Insulin receptor substrate-1 deficiency drives a proinflammatory phenotype in KRAS mutant lung adenocarcinoma

Significance

Kirsten rat sarcoma viral oncogene homolog (KRAS) mutant lung adenocarcinoma remains an intractable lung cancer subtype for which efficacious targeted therapies do not exist. This study identified a subpopulation of KRAS mutant lung adenocarcinoma with reduced insulin receptor substrate-1 (IRS-1) content and showed that this group is particularly amenable to Janus kinase/signal transducers and activators of transcription (JAK/STAT) inhibition. This finding is the first definitive report of a prohost role for IRS-1 in a solid tumor malignancy in humans. Furthermore, the data supports a novel role for IRS-1 in the sculpting of immune cell composition within the tumor microenvironment.

Abstract

Insulin receptor substrate-1 (IRS-1) is a signaling adaptor protein that interfaces with many pathways activated in lung cancer. It has been assumed that IRS-1 promotes tumor growth through its ability to activate PI3K signaling downstream of the insulin-like growth factor receptor. Surprisingly, tumors with reduced IRS-1 staining in a human lung adenocarcinoma tissue microarray displayed a significant survival disadvantage, especially within the Kirsten rat sarcoma viral oncogene homolog (KRAS) mutant subgroup. Accordingly, adenoviral Cre recombinase (AdCre)-treated LSL-Kras/Irs-1^fl/fl (Kras/Irs-1^−/−) mice displayed increased tumor burden and mortality compared with controls. Mechanistically, IRS-1 deficiency promotes Janus kinase/signal transducers and activators of transcription (JAK/STAT) signaling via the IL-22 receptor, resulting in enhanced tumor-promoting inflammation. Treatment of Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice with JAK inhibitors significantly reduced tumor burden, most notably in the IRS-1-deficient group.

Lung cancer continues to be the leading cause of cancer deaths worldwide (1), and in the United States it accounts for more than 160,000 deaths per year with 5-y survival rates of just ∼15% (2). Lung cancer is a heterogeneous disease that is typically subdivided into small-cell lung cancer (SCLC) and non–small-cell lung cancer (NSCLC), which constitute ∼15% and 85% of cases, respectively. NSCLC is further subclassified into multiple histologic subtypes, including lung adenocarcinoma (L-ADCA; ∼60%) and lung squamous cell carcinoma (L-SCCA; ∼20%) (3). L-ADCAs frequently possess a driving mutation in a key oncogene, such as Kirsten rat sarcoma viral oncogene homolog (KRAS) and the EGF receptor (EGFR). Whereas EGFR mutant cancers can be addressed with novel targeted therapies, KRAS mutant tumors remain largely untargetable (4).

Numerous studies have examined the impact of a specific driving mutation on the function of a particular signaling pathway, though none have addressed the role of insulin receptor substrate-1 (IRS-1) (5). Of the numerous proteins involved in pathway signaling, IRS-1 is somewhat unique in that it interfaces with many different pathways, including the PI3K, extracellular signal regulated kinase (MEK/ERK), and Janus kinase/signal transducers and activators of transcription (JAK/STAT) signaling pathways (6, 7). IRS-1 is afforded this promiscuity as a result of possessing numerous binding domains, including an N-terminal pleckstrin homology (PH) domain, a phosphotyrosine binding (PTB) domain, and a carboxy terminus with multiple serine and tyrosine phosphorylation sites (6). IRS-1 is a signaling adaptor protein best known for mediating canonical signaling from both the insulin receptor (IR) and the insulin-like growth factor receptor (8). The majority of research investigating the function of IRS proteins has naturally been performed in the setting of glucose metabolism and diabetes in metabolically active tissues (muscle and adipose). In this context, IRS-1 has been appropriately characterized as a positive effector of growth factor (9, 10). However, the impact of IRS-1 on pathway activity in cancer cells, where aberrant signaling is frequently encountered, has not been adequately addressed.

An emerging concept in cancer biology is the role of the tumor microenvironment (TME) in tumor progression (11), and the role of aberrant pathway signaling in sculpting the TME (12, 13). Immune cells represent a major component of the TME, frequently comprising over half of the cells in resected tumor specimens. It has become evident that inflammatory responses increase tumor initiation and progression, because immune cells are capable of supplying tumors with the growth factors, cytokines, and reactive oxygen species required to promote proliferation, survival, angiogenesis, invasion, and metastasis (14, 15). The mechanisms by which cancers generate tumor-promoting microenvironments must be further explored to identify novel therapeutic agents that address the TME.

Given the prominent role of IRS-1 in signaling pathways commonly hyperactive in L-ADCA, we undertook an independent study of this protein within the adenocarcinoma subtype. Surprisingly, we identified a counterparadigmatic and prohost role for IRS-1 specifically in KRAS mutant L-ADCA via alterations in cancer cell signaling that subsequently impacted the cellular composition of the TME.

Results

KRAS Mutant, IRS-1^low Lung Adenocarcinoma Subgroup Displays Reduced Patient Survival.

To evaluate associations between IRS-1 protein content and L-ADCA patient outcomes, we stained and analyzed a human tissue microarray (TMA) consisting of 135 primary tumor specimens that had been annotated for KRAS and EGFR mutation status, and for which survival data were available. Immunohistochemistry staining of IRS-1 was quantified using an Aperio Genie system. H-scores were generated by assessing the IRS-1 staining intensity within the tumor compartment. Examples of H&E-stained, IRS-1-stained, and Genie overlay images from cases displaying abundant (IRS-1^high) or reduced IRS-1 staining (IRS-1^low) are shown in Fig. 1 A–C.

Fig. 1.

IRS-1 predicts patient survival in L-ADCA. (A and B) Representative images of IRS-1^low (H-score <130) and IRS-1^high (H-score >130) human L-ADCA TMA sections stained for (A) H&E and (B) IRS-1. (C) Example of tumor compartment-specific quantification of IRS-1 staining using Aperio Genie. (D–G) Kaplan–Meier estimates of overall survival for the (D) entire lung adenocarcinoma cohort (n = 123, P = 0.02, (E) KRAS mutant cases (n = 62, P = 0.006), (F) EGFR mutant cases (n = 29, P = 0.18), and (G) Non-KRAS, non-EGFR mutant cases (n = 32, P = 0.83). All P values determined from log-rank tests.

As a continuous variable, IRS-1 expression did not predict overall survival (OS) (hazard ratio = 0.99, 95% CI 0.98–1.00, P = 0.16). Threshold effects were explored using recursive partitioning, identifying an H-score cut point of 130 (Fig. 1D; log-rank test P value = 0.02). Classification by IRS-1 did not appear a proxy for other patient or tumor characteristics (Table S1). Evaluated in subgroups by mutational status, the prognostic signal for dichotomized IRS-1 appeared to be driven by tumors with KRAS mutations (Fig. 1E). In these patients, the median OS was 50 mo for IRS-1^low (95% confidence interval 30–84 mo) and 131 mo for IRS-1^high (71–131 mo). EGFR mutant, and EGFR and KRAS wild-type (not KRAS or EGFR mutant) subgroups did not demonstrate IRS-1 as a strong prognostic factor (Fig. 1 F and G). The effect of IRS-1 persisted in a multivariable model. Smoking status (current/former/never, P = 0.18), smoking consumption (pack years, P = 0.55), and age (P = 0.96) were not associated with overall survival in univariate or multivariate models (Table S1). Therefore, the primary multivariable model included pathologic stage and sex. In this model, the hazard ratio for dichotomized IRS-1 was 1.8 (95% CI 1.1–3.1): patients with IRS-1^low tumors had a hazard of death nearly twice that of those with IRS-1^high (Wald test P value = 0.03).

Table S1.

TMA patient characteristics by IRS-1 category

Characteristic	All	Average IRS-1 H-score		P value
	(n = 123)	Below 130 (n = 47)	130 or higher (n = 76)
	N (%)	N (%)	N (%)
Sex				0.85*
Female	72 (59)	28 (60)	44 (58)
Male	51 (41)	19 (40)	32 (42)
Age, y				0.96†
Mean (SD)	65 (10)	65 (12)	66 (9)
Median (IQR)	65 (58, 74)	66 (56, 75)	65 (60, 74)
(Min, max)	(39, 84)	(39, 84)	(41, 84)
Race				0.1.0‡
White	111 (90)	43 (91)	68 (89)
Black	6 (5)	2 (4)	4 (5)
Pacific Islander	2 (2)	1 (2)	1 (1)
Unknown	4 (3)	1 (2)	3 (4)
Pathologic stage				0.37†
1	68 (55)	25 (53)	43 (57)
2	25 (20)	7 (15)	18 (24)
3/4	29 (24)	14 (30)	15 (20)
NA§	1 (1)	1 (0)
Phenotype				0.99*
EGFR	29 (23)	11 (23)	18 (23)
KRAS	62 (50)	24 (51)	38 (50)
WT	33 (27)	12 (26)	20 (26)
Smoking status				0.18*
Active	36 (29)	15 (32)	21 (28)
Former	71 (58)	23 (49)	48 (63)
Never	16 (13)	9 (19)	7 (9)
Pack years				0.55*
30 or more	68 (55)	25 (53)	43 (57)
Under 30	40 (33)	17 (36)	23 (30)
Unknown	15 (12)	5 (11)	10 (13)
IRS-1 H-score
Mean (SD)	139 (30)
Median (IQR)	138 (120, 158)
(Min, max)	(61, 229)

IQR, interquartile range.

^*χ² test.

^†Wilcoxon rank-sum test (pathologic stage IA-4).

^‡Fisher’s exact test.

^§Pathologic stage incomplete; no lymph nodes removed.

Kras/Irs-1^−/− Mice Display Increased Tumor Burden and Decreased Survival.

To further interrogate the phenotype identified in the human TMA study, we generated Lox-Stop-Lox-Kras/Irs-1^+/+ and Lox-Stop-Lox-Kras/Irs-1^fl/fl mice on a pure 129.SvJ genetic background. Mice were treated intratracheally with adenoviral Cre recombinase (AdCre) at 8 wk of age to activate mutant Kras expression and to delete Irs-1 (Fig. 2D). AdCre-treated Lox-Stop-Lox-Kras/Irs-1^+/+ mice will be denoted as Kras/Irs-1^+/+ and AdCre-treated Lox-Stop-Lox-Kras/Irs-1^fl/fl mice will be denoted as Kras/Irs-1^−/−. The mice were subsequently studied over a time course ranging from 4 to 20 wk post-AdCre. Similar to the findings in human L-ADCA, Kaplan–Meier survival curves demonstrated inferior survival for Kras/Irs-1^−/− mice compared with Kras/Irs-1^+/+ control mice (P < 0.0001; Fig. 2A). Kras/Irs-1^−/− mice displayed a median survival of 14 wk, compared with a median survival of 20 wk in the Kras/Irs-1^+/+ control group. Additionally, tumor burden was quantified at 4, 8, and 12 wk post-AdCre (Fig. 2 B and C). Statistically significant increases in tumor burden were observed in IRS-1-deficient mice at the 8- and 12-wk time points, representing a doubling and tripling of the tumor burden identified in control mice. Although we assessed for local and distant metastases, we were unable to identify any in either group.

Fig. 2.

Increased tumor burden and mortality in Kras/Irs-1^−/− mice. (A) Kaplan–Meier estimates of overall survival for AdCre-treated Kras/Irs-1^+/+ (n = 41) and Kras/Irs-1^−/− (n = 44) mice. P < 0.0001, log-rank test. (B) Tumor area percent (percent of lung tissue occupied by tumor) for Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice at 4, 8, and 12 wk post-AdCre. (C) Representative H&E-stained images for Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice 4, 8, and 12 wk post-AdCre. (D–F) Representative (D) Irs-1, (E) CC3, and (F) K_i-67 stained images for Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice at 12 wk post-AdCre. (G) Quantification of K_i-67 staining. All n = 6 per group. Bars ± SEM. *P < 0.05.

In attempts to distinguish between differential rates of apoptosis vs. cellular proliferation as the cause of increased tumor burden in Kras/Irs-1^−/− mice, we performed cleaved caspase-3 and K_i-67 staining (Fig. 2 E and F). We were unable to identify tumor cells staining positively for cleaved caspase-3 (CC3) in either group, and did not quantify apoptosis further. Quantification of K_i-67 staining demonstrated a statistically significant increase in tumor cell proliferation in the Kras/Irs-1^−/− group (Fig. 2G). Therefore, the increased tumor burden in Kras/Irs-1^−/− mice results from enhanced proliferation, and not decreased apoptosis.

Increased Neutrophilic Inflammation in Kras/Irs-1^−/− Mice.

Surprisingly, the most striking initial phenotypic difference apparent in the Kras/Irs-1^−/− mice was elevated inflammatory cell content in the bronchoalveolar lavage (BAL) fluid (Fig. 3 A and B). Further investigation revealed robust increases in macrophage and neutrophil content (Fig. 3 A and B). At 12 wk, neutrophils were approximately fivefold higher in Kras/Irs-1^−/− BAL, and Ly6G staining revealed the presence of tumor-associated neutrophils (Fig. 3C). Quantitative PCR (qPCR) was performed on lung tissue to assess differential expression of CC and CXC chemokines known to recruit myeloid lineage cells to the TME. Kras/Irs-1^−/− mice displayed increases in most of these chemokines compared with Kras/Irs-1^+/+ mice (Fig. 3D). Specifically, we identified significant increases in CCL-2, -3, -4, and CXCL-1, -2, and -5, consistent with the BAL cellular content data. To validate the qPCR findings at the protein level, we performed ELISA analysis for CCL-2, CCL-4, CXCL-1, and CXCL-2. Consistent with the qPCR data, Kras/Irs-1^−/− mice possess increased amounts of all four chemokines in their BAL fluid, compared with Kras/Irs-1^+/+ mice (Fig. 3E).

Fig. 3.

Increased tumor-associated inflammation in Kras/Irs-1^−/− mice. (A) BAL fluid content of neutrophils, macrophages, and lymphocytes from Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice at 4, 8, and 12 wk post-AdCre. n = 6 per group. (B) Representative Hema-3-stained cytospins performed on BAL fluid from Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice at 12 wk post-AdCre. (C) Representative anti-Ly6G (neutrophil marker)-stained sections from Kras/Irs-1^+/+ and Kras/Irs-1^−/− at 12 wk post-AdCre. (D) Real-time PCR values for the listed genes from Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice at 8 wk post-AdCre. n = 4 per group. Results expressed as fold change from Kras/Irs-1^+/+. (E) Chemokine expression (pg/mL) in BAL fluid from Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice at 8 wk post-AdCre. n = 4 per group. Results expressed as fold change from Kras/Irs-1^+/+. All bars ± SEM. *P < 0.05.

To verify a tumor-promoting role for tumor-associated neutrophils, we treated both cohorts of mice with a neutrophil-depleting Ly6G antibody (clone 1A8), which reduced both tumor-associated neutrophilic inflammation and tumor growth (Fig. S1), consistent with prior reports (16, 17). Not surprisingly, the impact of anti-Ly6G on tumor burden was greater in the IRS-1-deficient group, in which the tumor microenvironment is enriched in neutrophil content.

Fig. S1.

Neutrophil depletion reduces tumor burden in Kras/Irs-1^−/− mice. (A) Representative H&E-stained sections from Kras/Irs-1^+/+ mice treated with anti-LY6G or isotype control beginning at 6 wk post-AdCre, three times per week for 4 wk. (B) Tumor area percent (percent of lung occupied by tumor) for Kras/Irs-^+/+ mice treated with either anti-LY6G or isotype control, as above. n = 3 each group. Bars ± SEM. *P < 0.05. (C) Representative H&E-stained sections from Kras/Irs-1^−/− mice treated with anti-LY6G or isotype control beginning at 6 wk post-AdCre, three times per week for 4 wk. (D) Tumor area percent (percent of lung occupied by tumor) for Kras/Irs-1^−/− mice treated with either anti-LY6G or isotype control, as above. n = 3 each group. Bars ± SEM. *P < 0.05.

IL-22 Is Required for Increased Chemokine Response in IRS-1-Deficient Cancer Cells.

To further evaluate the differences in chemokine production observed between Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice, we generated IRS-1-deficient A549 (KRAS mutant) L-ADCA cells using shRNA approaches. We were not able to observe an increase in cytokine or chemokine production in a microarray study of IRS-1-silenced A549 cells under standard serum conditions (Table S2). We then reasoned that a factor present in the TME, but absent in cultured cancer cells, might be responsible for the in vivo observations in Kras/Irs-1^−/− mice. Of the factors known to be present within the TME, strong evidence exists to support a role for the Th17 family of cytokines in the amplification of myeloid cell-recruiting chemokines (18). Recently, Chang et al. (19) found that IL-17A-deficient KRAS mutant mice displayed reduced tumor burden, tumor-associated inflammation, and CC and CXC chemokine production in vivo. Therefore, we tested the ability of IL-6, IL-17A, and IL-22 to induce greater CC and CXC chemokine expression from IRS-1-deficient cancer cells in vitro. A549 cells silenced for IRS-1 (shIRS-1) and short hairpin RNA (shRNA) vector control (shCon) were stimulated with the above cytokines and qPCR was performed to assess for differences in cytokine and chemokine production. As expected, IL-6, IL-17, and IL-22 increased cytokine and chemokine expression in A549 cells. However, only IL-22 induced a greater increase in cytokine and chemokine production in IRS-1-silenced cells than it did in shCon cells (Fig. 4A). Because survival differences in human L-ADCA were limited to the KRAS mutant group, we repeated these experiments in 201T L-ADCA cells, which harbor WT KRAS alleles. TGF-β was induced by IL-6, IL-17A, and IL-22 treatments, but other cytokines and chemokines were not up-regulated by IRS-1 deficiency in the absence of KRAS mutation (Fig. 4B). We were further able to reproduce these findings in the KRAS mutant H460 lung cancer cell line (Fig. 4J). Because IRS-1 did not impact patient outcomes in EGFR mutant patients, we repeated these assays in H1650 cells, which are EGFR mutant but KRAS WT. Similar to the findings in KRAS WT 201T cells, H1650 cells did not produce excessive CC/CXC chemokines in response to IL-22 (Fig. 4L), further highlighting the requirement for mutant KRAS in this setting.

Table S2.

Selected gene expression data for IRS-1 deficient cells

Target ID	logFC	Avg exp	Adj P value	Sig
CCL2	−0.156268318	9.100781583	0.967169208	0
CCL3	−0.136369606	7.075929816	0.966896737	0
CCL4	−0.058376955	6.949231382	N.A.	0
CCL5	−0.063231897	8.038061882	0.968616821	0
CXCL1	−0.125905515	8.700773233	0.976061469	0
CXCL2	−0.078966537	7.97277147	0.979444897	0
CXCL3	−0.006735041	7.136750627	0.994868305	0
CXCL5	−0.044481868	10.81123675	0.979004734	0
CXCL12	−0.059039139	6.977275872	N.A.	0
IL-17A	−0.00687516	7.054173566	N.A.	0
IL-17F	−0.034182281	6.875021092	N.A.	0
IL-22	0.006286507	6.957046328	N.A.	0
IL-6	−0.014591663	7.103573952	0.991806462	0
TGFB1	−0.055214787	6.951293923	N.A.	0
CSF2	−0.009494854	7.060236202	N.A.	0

Gene expression assessed using HumanHT-12 Expression BeadChip (Illumina, Inc.) and imaged using an Illumina iScan system. Results reported as log fold-change of siIRS-1 compared with siSCR control A549 cells. Adj P value, adjusted P value; Avg exp, average expression; logFC, log fold change; N.A., not applicable; Sig, significance; Target ID, target identification.

Fig. 4.

IRS-1 deficiency enhances IL-22 signaling. (A and B) Real-time PCR values for the listed genes from lentiviral control (shCon) and IRS-1-silenced (shIRS-1) (A) A549 and (B) 201T cells treated with IL-6, IL-17A, or IL-22. Results expressed as fold changed in gene expression from shCon values ± SEM. *P < 0.05. (C and D) Western blot and band densitometry (n = 4) for shCon and shIRS-1 A549 cells with and without IL-22 stimulation. (E and F) Western blot and band densitometry (n = 4) for shCon and shIRS-1 201T cells with and without IL-22. (G) Real-time PCR values for listed genes from shCon and shIRS-1 A549 cells pretreated with a JAK inhibitor (Millipore no. 420099) and stimulated with IL-22. Results expressed as fold change from control values. (H) Western blot probed for STAT3 and pSTAT3 from A549 cells treated with a JAK inhibitor over the indicated concentrations and stimulated with IL-22. (I) Viability assay comparing DMSO and JAK inhibitor-treated A549 cells. Results expressed as fold change from control values. (J and L) Real-time PCR values for the listed genes from shCon and shIRS-1 (J) H460 and (L) H1650 cells treated with IL-22. Results expressed as fold changed in gene expression from control. (K and M) Western blot for shCon and shIRS-1 (K) H460 cells and (M) H1650 cells with IL-22 stimulation. All bars ± SEM. *P < 0.05.

IL-22 Stimulation Causes Exaggerated pSTAT3 Production in KRAS Mutant, IRS-1-Silenced Cells.

IL-22 is a known inducer of JAK/STAT signaling (20). To investigate the impact of IRS-1 deficiency on IL-22-induced JAK/STAT signaling, we treated shIRS-1 and shCon A549 and 201T cells with IL-22, both in the presence and absence of a JAK inhibitor. As expected, IL-22 stimulation increased the production of pSTAT3 compared with non–IL-22-stimulated control cells, for all conditions (Fig. 4 C and E). Predictably, the addition of a synthetic JAK inhibitor (Millipore) abrogated pSTAT3 production and cytokine/chemokine induction under all conditions, though it did not impact cellular viability (Fig. 4 G–I). Notably, IRS-1-silenced A549 cells (KRAS mutant) displayed an even greater increase in pSTAT3 production upon IL-22 stimulation than did vector control A549 cells (Fig. 4 C and D). In contrast, IL-22-induced pSTAT3 production was similar in magnitude between IRS-1-silenced and vector control 201T cells (KRAS WT; Fig. 4 E and F). We were able to reproduce the alterations in pSTAT3 production in an independent KRAS mutant cell line (H460) and an additional KRAS WT cell line (H1650; Fig. 4 K–M). To demonstrate that this mechanism was operative in vivo, formalin-fixed, paraffin-embedded (FFPE) sections from Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice were subjected to IHC for pSTAT3 (Fig. 5A). Tabulation of the slides confirmed greater pSTAT3 production in Kras/Irs-1^−/− mice (Fig. 5B), compared with controls, thereby validating the in vitro findings with respect to pSTAT3.

Fig. 5.

Increased IL-22RA1 content in IRS-1-silenced cells. (A) Representative images and (B) quantification for pSTAT3-stained sections from Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice at 12 wk post-AdCre. n = 6 per group. (C) Representative images and (D) quantification for IL22RA1-stained sections from Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice at 12 wk post-AdCre. n = 5 per group. (E) Control and IRS-1-silenced A549 cells were pretreated with cycloheximide before IL-22 stimulation. Western blots probed for IL-22RA1 and p85 were performed on membrane fraction. (F) Representative Western blots for the indicated proteins in shCon and shIRS-1 A549, H460, 201T, and H1650 cells. All bars ± SEM. *P < 0.05.

IRS-1 Deficiency Prolongs IL-22 Receptor Alpha 1 Half-Life via pGSK-3β Production.

The increased production of pSTAT3 by IRS-1-deficient cancer cells compared with control was observed while using an equivalent concentration of IL-22. This finding suggests that the production of pSTAT3 induced by each molecule of IL-22 is greater for IRS-1-silenced cells compared with controls. One possible mechanism that would explain these observations would be a prolongation of the IL-22R half-life in IRS-1-deficient cells. IHC for IL-22 receptor alpha 1 (IL-22RA1) was performed on FFPE sections from Kras/Irs-1^−/− mice and Kras/Irs-1^+/+ controls. Similar to the findings with respect to pSTAT3, Kras/Irs-1^−/− tumors displayed greater than twice the IL-22RA1 staining compared with Kras/Irs-1^+/+ controls (Fig. 5 C and D); to demonstrate this more clearly, IRS-1-silenced and vector control A549 cells were treated with cycloheximide before IL-22 stimulation and subsequent tracking of IL-22RA1 cycling. The IL-22RA1 signal is gradually lost from the membrane as it is internalized following activation, tagged for degradation, and eventually completely absent in the membrane fraction (Fig. 5E). Because the cells have been treated with cycloheximide, they are incapable of repopulating the cell surface with new receptor. In contrast, by the conclusion of the experiment in IRS-1-deficient cells, the IL-22RA1 membrane content has been restored, because the receptor has been recycled to the membrane, never having been degraded.

Recently, pGSK-3β has been shown to phosphorylate the cytoplasmic tail of IL-22RA1, which inhibits the ability of ubiquitinases to tag the receptor for degradation (21); this results in a prolongation of IL-22RA1 half-life. Because IRS-1 is a homeostatic binding partner of the p85 subunit of PI3K, and well known to impact PI3K signaling, we investigated the impact of IRS-1 deficiency on pAKT and pGSK-3β production in KRAS mutant and KRAS WT cells. IRS-1 silencing had little effect on pAKT and pGSK-3β production in KRAS WT 201T and H1650 cells (Fig. 5F). However, IRS-1 deficiency in the presence of mutant KRAS (A549 and H460) actually increased both pAKT and pGSK-3β (Fig. 5F). As a signaling adaptor protein, IRS-1 is able to interact with p85 within the cytosol, following phosphorylation by a cell-surface receptor; this is in contrast to the usual receptor tyrosine kinases that interact with p85 near the inner leaflet of the cellular membrane, where lipid substrates are abundant. In the absence of IRS-1, the subcellular location of p85 relocates predominantly to the cell membrane, regardless of KRAS mutational status (i.e., it occurs equally in all cell lines tested; Fig. 5F). Similar observations were previously made in adipocytes (22). These data suggest that pAKT and pGSK-3β production is increased when p85 is located nearest to the phosphoinositols, if a continuous signaling mutated protein is present (Fig. S2).

Fig. S2.

Illustrative diagram depicting the proposed mechanism by which IRS-1 deficiency results in exaggerated pSTAT3 production in response to IL-22. The p85 subunit of PI3K, typically located both within the cytosol and at the cell membrane, is preferentially redistributed to the cellular membrane in the absence of IRS-1. In the presence of mutant KRAS, increased bioavailability of p85 results in increased pAkt, and subsequently pGSK-3β production. pGSK-3β prolongs IL-22RA1 half-life via protective phosphorylation of its cytoplasmic tail. Increased IL-22RA1 half-life results in enhanced pSTAT3 production in response to IL-22 stimulation.

JAK Inhibition Reduces Inflammation and Tumor Burden in Kras/Irs-1^−/− Mice.

Although the increased tumor-associated inflammation and tumor burden identified in Kras/Irs-1^−/− mice was the result of a multistep process, all of these steps culminated in JAK/STAT activation. Therefore, we administered a JAK1 selective antagonist, AZ-289, and a JAK1/2 antagonist, AZD-1480, to both Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice. The mice received AZ-289, AZD-1480, or vehicle control via oral gavage 6 d/wk for 3 wk, starting 8 wk post-AdCre. Analysis of the bronchoalveolar lavage fluid inflammatory cell content for both JAK antagonists revealed that Kras/Irs-1^−/− mice possessed substantially reduced inflammation that was similar in magnitude to that observed in Kras/Irs-1^+/+ mice (Fig. 6 E and F). Both inhibitors also reduced CC and CXC chemokine expression, as expected (Fig. 6 G–J). AZD-1480 and AZ-289 significantly reduced the tumor burden in both cohorts, though the tumor burden was not significantly different for one drug compared with the other (Fig. 6 A–C). Mechanistically, JAK inhibition did not induce apoptosis in any treatment group (Fig. S3). Instead, we identified reductions in K_i-67-positive tumor cells in response to both reagents (Fig. 6 K and L). These results provide evidence that JAK inhibition reduces tumor burden in Kras mutant mice, and is additionally capable of abrogating the enhanced tumor-promoting inflammation afforded by IRS-1 deficiency.

Fig. 6.

JAK inhibition reduces tumor burden and inflammation in Kras/Irs-1^−/− mice. (A) Representative H&E-stained sections from Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice treated with JAK-1/2 inhibitor (AZD-1480), JAK-1 inhibitor (AZ-289), or vehicle control for a total of 3 wk beginning at 8 wk post-AdCre. (B and C) Tumor area percent (percent of lung occupied by tumor) for (B) Kras/Irs-1^+/+ and (C) Kras/Irs-1^−/− drug-treated mice, as above. n = 6 per group. (D) Representative images and (E and F) quantification of Hema-3-stained cytospins of BAL fluid from Kras/Irs-1^+/+ and Kras/Irs-1^−/− drug-treated mice. (G–J) Real-time PCR values for the listed genes from (G) Kras/Irs-1^+/+ and (H) Kras/Irs-1^−/− drug-treated mice. Results are expressed as fold change from vehicle control ± SEM. n = 3 per group. (K) Representative images and (L and M) quantification for K_i-67 stained sections from Kras/Irs-1^+/+ and Kras/Irs-1^−/− drug-treated mice, as above. n = 3–5 per group. All bars ± SEM. *P < 0.05.

Fig. S3.

KRAS mutant, IRS-l^low human L-ADCAs have increased myeloid cell infiltration. Representative H&E images from myeloid and lymphocyte high and low cases. Quantification of total myeloid and lymphoid inflammation between IRS-1^high (n = 37) and IRS-1^low (n = 24) cases. Bars ± SEM. *P < 0.0001.

KRAS mutant, IRS-1^low Human Lung Adenocarcinomas Contain Increased Myeloid Cell Inflammation.

To demonstrate that the proinflammatory phenotype identified in Kras/Irs-1^−/− mice was present in human disease, we reanalyzed the human lung adenocarcinoma TMA for immune cell content. Specifically, we scored all KRAS mutant cases in our TMA for myeloid cell and lymphocyte infiltration (both on a semiquantitative score from 0 to 3, where absent is 0, mild is 1, moderate is 2, and heavy is 3). These scores were also combined to give a total immune cell content score. KRAS mutant, IRS-1^low cancers possessed ∼30% greater content of inflammatory cells, and ∼40% increased myeloid cell infiltration, but no significant difference in lymphoid inflammation was identified, consistent with the findings in the Kras mouse model (Fig. S4).

Fig. S4.

No change in CC3 with JAK inhibition. (A) Representative images for CC3-stained sections from Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice treated with JAK-1/2 inhibitor (AZD-1480), JAK-1 inhibitor (AZ-289), or vehicle control. (B) Quantification of total, myeloid and lymphoid infiltration in IRS-1^High versus IRS-1^Low cases.

Discussion

We began this study of IRS-1 in lung cancer by identifying that IRS-1-deficient lung adenocarcinomas displayed reduced survival, but that this phenotype was restricted to the KRAS mutant L-ADCA subtype. Controlled experiments in mice reproduced the phenotype observed in humans and uncovered a proinflammatory phenotype highlighted by enhanced pSTAT3 production, excessive CC and CXC chemokine expression, and consequential increases in inflammatory cell content. We suspect that most, but not all, of the differences in tumor burden were the result of increased tumor-promoting inflammation, because we were able to diminish tumor burden using a neutrophil-depleting antibody, as has been reported several times previously (16, 17). It appears that the enhanced tumor-associated inflammation in IRS-1-deficient tumors functions to drive tumor cell proliferation, as evidenced by increased Ki67 staining in Kras/Irs-1^−/− tumors. There does not appear to be a role for differential apoptosis here, because apoptosis is rare in the Kras model, independent of IRS-1 content.

Mechanistically, IRS-1-deficient cells display an altered subcellular distribution of the PI3K machinery, favoring a plasma membrane location. Under certain circumstances, such as the presence of mutant KRAS, abundant PI3K within close proximity to its lipid substrates can enhance PI3K pathway activity, as observed here. In this case, increased pGSK-3β production increased the half-life of IL-22RA1, ultimately causing enhanced pSTAT3 production, CC/CXC chemokine expression, and increased protumor inflammatory cell infiltration.

The importance of the Th17 cytokines IL-17 and IL-22 in solid tumor malignancies has been highlighted by several recent publications (18, 23). Though we were unable to identify a role for IL-17A specifically in the context of IRS-1 deficiency, IL-17A-deficient mice were recently shown to display reduced tumor burden and tumor-associated inflammation in the LSL-Kras model used here (19). Notably, IL-22 has recently been shown to play dual roles in cancer (24). Early in the process of tumorigenesis, IL-22 is essential for epithelial cell repair, and actually retards tumor formation. In contrast, in established lesions in proinflammatory colon cancer models, IL-22 drives myeloid cell infiltration and tumor burden via JAK/STAT-mediated amplification of CC/CXC chemokine production, similar to our findings. IL-22 accounts for most, but not all, of the CC/CXC chemokine amplifications identified in Kras/Irs-1^−/− tumors. Taking into consideration the differences between human and mouse, such as the fact that human lung cancer cell lines rarely express CCL-3 (none of our cell lines did) and that we used IL-8 as a surrogate for CXCL-1 and -2, only CXCL-5 was not reproduced using IL-22 stimulation. Given the complexity of the TME, it is highly likely that there are other factors at play in addition to IL-22. Identifying the mechanistic basis for elevated CXCL-5 production by IRS-1-deficient tumors is an area of active investigation in our laboratory.

To our knowledge, this is the first definitive report of a prohost role for IRS-1 in any human cancer type and is paradigm-shifting in nature. Traditionally, IRS-1 is considered a positive effector of growth factor and presumed to promote tumor growth via PI3K activation, which is likely the case in certain malignancies. Several immunohistochemistry (IHC)-based studies have correlated IRS-1 staining with poor outcomes (25). The only prior study of IRS-1 in human lung cancer reported that 46% of NSCLC cases had diminished IRS-1 content by IHC, though outcomes data were not available (26). Notably, we denoted 38% of L-ADCA on the TMA as being IRS-1^low, consistent with the prior study. Taken together, these studies strongly suggest that ∼40% of L-ADCA tumors display reduced IRS-1 content at the protein level. Specifically within the KRAS mutant group, 39% of cases were defined as IRS-1^low, such that ∼10% of all L-ADCA is comprised of this unique subtype. Furthermore, these findings may explain why initial clinical trials using IGF-1R antagonists have not produced positive results (27). IGF-1R inhibition would be expected to reduce the pTyr-IRS-1 cellular content, which would reduce IRS-1:p85 interaction, because only pTyr-IRS-1 binds to p85. Thus, IGF-1R inhibition may generate a relatively deficient IRS-1 state, which, in the presence of mutant KRAS, may produce untoward effects.

Notable limitations of the current study include the fact that the classification of IRS-1^high and IRS-1^low tumors was based on data-driven criteria. Therefore, external validation will be required before further clinical investigation. Additionally, we have not identified the mechanism by which IRS-1 protein content is reduced in human L-ADCA, though we suspect it is related to neutrophilic inflammation; it is unlikely to be the result of a mutation, because IRS-1 mutations were uncommon in the lung adenocarcinoma TCGA dataset (28). We previously identified a significant correlation between neutrophil elastase (NE) staining and reduced IRS-1 staining on serial sections of 38 human L-ADCAs using IHC (29). In that study, we were also able to show that NE entered tumor cells, which resulted in the elimination of intracellular IRS-1. Since the time of that report, NE has been shown to enter multiple different cancer cell types, including breast cancer (30, 31). NE also enters nonmalignant cells, such as fibroblasts (32) and hepatocytes (33), both of which caused the loss of intracellular IRS-1. It is intriguing to consider a positive feedback loop in which tumor-associated neutrophilic inflammation would lead to IRS-1 degradation via NE activity, with the subsequent reduction in intracellular IRS-1 content driving further recruitment of neutrophils, thereby propagating the cycle.

The results reported here highlight a novel concept with respect to pathway signaling in cancer cells. Whereas the focus of aberrant pathway signaling has logically centered upon hyperactive mutant proteins (KRAS, PIK3CA, BRAF, etc.), this report clearly demonstrates that signaling intermediaries can drastically impact cell behavior when their content is reduced. Signaling intermediates are numerous, and mutated versions of many of these proteins are not frequently encountered. However, relative deficiencies of promiscuous signaling proteins, such as IRS-1, can impact pathway output in a manner that is difficult to predict. KRAS mutant, IRS-1^low lung adenocarcinomas represent a unique subtype in which JAK/STAT activation is excessive and drives tumor-promoting inflammation. Because the constellation of events instigated by IRS-1 deficiency culminates in pSTAT3 production, we used JAK inhibitors to show that this tumor subtype is particularly susceptible to this therapeutic strategy. These experiments highlight the role of IRS-1 in lung adenocarcinoma. KRAS mutant cancers contain some degree of inherent tumor-associated inflammation due to the ability of mutant KRAS to induce cytokines, such as IL-8, through NFκB activation (12). Although there is likely to be some contribution of JAK/STAT activity in sculpting the TME, pathway output is enhanced in IRS-1-deficient tumors. Therefore, JAK inhibition and, more specifically, JAK1 inhibition, eliminates the excessive chemokine production and subsequent inflammation afforded by IRS-1 deficiency, resulting in decreased tumor cell proliferation and tumor burden. Because KRAS mutant IRS-1^low lung adenocarcinomas represent 10% of all lung adenocarcinomas, JAK inhibition may represent a viable therapeutic option, especially when considering the intractable nature of KRAS mutant cancers.

Methods

Mouse Studies.

Animal experiments were performed in accordance with approved Institutional Animal Care and Use Committee (IACUC) protocols in a pathogen-free barrier facility at the Fred Hutchinson Cancer Research Center. Mouse models, tissue collection, assessment of mortality, BAL fluid analysis, histology, immunohistochemistry, qPCR analysis of lung tissue, and therapeutic studies are described in detail in SI Methods.

In Vitro Studies.

Human lung cancer cell lines, IRS-1 silencing, microarray, IL-22 treatment, JAK inhibition, Western blotting, and qPCR analysis are described in SI Methods.

Human Tissue Microarray.

An L-ADCA cohort on TMA was obtained from the University of Pittsburgh Cancer Institute. The TMA is annotated for KRAS mutant, EGFR mutant, or WT-EGFR and WT-KRAS status and survival data. Patient identifiers were removed and the study was considered Institutional Review Board exempt. TMA was stained with an IRS-1 antibody (Santa Cruz no.sc-720) and analyzed using an Aperio Genie system. Experimental details described in SI Methods.

Statistical Analysis.

Detailed statistical analyses for human TMA study as well as in vitro and in vivo studies are described in SI Methods.

SI Methods

Mice.

Lox-Stop-Lox-K-ras (LSL-Kras) (34) and Irs-1^fl/fl (35) mice on a pure SVEV background (backcrossed >10 generations) were used in these studies. Mice were housed in a pathogen-free barrier facility at the Fred Hutchinson Cancer Research Center and experiments were performed in accordance with approved IACUC protocols.

Intratracheal Administration of AdCre.

A titer of 4 × 10⁷ pfu AdCre (University of Iowa Viral Vector core) was administered intratracheally to sex-matched anesthetized mice between 8 and 10 wk of age via a 22-g i.v. catheter placed in the trachea.

Assessment of Mortality.

Kras/Irs-1^+/+ (n = 41) and Kras/Irs-1^−/− mice (n = 44) were followed from the date of AdCre administration until death or signs/symptoms of impending mortality. This time-interval was used to calculate overall survival.

BAL Fluid Analysis.

At 8, 12, and 16 wk post-AdCre, mice were killed and BAL fluid was collected by inserting a 22-g i.v. catheter into the trachea. Lungs were lavaged with 0.75-mL saline four times. The BAL fluid was centrifuged at 1,690 × g for 3 min. Red blood cells (RBCs) were lysed using a RBC lysis buffer (BioLegend). Total cell counts were obtained using a hemocytometer. Cytospins were made from each suspension, and manual differentials were generated using Hema-3 staining (Thermo Fisher Scientific).

BAL fluid chemokine content was assessed by subjecting BAL fluid to ELISA for CCL-2, CCL-4, CXCL-1, and CXCL-2.

Histology and Immunohistochemistry.

At 8, 12, and 16 wk post-AdCre, mice were killed and the lungs were harvested. A split lung inflation technique was performed in which the left lung was ligated, removed, and frozen, and the right lung was subsequently inflated with 2× zinc fixative (BD Biosciences no.552658) at 25 cm H₂O pressure via intratracheal catheter for 15 min. Lungs were removed and fixed in 2× zinc fixative for 72 h. Lungs were then paraffin-embedded. Then, 5-µm paraffin-embedded sections were stained for H&E, IRS-1 (Abcam no. 52167), Ly-6G (BioLegend no. 127601), Phopho-Stat3 (pTyr705) (D3A7) (Cell Signaling no. 9145), IL-22 receptor alpha 1 (IL-22RA1; Millipore no. 06-1077), K_i-67 (Abcam no. ab15580), and CC3 (Biocare Medical no. CP229B). Staining was preformed using the avidin–biotin HRP technique with DAB as the chromogenic substrate as previously described (36).

Slides were scanned using Eclipse 80i microscope (Nikon Instruments Inc.). Total lung and tumor area was determined from H&E-stained slides using NIS-Elements Advanced Research software (Nikon). Tumor area percentage was calculated as follows: [(tumor area/ total lung area) × 100]. Phospho-Stat3, IL-22RA1, CC3, and K_i-67-stained lung tissues were scored using a grid at 40× magnification. Results are expressed as number of positively stained cells per high-powered field (hpf) from a total of 5 hpf per slide.

Therapeutic Studies in Mice.

Mice were treated intratracheally with AdCre at 8 wk of age to activate mutant Kras expression and to delete Irs1. After allowing 8 wk post-AdCre for tumors to develop, mice received the JAK1/2 inhibitor, AZD-1480 (30 mg/kg; AstraZeneca; n = 7–8 per group), JAK 1 selective inhibitor, AZ-289 (25 mg/kg; AstraZeneca; n = 3–4 per group), or vehicle control (n = 6 per group) via oral gavage. Mice were treated with drug or vehicle 6 d/wk for 3 wk. In a separate experiment, Kras/Irs-1^+/+ and Kras/Irs-1^−/− mice (n = 3 per group) were treated with either the neutrophil-depleting Ly-6G 1A8 monoclonal antibody or Rat IgG2a 2A3 isotype control (BioXCell) beginning at 6 wk post-AdCre for 4 wk at a dose of 500 µg three times per week.

Cells.

A549 (ATCC no. CCL-185), H460 (ATCC no. HTB-177), H1650 (ATCC no. CRL-5883), and 201T (obtained from University of Pittsburgh Cancer Institute) (37) human lung cancer cell lines were used for in vitro experiments. Cells were maintained in DMEM plus 10% (vol/vol) FCS and 1× penicillin–streptomycin (Gibco).

A549, H460, 201T, and H1650 cells were transfected with control and IRS-1 lentiviral transduction particles (Sigma-Aldrich SHC001V and TRCN0000039910 NM_005544.1-4044s1c1). Cells were plated in 96-well plates (4,000 cells per well) in DMEM containing 10% serum and allowed to adhere overnight. A total of 5 μL of transduction particles were added to each well for 18 h. Plates were washed 2× with PBS, and fresh 10% serum DMEM was added. One day later, puromycin was added to wells, and cells were selected for 10 d.

Quantitative PCR.

Frozen mouse lungs were homogenized in TRIzol (Invitrogen) and total RNA was extracted using chloroform and isopropanol precipitation. Cell lines were plated at 250,000 cells per well in six-well plates in DMEM containing 10% serum and allowed to adhere overnight. Plates were then washed with PBS twice and serum-free DMEM was added. Then, 10 ng of IL-22 (eBioscience no. 14-8229-63) was added to the existing serum-free for 1 h. Cells were washed twice with PBS and frozen in liquid nitrogen, and RNA was extracted using an RNeasy Mini Kit (Qiagen no. 74104).

cDNA was generated from 2 µg total RNA using a SuperScript cDNA synthesis kit (Invitrogen). qPCR was performed using a StepOnePlus Real-Time PCR Machine (Applied Biosystems) and TaqMan primer/probe sets (Applied Biosystems). Reactions were run in triplicate, and GAPDH was used as the endogenous housekeeping gene. Delta-CT values were used for data analysis and expressed as fold change from control.

Microarray.

Three different passages of A549 (ATCC no. CCL-185) cells were treated with three different siRNA mixtures, representing biological replicates. Cells were plated at a concentration of 750,000 cells per well in six-well plates in DMEM (Gibco) plus 10% FCS and 1× penicillin–streptomycin (Gibco) and allowed to adhere overnight. Plates were then washed with PBS twice and siRNA mixtures [20 nM siRNA, either siSCR (Dharmacon ON-TARGETplus Nontargeting Pool D-001810-10-20) or siIRS-1 (Dharmacon IRS1 ON-TARGETplus SMARTpool L-003015-00-0020) plus Lipofectamine 2000 (Thermo Fisher Scientific no. 11668027) in Opti-MEM (Thermo Fisher Scientific no. 31985070) plus serum-free DMEM] were added overnight. RNA was isolated from samples using an RNeasy Kit (Qiagen no. 74104).

RNA integrity was tested using an Agilent 2200 TapeStation (Agilent Technologies, Inc.), quantified using a Trinean DropSense96 spectrophotometer (Caliper Life Sciences), converted to cDNA and biotin-labeled for microarray analysis using Ambien’s Illumina TotalPrep RNA Amplification kit (Life Technologies). Samples were processed on a HumanHT-12 Expression BeadChip (Illumina, Inc.) and imaged on the Illumina iScan system.

IL-22 Treatment.

A549, H460, 201T, and H1650 cells were plated at 250,000 cells per well in six-well plates in DMEM containing 10% serum and allowed to adhere overnight. Plates were then washed with PBS twice, and serum-free DMEM was added. At 18 h later, 25 ng of IL-22 (eBioscience no. 14-8229-63) was added to the existing serum-free media for various time points (10, 15, 30, and 60 min). Cells were washed twice with PBS, frozen in liquid nitrogen, and lysed with 2× cell lysis buffer plus Protease/Phosphatase Inhibitor Mixture (Cell Signaling no. 9803 and no. 5872).

In Vitro JAK Inhibition.

For chemokine experiment, cells were plated at 250,000 cells per well in six-well plates in DMEM containing 10% serum and allowed to adhere overnight. Plates were then washed with PBS twice, and serum-free DMEM was added. A JAK inhibitor (Millipore no. 42009918) diluted in DMSO was added to the serum-free media along with DMSO control for 2 h. After 2 h, 10 ng of IL-22 (eBioscience no. 14-8229-63) was added to the existing serum-free media plus inhibitor and DMSO control for 1 h. Cells were washed twice with PBS and frozen in liquid nitrogen, and RNA was extracted for qPCR analysis.

For protein analysis, cells were plated at 250,000 cells per well in six-well plates in DMEM containing 10% serum and allowed to adhere overnight. Plates were then washed with PBS twice and serum-free DMEM was added. A JAK inhibitor (Millipore no. 42009918) diluted in DMSO was added to the serum-free media along with DMSO control for 2 h at various concentrations. After 2 h, 25 ng of IL-22 (eBioscience no. 14-8229-63) was added to the existing serum-free media plus inhibitor and DMSO control for 15 min. Cells were washed twice with PBS, frozen in liquid nitrogen, and lysed with 2× cell lysis buffer plus Protease/Phosphatase Inhibitor Mixture (Cell Signaling nos. 9803 and 5872).

For viability analysis, cells were plated at 2,000 cells per well in 96-well plates in DMEM containing 10% serum and allowed to adhere overnight. Plates were then washed with PBS twice, and serum-free DMEM was added. A JAK inhibitor (Millipore no. 42009918) diluted in DMSO was added to the serum-free media along with DMSO control for 4 h at listed concentrations. ApoTox-Glo Triplex Assay (Promega no. 9FB104) was then performed based on manufacturer’s instructions.

Immunoblotting.

Protein concentrations were calculated using a BCA Protein Assay Kit (Pierce no. 23225) and separated using NuPAGE 4–12% Bis-Tris gels (Life Technologies) followed by transfer to PVDF membranes. Membranes were blocked for either 2 h at room temperature or overnight at 4C in 5% milk in PBS Tween. Primary antibodies were incubated either overnight at 4 °C or for 2 h at room temperature followed by secondary antibody for 1 h at room temperature. Membranes were incubated in SuperSignal West Pico Chemiluminescent Substrate (Thermo Scientific) for 5 min and then developed using an X-ray film processer. Primary antibodies include IRS-1 (1:250; Cell Signaling no. 2382), pSTAT3 (1:1,000; Cell Signaling no. 9271), STAT3 (1:1,000; Cell Signaling no. 9145), pAKT (ser473) (1:250; Cell Signaling no. 9271), AKT (1:1,000; Cell Signaling no. 9272), pGSK-3β (ser9) (1:1,000; Cell Signaling no. 9336), GSK-3β (27C10) (1:1,000; Cell Signaling no. 9315), PI3K-p85 (1:1,000 Millipore no. 06-497), and IL-22 receptor (1:1,000; Thermo Scientific no. PA5-19987). GAPDH (1:10,000; Cell Signaling no. 2118) and Ponceau stain (Sigma no. P3504) were used as load control.

Human Tissue Microarray.

A lung adenocarcinoma cohort on tissue microarray was obtained from the University of Pittsburgh Cancer Institute (UPCI). The TMA consists of 135 cases. Each case is annotated as KRAS mutant, EGFR mutant, or WT for both EGFR and KRAS. Survival data and known prognostic factors (age, tumor stage, mutational status, etc.) were available through the UPCI cancer registry. Patient identifiers were removed and the study was therefore considered “not human subjects” research (no informed consent required and IRB exempt). The analysis was performed on 124 specimens from 123 unique subjects, because inadequate tissue precluded an assessment of IRS-1 staining on the remaining cases.

FFPE sections were stained with an IRS-1 antibody (Santa Cruz no. sc-720). Digital images of IHC-stained (with IRS-1 antibody) TMA slides were obtained at 20× magnification (0.5 μm per pixel) using a whole-slide scanner (ScanScope AT; Aperio) fitted with a 20×/0.75 Plan Apo objective lens (Olympus). Images were saved in SVS format (Aperio), managed with server software (ImageServer; Aperio), and retrieved with a file management web interface (Spectrum; Aperio).

Under pathologist supervision, the TMA cores were annotated using Aperio’s annotation software (ImageScope v12.2; Aperio). For automated image classification, image areas from TMAs were annotated that represented five user-defined image classes (tumor, stroma, inflammation, other, and clear glass) and ranged in morphologic appearance and staining intensity of DAB and hematoxylin (counterstain). These image areas were used as input parameters for the histologic pattern recognition training software (Genie Training; Aperio) to produce a Genie Training Set. The effectiveness of the Genie Training Set was visualized on the TMA image test regions (TMA spots) using the image classifier algorithm (Genie Classifier; Aperio), which overlaid an image markup pseudo colored for each image class. Annotated image areas from the TMA were adjusted (adding or removing image areas) for each image class to improve the classifier accuracy. For example, if the Genie Classifier algorithm overclassified regions of stroma as tumor, additional stromal annotations were added to the Genie Training algorithm to better represent the stromal image class. This process of adjusting annotations, rerunning the Genie Training algorithm, and visually inspecting pseudo colored markup images output by Genie Classifier was iteratively repeated until a Genie Training Set was developed to classify the TMA slides optimally, as visually validated by a pathologist. The optimized Genie Classifier was then run on the TMAs.

IHC staining was evaluated within tumor areas only of each TMA spot that had been manually annotated, and a separate analysis was performed on areas from each TMA spot that had been classified as tumor by the Genie Classifier. The Color Deconvolution algorithm (Aperio) was used to isolate individual stains for quantification: the red, green, and blue (RGB) OD color vectors were measured for each stain using default software settings for hematoxylin and DAB. The average RGB OD values (hematoxylin: 0.65, 0.704, 0.286; DAB: 0.268, 0.57, 0.776) were used in the Color Deconvolution software to define each stain component in the final analysis settings. Staining was quantified by the following metrics: the percentage of tumor with IRS-1 staining (percentage of strong, medium, and weak positivity) and the score. The H-score was calculated by a simple formula involving the positive percentages: score = 1.0 × (% weak) + 2.0 × (% medium) + 3.0 × (% strong).

H&E-stained sections were obtained from the UPCI and scored for myeloid and lymphocytic cell inflammation. Cellular morphologies under high power were used to distinguish lymphocytes from myeloid cells. The specimens were scored as follows for each: absent, 0; sparse cellular content, 1; moderate cellular content, 2; heavy cellular infiltration, 3.

Statistical Analysis.

For human TMA, overall survival was calculated from the date of diagnosis to the date of death or last confirmation of vital status. Associations between overall survival and clinical and novel predictors were assessed using the log-rank test and by Cox proportional hazards regression. Recursive partitioning, a tree-based survival analysis method, was used to identify an optimal IRS-1 H-score cut point in a univariate survival model (38). Dichotomized IRS-1 and clinical predictors with strong associations with survival in our data or in the literature [sex, age, mutational phenotype (K-ras, EGFR, wild-type), tumor, node, metastasis (TNM) stage, smoking status (never/former/current)] were considered for a multivariable Cox regression model. Smoking status and age were not associated with overall survival in univariate or multivariate models predicting overall survival. Therefore, the primary multivariable model included sex and pathologic stage (as a continuous variable with stages 1A–4 all represented) as factors in addition to the dichotomized IRS-1 score.

In vitro and in vivo studies compared average values between groups using Student’s t test or one-way ANOVA with Tukey test adjustments for multiple comparisons. Data displays and statistical analyses were conducted using Prism Software, SAS/STAT software, version 9.4 (SAS Institute, Inc.) and R version 3.1.3 (R Foundation for Statistical Computing) including the rpart package. Tests were two-sided, with P < 0.05 considered statistically significant. For the A549 microarray, the Bioconductor package lumi (39) and genefilter were used for data assessment and filtering. Limma (40) was used to determine differential gene expression where significant values were defined as |log₂ (ratio) | ≥ 0.585 (±1.5-fold) with the false detection rate set to 5%.