbp53 mice received 1 × 108 PFU of AdCre; given the long latency period, these animals were studied 9 mo postinduction.
CTLA4 Ab, clone 9D9 (MedImmune) or isotype control (mIgG2b) was administered to an additional cohort of Kras mice twice weekly via i.p. injection for a total of 4 wk, starting at 8 wk post-AdCre, at a dose of 10 mg/kg.
Tissue collection and histology
Lung tissue specimens were collected and processed as described (32). Briefly, the left lung was ligated and snap-frozen for later analysis. The right lung was inflated with 10% neutral buffered formalin at 25 cm H2O pressure before fixing in neutral buffered formalin overnight. Five-micrometer paraffin-embedded sections were stained with H&E or immunostained for CD45 (BD Bioscience, San Diego, CA), Foxp3 (eBioscience, San Diego, CA), or CD3 (Serotec, Raleigh, NC) using 3, 3′-diaminobenzidine development and hematoxylin counterstaining. Global adjustments to white balance, brightness, and/or contrast were made to some photomicrographs using Photoshop (Adobe Systems, San Jose, CA).
Slides were examined with an Eclipse 80i microscope (Nikon Instruments, Melville, NY), excluding the whole-lobe images presented in Fig. 1A, which were acquired with an Aperio digital pathology slide scanner (Leica Biosystems, Buffalo Grove, IL). Total lung and tumor areas (mm2) were measured from H&E-stained slides using NIS-Elements Advanced Research software (Nikon Instruments). Results are expressed as the percentage of lung occupied by tumor [(area tumor ÷ area lung) × 100]. Each lung was also scored for tumor grade, as described (33). Foxp3- and CD3-stained lung lobes (n = 5 mice per genotype) were scored for the presence or absence of tumor-associated (TA) cells, tumor-infiltrating (TI) cells, and cells within a lymphoid aggregate (LA).
Lung tissue single-cell preparation
Single-cell suspensions were generated from saline-perfused mouse lungs using mechanical disruption, followed by a 1-h digestion at 37°C in RPMI 1640 containing 10% FCS and penicillin/streptomycin, 80 U/ml DNase, 300 U/ml collagenase Type 1 (both from Worthington Biochemical, Lakewood, NJ), and 60 U/ml hyaluronidase (Sigma, St Louis, MO). Digested lungs were sheared through a 19-gauge needle, strained through 70-μm nylon mesh, centrifuged, lysed (RBCs), washed, strained through 40-μm mesh, centrifuged, and resuspended in Dulbecco’s PBS + 2% FCS. Cell viability was determined using trypan blue staining and a TC20 Automated Cell Counter (Bio-Rad, Hercules, CA). We obtained two SCLC and five ADCA surgical specimens, each with nonadjacent normal lung tissue, using the approved Institutional Review Board file number 6663 in association with Fred Hutchinson Cancer Research Center, University of Washington Medical Center, and Northwest BioTrust. Single-cell suspensions were generated using the above digestion protocol.
Single-cell suspensions were incubated with 1.0 μg of Mouse TruStain FcX or 1.0 μl of Human TruStain FcX (both from BioLegend, San Diego, CA) per 106 cells prior to immunostaining. Twenty-seven fluorochrome-labeled Abs were distributed among four multicolor panels for mouse specimens (all flow Abs are detailed in Supplemental Table I). Immunostaining was performed for 30 min on ice, protected from light. Dead cells were excluded with Fixable Viability Dye (FVD) eFluor 780 (eBioscience), per the manufacturer’s protocol. Stained cells were washed, fixed with IC Fixation Buffer (eBioscience), and stored at 4°C until analysis.
Intracellular cytokine production was assessed using PMA (25 ng/ml), ionomycin (1 μg/ml; both from Sigma), and monensin (1.5 μl/ml; BD Bioscience) stimulation for 5 h at 37°C, 5% CO2. An unstimulated sample was incubated in the absence of PMA and ionomycin. After stimulation, cells were washed, stained with FVD, fixed, and permeabilized with a Transcription Factor Buffer Set (BD) prior to immunostaining.
Samples were analyzed on an LSR II flow cytometer with FACSDiva software (BD), and ≥1 × 105 events were recorded per sample. Data were compensated and analyzed with FlowJo software (TreeStar, Ashland, OR). Gates were defined by fluorescence-minus-one samples and verified with appropriate isotype controls. The unstimulated control was used to define cytokine gates. Total cell content was calculated by multiplying the overall number of live cells recovered from each animal (i.e., the trypan blue–negative hemocytometer count) by the percentage of live cells for each gated parameter. Cytokine-producing T cell subsets were calculated by multiplying the percent parent gate with the previously determined parent population count. The median fluorescence intensity (Med.F.I.) of NK group 2, member D receptor (NKG2D)-BV421 and PD-L1–BV421 parameters was calculated in FlowJo, and Med.F.I. of the relevant fluorescence-minus-one control was subtracted from all experimental values for normalization.
Splenocytes from wt mice were labeled with 50 μM CFSE (Molecular Probes, Eugene, OR), per the manufacturer’s instructions. A total of 1 × 106 CFSE-labeled splenocytes was transferred to six-well tissue culture plates coated with anti-CD3/anti-CD28 Abs (BioLegend). Cells were incubated with 200 μg of homogenate, generated from 10-wk Egfr, Kras, or wt lungs, at 37°C, 5% CO2 for 4 d. Lymphocyte proliferation was determined by harvesting the cells, staining with CD8a-BV421 (BioLegend) and FVD, and measuring ≥1 × 103 live CD8+ cells on an LSR II flow cytometer.
Gene expression analysis
Total RNA was isolated from frozen mouse lungs using TRIzol reagent (Life Technologies, Carlsbad, CA) and subsequently purified with an RNeasy Mini Kit (QIAGEN, Hilden, Germany). cDNA was generated from 2 μg of total RNA using SuperScript II Reverse Transcriptase and oligo(dT) (Life Technologies). The expression of indicated target genes was analyzed using a StepOnePlus Real-Time PCR System and TaqMan primer/probe sets (Applied Biosystems, Foster City, CA), with all reactions run in triplicate. Δ Cycle threshold values were calculated using Gapdh as the endogenous housekeeping gene.
A lung ADCA cohort on a tissue microarray (TMA), consisting of 135 cases, was obtained from the University of Pittsburgh Cancer Institute. Patient identifiers were removed; therefore, the study was considered not human subjects research (i.e., Institutional Review Board exempt). Each case was previously annotated as EGFR mutant (n = 31), KRAS mutant (n = 69), or wt for EGFR and KRAS (n = 35). Formalin-fixed paraffin-embedded sections were stained with an anti-human CD8 Ab (catalog number ab4055; Abcam, Cambridge, MA). Immunohistochemical (IHC)-stained TMA slides were scanned in brightfield with a 20× objective using a NanoZoomer Digital Pathology System (Hamamatsu, Hamamatsu City, Japan). Twenty TMA cases (n = 5 EGFR and 16 KRAS) were excluded as lost, noninformative (e.g., poorly stained or nontumorous), or exhibiting high interpunch variability (i.e., SEM ≥ 50% of mean). The number of CD8+ cells per TMA core was recorded blind to genotype and normalized to core area. Individual core counts from two or more replicates were available for most cases, and CD8+ cell counts per square millimeter were averaged across replicates. Cut-off values of low versus high CD8+ cell content were defined by the midpoint. Comparison of TMA cohorts was conducted using the Fisher exact test with a one-tailed p value.
Significant differences between experimental groups were determined in Prism 6 (GraphPad, La Jolla, CA) using unpaired t tests or, for comparing at least three groups, one-way ANOVA with the indicated post hoc test for correction of multiple comparisons. The incidence of CD3+ and/or Foxp3+ cells per lobe was compared between genotypes using the χ2 test. Unless indicated otherwise, data are presented as mean ± SEM. The p values < 0.05 were considered statistically significant.
Egfr- and Kras-driven ADCAs induce a strong inflammatory response
Lung tumor development and associated inflammation were assessed in Egfr, Kras, and Kp53 mice. To allow for the dynamic assessment of TA immune responses, lung tumor–bearing animals and appropriate littermate controls were studied for 6, 10, and 14 wk postinitiation. Consistent with previous studies (24, 26, 34), Egfr, Kras, and Kp53 mice developed neoplastic lesions reminiscent of human disease, from benign hyperplasia and adenomas to malignant ADCAs (Fig. 1A–C). Hyperplasia was observed at all time points, although it was less prevalent in Egfr mice than in Kras-mutant mice (data not shown). The introduction of a secondary mutation in Trp53 increased ADCA formation and amplified tumor growth. Accordingly, tumor burden in 10-wk Kras mice was significantly less than observed in age-matched Kp53 mice (Fig. 1D, p = 0.0231). Analysis of Kp53 mice at the 14-wk time point was precluded by early mortality, but Kras mice remained viable and exhibited 38% lung tumor burden. Body mass measurements, used to noninvasively monitor lung TA morbidity, correlated with tumor burden for all genotypes (Fig. 1E).
Previous investigations of pulmonary inflammation were largely based on the assessment of bronchoalveolar lavage fluid (BALF). Although a well-accepted methodology, BALF studies confine analysis to the airway compartment and limit the number of immune cell types that can be identified. Therefore, to more thoroughly investigate the immune cell composition of the TME, we performed flow cytometric analyses on single-cell suspensions generated from whole-lung tissues. Using the gating strategy shown in Fig. 2A, we identified 13 unique leukocyte populations defined by 16 Ab markers (2Materials and Methods, Supplemental Table I). Kras-, Kp53-, and Egfr-mutant mice display a robust immune response, as evidenced by 3–5-fold increases in total CD45+ cell content compared with normal lung (Fig. 2B–G).
Macrophages were by far the most prevalent immune cell type in the lungs of the three murine ADCA models. Ten weeks after starting doxycycline, Egfr mice exhibited 11-fold increases in macrophage content compared with controls (Fig. 2B). Similarly, by 6 and 10 wk post-Cre induction in Kp53 and Kras animals, respectively, total macrophage cell counts had increased 14- and 17-fold (Fig. 2D, 2F). Of note, macrophage content increased with time only in Kras and Kp53 animals (Supplemental Table II). Because myeloid-derived suppressor cells (MDSC) are composed of monocytic MDSC and granulocytic MDSC subsets, we simply defined these cells as monocytes (CD11b+Ly6C+) and neutrophils (CD11b+Ly6G+). We observed small, but significant, differences in neutrophil content in 10- and 14-wk Egfr tumor-bearing lungs (Fig. 2B, 2C), as well as in 14-wk Kras tumor-bearing lungs (Fig. 2E). By 6 and 10 wk, TA neutrophil content increased 2- and 5-fold, respectively, in Kp53 animals compared with control (Fig. 2F, 2G). Indeed, neutrophil counts in Kp53 lungs were statistically significantly increased compared with all other genotypes at the 10-wk time point (Supplemental Table III). However, this neutrophil signature may merely reflect overall tumor burden, because statistical analysis of cohorts with approximately matched tumor area (i.e., 6-wk Kp53, 10-wk Kras, and 14-wk Egfr) failed to identify significant differences in pulmonary polymorphonuclear cell content.
Impaired NK cell function in Kras-driven ADCA
NK cells are lymphocytes of the innate immune system that play an important role in the host defense against inhaled pathogens (35). NK cell counts in nontumor-bearing lungs were comparable to those of macrophages and granulocytes (Fig. 2). However, unlike the myeloid cell expansion observed with ADCA development, NK populations remained largely unaltered in tumor-bearing lungs. When significant increases were identified (Fig. 2B, 2D–G), the fold changes were small, and NK cell counts decreased over time in Kras animals (Supplemental Table II). NK cells are required for effective tumor immunosurveillance (36), but cancer cells have developed multiple strategies to escape NK cell–mediated cytotoxicity, including downregulation of NKG2D (also known as CD314) (37, 38). Surface expression of NKG2D on NK cells was decreased significantly in Kras- and Kp53-tumor–bearing animals at all time points but exhibited no change in Egfr mice (Fig. 2H). Thus, these murine models of Kras-driven lung ADCA recapitulate an immune-escape mechanism previously described in human lung cancer (39).
Oncogenic drivers dictate lymphocyte recruitment into the ADCA microenvironment
The majority of immunotherapeutic approaches are based on the ability of the adaptive immune system to infiltrate tumors and identify tumor-specific Ags. Therefore, we comprehensively surveyed the lymphocyte subpopulations present within the TME. B cell populations remained unaltered in tumor-bearing Egfr mice (Fig. 2B, 2C) but increased ≥2-fold in Kras and Kp53 mice at all time points (Fig. 2D–G). CD3+ populations were significantly increased in multiple groups (Fig. 2B, 2D–H), but T cell counts were demonstrably greater in Kras and Kp53 mice compared with Egfr mice (Supplemental Table III), even after controlling for tumor burden.
CD4+ Th cell expansion was observed in both Kras-driven tumor models but was limited in Egfr mice (Fig. 3A–F). This pattern was reflected in Treg content, which was significantly lower in tumor-bearing lungs from Egfr mice compared with Kras mice (Supplemental Table III). Interestingly, Treg content increased over time in Kras and Kp53 animals (Supplemental Table II), the only cell type other than macrophages and neutrophils to exhibit such dynamic behavior. Expression of IFN-γ by CD4+ T cells (Th1 cells) and IL-17A by CD3+ T cells was also significantly upregulated in Kras and Kp53 animals compared with control at early and late time points (Fig. 3C–F) but exhibited little to no increase in Egfr animals (Fig. 3A, 3B). Direct comparison of tumor-bearing lungs from all genotypes found higher Th1 and CD3+IL17A+ cell counts in both Kras-driven models compared with Egfr mice (Supplemental Table III). Despite repeated attempts, we were unable to identify IL4+ Th2 cells in our animals (Supplemental Fig. 1); it remains unclear whether this deficiency reflects a true biological absence or, more likely, a technical barrier.
To assess the spatial relationship between lymphocytes and tumor cells, we performed IHC staining for CD3 and Foxp3 on Egfr-, Kras-, and Kp53-tumor–bearing mice. For each model of lung ADCA, immune cell location was identified as TA (or peripheral), TI, or within a neighboring LA. Examples of each are provided in Fig. 3G–J. Staining for CD3 illustrated key differences by genotype, because Egfr-mutant mice displayed markedly fewer TA CD3+ T cells than Kras and Kp53 specimens (Fig. 3K, left panel). TI CD3+ cells were identified in all ADCA models, although they were more prevalent in Kp53 mice (Fig. 3K, middle panel). Similarly, CD3+ cells were present in all LAs, but LAs were significantly less common in Egfr mice but were uniformly present in the other genotypes (Fig. 3K, right panel). Approximately 15% of lobes from Egfr mice contained TA Tregs compared with ∼40% for Kras mice (Fig. 3L, left panel). The primary location of Tregs in all genotypes was within LA structures (Fig. 3L, right panel), with essentially no tumor infiltration observed (Fig. 3L, middle panel). Taken together, the IHC studies confirm the flow cytometry data showing that Egfr mice contain fewer Tregs than the other genotypes and, moreover, demonstrate a paucity of TI lymphocytes, especially compared with Kp53 tumors.
CD8+ lymphocyte content and function differ by lung ADCA subtype
CD8+ T cells are capable of detecting and discriminately eliminating tumor cells (40). Notably, CD8+ cell content differed significantly between models driven by mutant Kras versus Egfr. Expansion of CD8+ cells was observed at all time points in Kras and Kp53 mice but did not occur in the Egfr-mutant cohort (Fig. 3A–F). Even after controlling for tumor burden, Kras-mutant mice displayed greater CD8+ cell content than did Egfr mice (Supplemental Table III). Therefore, we carried out a number of experiments in an attempt to determine the mechanistic basis for this finding and to translate this observation to human disease. Initially, we performed quantitative real-time PCR for key CC and CXC chemokines known to impact immune cell recruitment. Notably, we identified an increase in Cxcl-9 and Cxcl-10 in Kras-mutant lungs compared with Egfr (Fig. 3M), which may explain, at least in part, the differences in lymphocyte content seen between the two lung ADCA models.
To translate these findings to human disease, we performed IHC staining for CD8 on a lung ADCA TMA annotated for KRAS (n = 53 cases) and EGFR (n = 26) mutational status. The frequency of KRAS and EGFR mutations in the cases displaying high versus low CD8 content (the top 50% and bottom 50% of cases, respectively) was assessed, and EGFR mutations were found to be significantly overrepresented in the CD8-low cohort (Fig. 4A). Specifically, 65.4% of EGFR-mutant cases were scored as CD8-low versus 41.5% of KRAS-mutant cases. Thus, similar to the findings in the GEM models presented above, EGFR-mutant lung ADCAs exhibit reduced CD8+ lymphocyte infiltration in human lung cancers compared with KRAS-mutant lung ADCA.
Because CD8+ responses can be blunted by immune checkpoint ligands, we measured PD-L1 expression on macrophages and tumor cells (EpCAM+) by flow cytometry. Interestingly, although PD-L1 expression was decreased in tumor-bearing versus control lungs for Egfr and Kras mice, there was no difference in PD-L1 expression between oncogenic subtypes (Fig. 4B). Because other tumor microenvironmental factors can perturb lymphocyte function, we assessed whether the TME of Egfr mice was more suppressive to lymphocyte proliferation than that found in Kras mice. Egfr and Kras tumor homogenates reduced CD8+ T cell proliferation using CFSE-labeled lymphocytes, but the Egfr homogenates were not more suppressive than Kras (Fig. 4C).
Because the intriguing lack of CD8+ cell expansion in Egfr-mutant tumors suggests a failure of CD8+ cell activation in Egfr mice, we performed a detailed assessment of T cell effector and memory status at the 10- and 14-wk time points using the markers CD62L, CD44, and PD1 (Fig. 4D). No evidence of CD8+ T cell activation was observed, as reflected by the lack of an increase in CD8+PD1+ cells in tumor-bearing Egfr mice (Fig. 4E, 4G). Additionally, the proportion of central memory (CD62L+CD44+) and effector/effector memory (CD62L−CD44+) CD8+ T cells was unchanged, with the majority of these cells still falling into the naive (CD62L+CD44−) category in Egfr mice. In contrast, and consistent with the small, but significant, increase in Th cells shown in Fig. 3A and 3B, CD4+ T cells demonstrated a significant increase in effector/effector memory populations in 10- and 14-wk Egfr mice (Fig. 4F, 4H).
Given the robust increase in CD8+ T cell content observed in tumor-bearing lungs from Kras mice compared with control, we elected to assess the status and functionality of the lymphocyte populations in Kras mice using a CTLA4 mIgG2b (clone 9D9) Ab (MedImmune). Although administration of anti-CTLA4 increased the proportion of CD8+ (Fig. 4J) and CD4+ (Fig. 4K) effector/effector memory cells, this cellular phenotype failed to translate into an altered tumor burden in Kras mice (Fig. 4I). Thus, despite an activated CD8+ T cell response in Kras-mutant mice, tumor progression continued unabated.
Role of IL-17A–producing γδ T cells in Kras-driven lung ADCA
Given the robust expansion of IL-17A+ T cells in Kras-mutant tumor–bearing mice and the known protumor role of Th17 cells in lung ADCA (41), we elected to examine the cellular sources of IL-17A in the lungs of our Kras-mutant mouse models. Surprisingly, the predominant source of IL-17A was found to be γδ T cells rather than CD4+ Th17 cells (Fig. 5A, 5B). An attempt to interrogate the role of IL-17A in lung tumorigenesis by crossing Kras-mutant mice to mice lacking the transcription factor for IL-17A (i.e., RORγt) (42) was stymied by the frequent occurrence of lymphoid neoplasms in these animals. The spontaneously arising lymphomas exhibited thymic (Fig. 5C) and splenic (Fig. 5D) involvement, as well as diffuse infiltration of the liver (Fig. 5E) and lungs (Fig. 5F). Previous studies of a related mouse model of RORγt deficiency (43) similarly identified a high incidence of lymphoma but did not detect pulmonary metastases that, unfortunately, preclude the use of this model in lung tumorigenesis studies. Further efforts to interrogate the specific role of γδ T cells in Kras-mutant lung ADCA revealed that deletion of γδ T cells did not impact tumor burden (Fig. 5G) or the immune cell composition of the TME (Fig. 5H).
A paucity of TI leukocytes in murine and human SCLC
The immune cell composition of the SCLC TME has not been investigated to any extent. Therefore, we profiled the immune content of SCLC using cohorts of Rbp53 mice infected with AdCre. As described previously (44), lung tumors of mainly neuroendocrine histology arose within 40–50 wk of AdCre administration (Fig. 6A). Flow cytometric analysis of SCLC tumor–bearing lungs (gated as shown in Fig. 2) identified a small, but noteworthy, inflammatory presence in Rbp53 mice compared with control (Fig. 6B). The total number of CD45+ leukocytes was increased 2-fold in SCLC, and TA CD45+ cells were also visible by IHC staining (Fig. 6C). Unlike Kras- and Egfr-mutant animals (Fig. 1A, 1B), SCLC tumors presented as large discrete foci, and little hyperplasia was observed. CD45+ cells were consequently clustered at the periphery of the SCLC lesions (Fig. 6D), without the inflammatory field effect frequently observed in the ADCA models. Few TI leukocytes were detected (Fig. 6E).
The major immune component of SCLC was found to be CD3+ T lymphocytes (Fig. 6B). This population included a 7-fold increase in the number of γδ T cells and a strong trend toward increased CD4+ Th cells (p = 0.0698). In marked contrast to the ADCA models, expansion of innate immune cells in SCLC tumor–bearing lungs was minimal, with only a 2-fold increase observed in macrophages and a nonsignificant increase in neutrophils (p = 0.0886). To further investigate this phenomenon, we compared the ratio of CD3+ T cells/myeloid cells (macrophages, neutrophils, monocytes, and eosinophils) and found a pronounced lymphocyte-dominant signature in SCLC versus all ADCA models (Fig. 6F). Egfr mice presented the smallest CD3/myeloid ratio and were also significantly different from Kras mice.
As part of an ongoing study of the immune composition of human lung cancer, we obtained two surgical specimens with confirmed small cell pathology. Because resecting SCLC is rarely clinically indicated, these two specimens represented a unique opportunity to measure the immune cell composition present within the SCLC TME. Therefore, we performed flow cytometry analyses on single-cell suspensions generated from these two cases. Similar to the findings in the GEM models, TA inflammation was discernibly lower in SCLC compared with five ADCA specimens; CD45+ cells accounted for a mere 16.2% of live cells in the SCLC resections compared with 81.9% in ADCA (Fig. 6G).
Lung cancer is a heterogeneous disease that can be divided into distinct subtypes based on molecular and cellular characteristics (45). In this study, we tested the hypothesis that these subtypes dictate the inflammatory response to cancer by immune profiling the lung TME in a mouse model of SCLC and in three molecularly distinct models of NSCLC (Fig. 6H). We found that Egfr and Kras mutations give rise to distinct immune responses characterized by differential expansion of B cells, CD8+ T cells, Tregs, and IL-17A–producing T cell populations. Although loss of Trp53 promoted malignancy, it had minimal effect on immune cell composition within the Kras TME. We further demonstrate that SCLC possesses an overall reduced inflammatory presence compared with NSCLC, and one in which lymphocytes predominate over myeloid lineage cells. Therefore, mutational profile and histological origin actively shape the immune contexture of lung cancer, a finding that may have important clinical ramifications.
The strong macrophage field responses that occur in mutant Kras- and Egfr-driven mouse ADCAs are seldom observed in human lung cancer and represent a potential limitation of these GEM models. In Kras mice, this phenomenon of alveolar macrophages flooding the airspaces was likened to desquamative interstitial pneumonitis, a rare interstitial lung disease with similar pathology (46). Moreover, the conditional mouse models in this study use varied induction methodologies (i.e., AdCre- or doxycycline-regulated transgene expression). Although we cannot exclude potential confounding effects of viral infection or doxycycline consumption on the tumor immune response, we attempted to correct for these variables by using adenovirus- or doxycycline-exposed wt animals for the relevant control cohorts.
Few gene-specific investigations of the mouse lung TME have been conducted, and no comprehensive effort has been made to compare and contrast different molecular and histological models of lung cancer. However, our results validate findings from several earlier studies of lung TA inflammation. We were unsurprised to identify macrophages as the dominant immune cell presence in mouse ADCAs given that strong TA macrophage responses were identified in mutant Egfr (21) and Kras (19, 41, 47) mouse lung tumor models. Likewise, as we described in this article, neutrophils were shown to be a modest, but important, component of Kras-driven, but not Egfr-driven, mouse lung ADCA (19, 21, 41). Because the majority of prior data in this regard relied on BALF cell counts, we used flow cytometry to better define the quality of the immune response. Using this methodology, we found that recruitment of lymphoid lineage cells varies greatly among ADCA models, because EgfrL858R mice exhibited a paucity of B cells, CD8+ T cells, Tregs, and IL-17A–producing T cells compared with the Kras and Kp53 lung TME.
The most clinically relevant finding in this study is the lack of a CD8+ lymphocyte response in Egfr-mutant mice and EGFR-mutant human lung ADCA specimens compared with their KRAS-mutant counterparts. Markers of effector/memory status failed to reveal any evidence of CD8+ T cell activation or differentiation in Egfr mice. This suggests that EGFR mutation may not elicit an Ag-driven immune response. Although the same could be said for mutant KRAS, we were able to demonstrate an increase in activated and effector memory CD8+ cells in the Kras mouse model. Furthermore, TI lymphocyte populations that specifically target mutant KRASG12D were identified recently in colon cancer (48). Despite the presence of activated CD8+ cells, tumor growth continued in Kras mice, even with the addition of an anti-CTLA4 therapeutic Ab. Our interpretation of this data is that increases in activated CD8+ cells within the TME in Kras-mutant mice do not impact tumor growth unless they are tumor reactive. Specifically in this case, tumor-derived chemokines, such as Cxcl-10, are likely to increase the number of TA lymphocytes. Although anti-CTLA4 Ab therapy drove an increase in effector T cells, these cells would not be expected to reduce tumor burden if they did not recognize a TA Ag. It is also possible that anti-CTLA4 monotherapy will prove ineffective but that combined immunotherapeutic regimens (e.g., anti-CTLA4 + anti–PD-L1) will prove effective. With respect to human lung ADCA, the association of EGFR-mutant cancers with never-smoker status suggests that these tumors are genetically simplified and, unlike smoking-associated KRAS-mutant cancers, may not possess sufficient mutational burden to harbor neoantigens (49, 50). The mouse models described in this article were not exposed to cigarette smoke or other carcinogens, eliminating this proposed explanation for the differential CD8+ responses that we observed. However, at this time we cannot exclude the possibility that KRAS-mutant human and mouse lung ADCAs achieved the same phenotype of high CD8+ T cell infiltration through different mechanisms of action.
Tregs and IL17A+ T cells have emerged as important cell populations in multiple mouse models of cancer (41, 51, 52), and both cell types exhibited notable patterns of expression or localization in the murine ADCA models. Although TGF-β and IL-6 generate gradients leading independently to Treg or Th17 differentiation, we observed concurrent increases in both populations in Kras- and Kp53-mutant tumor–bearing lungs. Notably, we found that the major source of IL-17A in Kras-mutant ADCA was γδ T cells and not Th17 cells. Because IL-17A deficiency was shown to reduce lung tumor growth (41), and IL-17A–producing γδ T cells are known to promote breast (53) and pancreatic neoplasia (52), these findings suggested to us that expansion of a pulmonary IL-17A–producing γδ T cell subset might overshadow the tumor-surveillance role traditionally ascribed to γδ T cells (54). However, Kras-mutant γδ T cell–deficient mice displayed equivalent lung tumor burden and strikingly similar immune profiles to their γδ T cell–competent counterparts. Therefore, although IL-17A is an important signaling component in the immune landscape of lung ADCA, IL-17A+ γδ T cells appear to contribute little to the process of lung tumorigenesis.
Tregs, in contrast, appear to play a particularly important role in the Kras-mutant lung TME, because they were the only nonmyeloid lineage population to expand over the course of tumor development. Moreover, Tregs display a unique anatomic location in lung ADCA. They are rarely associated with the tumor itself; instead, they are frequently found within LA structures that are believed to function as a local site of Ag presentation and are correlated with good clinical outcomes in NSCLC (55). The presence of Tregs in these structures was recently shown to be detrimental to the generation of an effective immune response in murine lung ADCA (56), highlighting the importance of Treg-targeting strategies for the clinical management of lung cancer patients.
The immune cell composition of SCLC has not been well studied. Our findings in Rbp53 mice point to a less robust, but more lymphocyte-predominant, host immune response to murine SCLC than to ADCA. Moreover, when we analyzed the immune cell content of two human SCLC cases, we identified a strikingly similar immune profile of sparse CD45+ cell content. Although we acknowledge the inherent limitations of n = 2 studies, patients diagnosed with SCLC seldom undergo lung resection (57), making access to such specimens exceedingly rare. Solid tumor malignancies demonstrating the best responses to current ICI therapies are those with high mutational burdens and/or a history of cigarette smoke exposure (e.g., melanoma, head and neck squamous cell carcinoma, and urinary bladder cancer), both traits common to SCLC (2). In light of these correlations, it is tempting to speculate that SCLC patients would exhibit good responses to ICI therapy. However, initial reports suggest that success rates for anti-PD1 therapy in SCLC are, at best, only comparable to NSCLC (58, 59). Our preliminary findings with respect to the immune cell composition in SCLC suggest that the presence of redundant immune-suppressive factors would not be a likely source of treatment failure, which is almost certainly an important concept in NSCLC. These findings point to potentially unique features of the SCLC TME (e.g., matrix protein composition) that require additional study.
A robust CD45+ immune response was observed in the ADCA mouse models and human lung ADCA patients. Leukocytes account for nearly 75% of total cellular content in human ADCA, which is an even greater proportion than we identified in mice (∼55%). With the exception of the aforementioned exaggerated macrophage responses, the robust and diverse immune landscape observed in GEM models of ADCA approximates that seen in human lung cancers (60). Driving mutations, such as in Egfr and Kras, substantially impact the TME through the release of bioactive molecules, which is very well reflected in these GEM models. One potential shortcoming of these models is the genetic simplicity of the tumors, which rely on a single driving mutation. In contrast, human NSCLC harbors an average of ∼150 distinct mutations per case (61). Efforts are underway to construct mouse models of cancer that harbor a greater abundance of single nucleotide variations and, thus, potential neoantigens. However, EGFR-mutant cancers in nonsmokers are typically genetically simplified (5), such that the Egfr-mutant mice described in this article likely constitute an excellent representation of the genetic component of the cancer cell and the immune composition of the TME.
The emergence of ICIs has been a tremendous advance; unfortunately, the majority of lung cancer patients in clinical trials failed to respond to ICI therapy (8–11). In addition to the PD-1/PD-L1–based drugs currently in use, novel ICI agents are likely to emerge in the near future. Our findings argue that the cellular and molecular characteristics of lung cancer may provide an important framework for patient-targeted immunotherapy. Furthermore, preclinical testing of future immunotherapy agents should be performed in genetically and histologically diverse model systems to enable the assessment of tumor subtype–specific efficacy.
We thank the Fred Hutchinson Cancer Research Center Experimental Histopathology and Flow Cytometry shared resource facilities, the University of Washington Histology and Imaging Core, and the Viral Vector Core Facility at the University of Iowa Carver College of Medicine. We also thank MedImmune for supplying the anti-CTLA4 therapeutic Ab.
This work was supported by National Institutes of Health/National Heart, Lung, and Blood Institute Grant R01 HL108979 (to A.M.H.), European Commission Grant FP7-PEOPLE-2012-IOF 331255 (to J.K.), and the Fred Hutchinson Cancer Research Center.
The online version of this article contains supplemental material.
Abbreviations used in this article:
adenoviral Cre recombinase
bronchoalveolar lavage fluid
Fixable Viability Dye
genetically engineered mouse
immune checkpoint inhibitor
myeloid-derived suppressor cell
median fluorescence intensity
NK group 2, member D receptor
non–small cell lung cancer
small cell lung cancer
regulatory T cell
The authors have no financial conflicts of interest.