The Underlying Tumor Genomics of Predominant Histologic Subtypes in Lung Adenocarcinoma

Abstract

Introduction:

The purpose of the study is to genomically characterize the biology and related therapeutic opportunities of prognostically important predominant histological subtypes in lung adenocarcinoma (LUAD).

Methods:

We identified 604 patients with stage I-III LUAD who underwent complete resection and targeted next-generation sequencing using the MSK-IMPACT platform. Tumors were classified according to predominant histologic subtype and grouped by architectural grade (lepidic [LEP], acinar or papillary [ACI/PAP], and micropapillary or solid [MIP/SOL]). Associations between clinicopathologic factors, genomic features, mutational signatures, and recurrence were examined within subtypes and, when appropriate, quantified using competing-risks regression, with adjustment for pathologic stage and extent of resection.

Results:

MIP/SOL tumors had higher tumor mutational burden (p<0.001), fraction of genome altered (p=0.001), copy number amplifications (p=0.021), rate of whole-genome doubling (p=0.008), and number of oncogenic pathways altered (p<0.001), compared with LEP and ACI/PAP tumors. Across all tumors, mutational signatures attributed to APOBEC activity were associated with the highest risk of postresection recurrence: SBS2 (p=0.021) and SBS13 (p=0.005). Three oncogenic pathways (p53, Wnt, Myc) were altered with statistical significance in MIP/SOL tumors. Compared with LEP and ACI/PAP tumors, MIP/SOL tumors had a higher frequency of targetable BRAF-V600E mutations (p=0.046). Among ACI/PAP tumors, alterations in the cell cycle (p<0.001) and PI3K (p=0.002) pathways were associated with recurrence; among MIP/SOL tumors, only PI3K alterations were (p=0.049).

Conclusions:

These results provide the first in-depth assessment of tumor genomic profiling of predominant LUAD histologic subtypes, their associations with recurrence, and their correlation with targetable driver alterations in patients with surgically resected LUAD.

Introduction

To address the robust intratumoral heterogeneity of lung adenocarcinoma (LUAD), the International Association for the Study of Lung Cancer, American Thoracic Society, and European Respiratory Society proposed a histologic-subtype classification system for LUAD in 2011.¹ This classification system uses comprehensive histologic subtyping to semi-quantitatively assess histologic patterns to define a single predominant pattern for invasive LUAD.¹ Multiple subsequent studies have shown this classification system is associated with clinicopathologic characteristics and patient prognosis.^{2, 3}

Parallel to LUAD histologic-subtype classification, broad-panel next-generation sequencing (NGS) has increasingly been used to elucidate tumor biologies, identify driver-gene perturbations amenable to targeted therapies, examine chromosomal instability (CIN), and inform prognoses for patients with non-small cell lung cancer (NSCLC).⁴ In addition, it is well established that cancer biology and its resulting clinical phenotypes are driven by different pathways, many of which have upstream oncogenic alterations that result in further downstream alterations and phenotypes in the same pathway.^{5, 6}

Despite the known strong association between LUAD predominant histologic subtypes and prognosis, there remains a surprising paucity of data on genomic alterations associated with each predominant histologic subtype, as well as their potential contributions to clinical outcomes. To address this knowledge gap, we used broad-panel NGS to investigate tumor genomic features, oncogenic pathway alterations, CIN, mutational signatures, and targetable driver gene alterations and their association with predominant LUAD histologic subtypes and tumor recurrence in surgically resected LUAD.

Methods

Patient Cohort

This study was approved by the institutional review board at Memorial Sloan Kettering Cancer Center. We identified patients who underwent complete surgical resection for LUAD and had targeted NGS (Memorial Sloan Kettering–Integrated Mutation Profiling of Actionable Cancer Targets [MSK-IMPACT]⁷) performed on their primary tumor between February 2010 and December 2018. Exclusion criteria included mucinous LUAD, receipt of induction therapy, microscopic or macroscopic residual disease, pathologic stage IV disease, and low-quality NGS (Supplementary Figure 1).

Clinical characteristics, preoperative computed tomographic (CT) and positron emission tomographic (PET) images, and pathologic reports (8^th edition AJCC Cancer Staging Manual) were reviewed. Follow-up was performed in accordance with National Comprehensive Cancer Network guidelines.⁸ Recurrences were distinguished from metachronous tumors using Martini and Melamed criteria, with confirmation from pathologic and genomic relatedness when available.⁹

Histologic Evaluation

Tumors were classified as lepidic (LEP), acinar (ACI), papillary (PAP), micropapillary (MIP), or solid (SOL).¹ The predominant histologic subtype was the pattern with the highest percentage, as reported by Travis et al.¹⁰ Tumors were grouped according to architectural grade: low (LEP), intermediate (ACI/PAP), and high (MIP/SOL). This grading system was previously shown to be prognostic for disease-free survival and recurrence.^11-14

Tumor Genomic Analysis

MSK-IMPACT NGS was performed and analyzed as previously described.^{7, 15} The sequencing breadth of the IMPACT panel has increased over time, resulting in 8, 190, and 406 patients sequenced with 341-, 410-, and 468-gene panels, respectively. Tumor mutational burden (TMB) was defined as the fraction of nonsynonymous single-nucleotide or insertion or deletion mutations divided by the length of the coding region (in megabases [Mb]) sequenced by each panel (0.98, 1.06, and 1.22 Mb in the 341-, 410-, and 468-gene panels, respectively). Fraction of genome altered (FGA) was defined as the fraction of log2 copy number variation (gain or loss) >0.2 divided by the size of the genome whose copy number was profiled. FGA was corrected for tumor purity, ploidy, and clonal heterogeneity using the FACETS tool and was used as a surrogate for CIN.¹⁶

Copy-number alteration frequency plots were generated using the Integrative Genomics Viewer (Broad Institute).¹⁷ Statistically significant focal copy number alterations were identified from segmented data using GISTIC 2.0.¹⁸ Copy number deletions, amplifications, and whole-genome doubling (WGD) and arm-level copy number estimates were calculated from the FACETS output.¹⁹ Mutational signatures were computed for tumors with a mutation burden of ≥13.8 mutations/megabase²⁰ (268 patients [44%]: LEP, 29; ACI/PAP, 167; MIP/SOL, 72).

Ten canonical oncogenic signaling pathways (cell cycle, Hippo, Myc, Notch, Nrf2, PI3K, RTK/RAS, TGFβ, p53, and Wnt) were investigated by predominant histologic subtype, as previously described.^{15, 21} Analysis of specific somatic alterations was performed using OncoKB to remove variants of unknown significance.²² Therapeutic-actionability information was annotated using OncoKB, and each genomic alteration was stratified into 1 of 4 levels of clinical actionability.²² All of the genomic and clinical data used in our analyses are publicly available through the cBioPortal for Cancer Genomics.²³ More information is included in the Supplementary Methods.

Statistical Analysis

The distributions of clinical factors were summarized as frequency (percentage) or median (interquartile range [IQR]) and compared between histologic subtypes using Fisher’s exact test for categorical variables or the Kruskal-Wallis test for continuous variables. Genomic variables were similarly compared across histologic subtypes, with the application of false discovery rate to account for multiple testing. Spearman’s correlation was used to measure the association between TMB and FGA by histologic subtype.

Median follow-up was estimated using the reverse Kaplan-Meier method. The primary endpoint of recurrence was analyzed using a competing-risks approach. Cumulative incidence of recurrence (CIR) was defined from the time of surgery to recurrence, with death without recurrence treated as a competing-risks event. Within each histologic subtype, the relationship between variables and recurrence was quantified using Fine and Gray’s competing-risks regression.

Analyses were conducted using Stata 15.0 (StataCorp, College Station, TX) and R 3.5.1 (R Core Team, Vienna, Austria). Statistical tests were 2-sided, and p<0.05 was considered statistically significant.

Results

Clinicopathologic Characteristics

In total, 604 patients were included in the study (median age at resection, 68 years [IQR, 62-74]), and 402 (67%) were women (Table 1). Most patients (77%) had a history of smoking, with a median of 20 pack-years (IQR, 1-40). At diagnosis, 40 patients (7%) were clinically nodepositive on the basis of CT and PET criteria. Most patients underwent a minimally invasive approach (n=521 [86%]), and 83 (14%) had an open resection. A majority of patients underwent lobectomy (n=390 [65%]). Postoperatively, 97 patients (16%) had pathologic nodal metastasis. Predominant invasive LUAD histologic patterns were as follows: LEP, 88 (15%); ACI, 368 (61%); PAP, 43 (7%); MIP, 37 (6%); SOL, 68 (11%). Pathologic stages were as follows: I, 447 (74%); II, 95 (16%); III, 62 (10%). Postoperatively, 116 patients (19%) received adjuvant therapy. Adjuvant therapy modalities included chemotherapy/immunotherapy (96/116 [83%]), chemoradiation (18/116 [15%]), and radiation (2/116 [2%]). Median follow-up was 2.51 years (IQR, 2.08-3.14).

Table 1.

Clinicopathologic characteristics by lung adenocarcinoma histologic subtype

Characteristic	Total (N=604)	LEP (N=88 [15%])	ACI/PAP (N=411 [68%])	MIP/SOL (N=105 [17%])	p
Age at resection, years	68 (62-74)	69 (64-73)	69 (61-74)	68 (63-73)	0.803
Sex
Male	202 (33)	18 (20)	143 (35)	41 (39)	0.012
Female	402 (67)	70 (80)	268 (65)	64 (61)
Smoking status
Never	138(23)	20 (23)	102(25)	16 (15)	0.11
Ever	466 (77)	68 (77)	309 (75)	89 (85)
Pack-years (N=603)	20(1-40)	14 (0.8-30)	17 (0.1-38)	30 (13-48)	<0.001
FEV1 (N=592)	94 (83-106)	96 (83-106)	94 (84-107)	90 (75-106)	0.086
DLCO (N=586)	84 (69-97)	90 (80-99)	85 (70-98)	74 (62-89)	<0.001
Tumor size on CT, cm (N=600)	2 (1.4-2.9)	1.8 (1.3-3)	2 (1.4-2.7)	2.2 (1.5-3.3)	0.117
SUVmax (N=537)	3.7 (1.9-7.2)	1.8 (1.3-2.8)	3.7 (1.9-6.4)	7.3 (4.2-10.9)	<0.001
cN status (N=602)
Node-negative	562 (93)	85 (98)	388 (95)	89 (85)	<0.001
Node-positive	40 (7)	2 (2)	22 (5)	16 (15)
Pathologic tumor size, cm	1.8 (1.2-2.8)	1.4 (0.9-2)	1.8 (1.3-2.7)	2.2 (1.6-3.5)	<0.001
VPI	109 (18)	4 (5)	71 (17)	34 (32)	<0.001
LVI (N=599)	248 (41)	9 (10)	166 (41)	73 (70)	<0.001
Tumor STAS (N=544)	337 (62)	23 (29)	240 (64)	74 (83)	<0.001
pN status (N=603)
Node-negative	506 (84)	84 (95)	346 (84)	76 (72)	<0.001
Node-positive	97 (16)	4 (5)	64 (16)	29 (28)
pStage
I	447 (74)	78 (89)	311 (76)	58 (55)
II	95 (16)	6 (7)	63 (15)	26 (25)	<0.001
III	62 (10)	4 (5)	37 (9)	21 (20)

Data are no. (%) or median (interquartile range). ACI, acinar; CT, computed tomography; DLCO, diffusing capacity of the lungs for carbon monoxide; FEV1, forced expiratory volume in 1 second; LEP, lepidic; LVI, lymphovascular invasion; MIP, micropapillary; PAP, papillary; SOL, solid; STAS, spread through air spaces; SUVmax, maximum standardized uptake value; VPI, visceral pleural invasion.

Invasive LUAD histologic subtypes were grouped according to architectural grade (LEP, 88 [15%]; ACI/PAP, 411 [68%]; MIP/SOL, 105 [17%]). Associations between histologic subtypes and clinicopathologic characteristics are shown in Table 1.

Association Between Subtype and Genomic Features

We first evaluated the frequencies of genes altered in ≥2% of the study cohort (Figure 1A). Three genes were significantly altered in LEP tumors, compared with ACI/PAP and MIP/SOL tumors: EGFR (42% vs. 31% vs. 19%; p=0.001), RBM10 (26% vs. 12% vs. 10%; p=0.001), and TERT (14% vs. 5% vs. 10%; p=0.012) (Figure 1B). Four genes were significantly altered in MIP/SOL tumors, compared with ACI/PAP and LEP tumors: TP53 (62% vs. 33% vs. 16%; p<0.001), SETD2 (8% vs. 3% vs. 1.1%; p=0.048), MGA (11% vs. 2% vs. 2.3%; p<0.001), and SMARCA4 (6% vs. 1.5% vs. 2.3%; p=0.035) (Figure 1B).

Figure 1. Association of histologic subtypes of invasive lung adenocarcinoma and genomic features.

(A) Oncoprint of alteration frequencies of genes altered in ≥2% of the study cohort. Gene alteration data are presented by increasing pathologic stage. (B) Frequency plot of significantly altered genes according to histologic subtype. (C) Box plot of tumor mutational burden (TMB) vs. histologic subtypes. (D) Box plot of fraction of genome altered (FGA) vs. histologic subtypes. (E) Scatterplot of TMB vs. FGA by histologic subtype (ρ=Spearman’s rho correlation coefficient). The distribution of TMB values by histologic subtype is displayed on the X-axis, and the distribution of FGA values by histologic subtype is displayed on the Y-axis. ACI, acinar; LEP, lepidic; MIP, micropapillary; PAP, papillary; SOL, solid.

Median TMB increased with subtype invasiveness: LEP, 3.9 (IQR, 2-6.9); ACI/PAP, 4.9 (IQR, 2.6-7.9); MIP/SOL, 7.9 (IQR, 3.9-14.5) (p<0.001) (Figure 1C). Median FGA also increased with subtype invasiveness: LEP, 0.174 (IQR, 0.069-0.328); ACI/PAP, 0.222 (IQR, 0.095-0.406); MIP/SOL, 0.304 (IQR, 0.119-0.505) (p=0.001) (Figure 1D). Finally, the correlation between TMB and FGA became stronger with subtype invasiveness: LEP, correlation coefficient rho (ρ)=−0.03 (p=0.770); ACI/PAP, ρ=0.11 (p=0.029); MIP/SOL, ρ=0.28 (p=0.004) (Figure 1E).

Association Between Subtype and CIN

Investigation of the CIN landscape revealed increasing copy number alterations with histologic subtype invasiveness, despite an overall similar pattern on affected chromosomes (Figure 2A). Tumor purity did not differ between histologic subtypes (p=0.189; Supplementary Figure 2A). Statistically significant differences in copy number alterations across subtypes at the following chromosome arm levels were identified: 2p (p=0.030), 2q (p=0.028), 4q (p=0.030), 5p (p=0.030), 6p (p=0.013), 6q (p=0.013), 7p (p=0.023), 7q (p=0.002), 11p (p=0.008), 11q (p=0.008), 12p (p=0.047), 17p (p=0.008), and 17q (p<0.001). Chromosome arms altered with statistical significance were annotated with copy number–amplified oncogenes according to OncoKB (Supplementary Figure 2B).²² Across subtypes, MIP/SOL tumors had the highest number of copy number amplifications (p=0.021) (Figure 2B), whereas copy number deletions did not differ between subtypes (p=0.208) (Figure 2C). Copy number–altered genes present in ≥1% were studied by subtype (Supplementary Figure 2C). Importantly, and consistent with their aggressive tumor biology, MIP/SOL tumors, compared with ACI/PAP and LEP tumors, had the highest rate of WGD (18% vs. 10% vs. 4.5%; p=0.008) (Figure 2D).

Figure 2. Analysis of copy number alterations by histologic subtype of invasive lung adenocarcinoma.

(A) Copy number heatmap with amplifications (red) and deletions (blue) by histologic subtype, arranged by decreasing FGA. (B) Box plot of FGA by amplifications vs. histologic subtypes. (C) Box plot of FGA by deletions vs. histologic subtypes. (D) Bar plot of percentage of whole-genome doubling vs. histologic subtypes. ACI, acinar; FGA, fraction of genome altered; LEP, lepidic; MIP, micropapillary; PAP, papillary; SOL, solid; WGD, whole-genome doubling.

Somatic Mutational Signature Analysis

Somatic mutational signatures were investigated by histologic subtype (Figure 3A). Median somatic mutations increased with histologic grade: LEP, 26 (IQR, 21.5-32.5); ACI/PAP, 30 (IQR, 21-44); MIP/SOL, 37 (IQR, 24-59) (p=0.007). We evaluated detectable signatures across all tumors (Figure 3B) and by histologic subtype (Figure 3C). LEP tumors exhibited a higher frequency of SBS3 (a signature attributed to defective DNA repair) than ACI/PAP or MIP/SOL tumors (Figure 3C). The frequency of SBS2 and SBS13, signatures attributed to activity of the APOBEC family of cytidine deaminases (APOBEC3A and APOBEC3B),^{24, 25} increased with subtype invasiveness (Figure 3C).

Figure 3. Analysis of somatic mutational signatures.

(A) Bar plots of detectable mutational signatures (colored) by individual tumors. Tracks above the bar charts indicate (i) somatic mutations, (ii) histologic subtype, (iii) recurrence status, (iv) pathologic stage, (v) tumor spread through air spaces (STAS), (vi) and smoking history. (B) Distribution of select detectable signatures across all tumors and their representative single-base substitution (SBS) mutational profiles and proposed etiology. (C) Select detectable mutational signatures across histologic subtypes. (D) Cumulative incidence of postresection recurrence curves according to SBS2 and SBS13 mutational signature status across all patients. ACI, acinar; HR, homologous recombination; LEP, lepidic; MIP, micropapillary; NA, not available; PAP, papillary; SOL, solid.

Investigation of mutational signatures associated with postresection recurrence (Supplementary Figure 3A) revealed that, across all tumors, SBS2 (subhazard ratio [SHR], 2.07; 95% confidence interval [CI], 1.11-3.84; p=0.021) and SBS13 (SHR, 2.27; 95% CI, 1.29-4.02; p=0.005) were associated with increased risk of tumor recurrence (Figure 3D). Among ACI/PAP tumors, SBS2 (SHR, 2.56; 95% CI, 1.16-5.64; p=0.02) and SBS13 (SHR, 2.85; 95% CI, 1.34-6.07; p=0.007) were again associated with increased risk of tumor recurrence (Supplementary Figure 3B); no signatures were significantly associated with recurrence among MIP/SOL tumors.

Association Between Subtypes and Oncogenic Pathways

We evaluated the alteration frequencies of 10 oncogenic signaling pathways by subtype (Figure 4A). Alteration frequencies of the 121 genes used for pathway analysis are reported in Supplementary Table 1. The RTK/RAS pathway was significantly altered in LEP tumors, compared with ACI/PAP and MIP/SOL tumors (92% vs. 91% vs. 81%; p=0.010) (Figure 4B). Three oncogenic pathways were significantly altered in MIP/SOL tumors, compared with ACI/PAP and LEP tumors: p53 (67% vs. 43% vs. 28%; p<0.001), Wnt (7.6% vs. 5% vs. 0%; p=0.042), and Myc (13% vs. 5.6% vs. 9.1%; p=0.022) (Figure 4B). Mean number of pathways altered (NPA) increased with subtype invasiveness: LEP, 1.65; ACI/PAP, 2.00; MIP/SOL, 2.34 (p<0.001). Tumors with 1 NPA (n=184) were most frequently LEP (n=34 [39%]); tumors with ≥4 NPA (n=56) were most frequently MIP/SOL (n=17 [16%]) (p<0.001) (Figure 4C).

Figure 4. Association of histologic subtypes of invasive lung adenocarcinoma and oncogenic pathways.

(A) Oncoprint of altered oncogenic pathways by histologic subtype. Pathway alteration data presented by increasing number of pathways altered (NPA). (B) Frequency plot of oncogenic pathways according to histologic subtype. (C) Frequency of NPA vs. histologic subtypes (darker tones indicate higher alteration frequencies). (D) Co-occurrence and mutual exclusivity between oncogenic pathways across all tumors and by histologic subtype. ACI, acinar; LEP, lepidic; MIP, micropapillary; PAP, papillary; SOL, solid.

Next, we investigated the mutual exclusivity and co-occurrences of the 10 oncogenic pathways (Figure 4D). Across the entire cohort, 4 pairs co-occurred with statistical significance: p53–cell cycle (odds ratio [OR], 4.47; p<0.001), PI3K-Nrf2 (OR, 6.48; p<0.001), p53-Notch (OR, 4.07; p=0.01), and p53-Hippo (OR, 6.93; p=0.03). Co-occurrences by histologic subtype include the following: LEP, p53–cell cycle (OR, 7.50; p=0.03); ACI/PAP, p53–cell cycle (OR, 3.56; p<0.001) and PI3K-Nrf2 (OR, 7.44, p<0.001); MIP/SOL, p53–cell cycle (OR, 16.35; p=0.02) (Figure 4D). Pathway co-occurrences at the gene level are presented in Supplementary Figure 4.

Cumulative Incidence of Recurrence Analysis

Across all tumors, the 3-year CIR for any recurrence was 23% (95% CI, 19%-28%). The 3-year CIR for any recurrence was 8% (95% CI, 3%-19%) in the LEP cohort, 23% (95% CI, 19%-29%) in the ACI/PAP cohort, and 37% (95% CI, 27%-51%) in the MIP/SOL cohort (p<0.001) (Supplementary Figure 5A), consistent with previous observations.¹³ On univariable analysis, a minimally invasive approach was associated with improved CIR, compared with an open approach, across all patients (SHR, 0.33; 95% CI, 0.22-0.48; p<0.001). To identify clinicopathologic and/or oncogenic pathways associated with recurrence after complete resection, competing-risks regression analyses were performed within each histologic subtype. Multivariable analyses, adjusted for extent of resection and pathologic stage, were performed, excluding LEP tumors (Supplementary Table 2), given the lack of recurrence events. Among ACI/PAP tumors (Supplementary Table 3), preoperative primary tumor maximum standardized uptake value (SHR, 1.11; 95% CI, 1.08-1.15; p<0.001), visceral pleural invasion (SHR, 1.88; 95% CI, 1.12-3.14; p=0.017), and pathway alterations in cell cycle (SHR, 2.74; 95% CI, 1.73-4.34; p<0.001; Supplementary Figure 5B) and PI3K (SHR, 2.11; 95% CI, 1.32-3.37; p=0.002; Supplementary Figure 5C) were independently associated with increased risk of recurrence. Among MIP/SOL tumors (Supplementary Table 4), alterations in the PI3K pathway were associated with increased risk of recurrence (SHR, 1.97; 95% CI, 1.00-3.88; p=0.049; Supplementary Figure 5D).

Therapeutic Actionabilities

Using OncoKB, we identified 735 actionable alterations across 30 genes (Figure 5A). The RTK/RAS pathway harbored the most actionable alterations (n=536 [73%]), of which 37% (200/536) had level I evidence. Total actionable alterations according to histologic subtype included 110 LEP (15%), 515 ACI/PAP (70%), and 110 MIP/SOL (15%) (Figure 5A). LEP tumors had a higher frequency of level I actionable targets, compared with ACI/PAP or MIP/SOL tumors (35% vs. 27% vs. 21%) (Figure 5B). We compared the TMB of MIP/SOL tumors with level 1 actionability to those with other levels of evidence (levels 2, 3A, 3B, 4, oncogenic, or variant of unknown significance) and identified a statistically significantly lower TMB among MIP/SOL tumors with actionable level 1 alterations (median, 3.25 [IQR 2-7] vs. 10.5 [IQR 6.1-15.8]; p<0.001) (Supplementary Figure 6A). MIP/SOL tumors had a higher frequency of actionable amplifications than LEP or ACI/PAP tumors (13% vs. 9% vs. 12%) (Supplementary Figure 6B). Among tumors with ≥3 NPA, MIP/SOL tumors had a higher frequency of actionable targets than LEP or ACI/PAP tumors (20% vs. 14% vs. 18%) (Supplementary Figure 6C).

Figure 5. Actionable alterations by histologic subtype of invasive lung adenocarcinoma.

(A) Frequency (%) of patients with an actionable alteration according to histologic subtype. The color code represents the highest level of evidence within the specific histologic subtype (note: among LEP tumors, alterations in BRAF-V600K, BRAF-K601E, and BRAF-K601N were level 3B, whereas among ACI/PAP and MIP/SOL, alterations in BRAF-V600E were level 1). (B) Levels of evidence by histologic subtype. (C) Targetable lung adenocarcinoma alterations by histologic subtype. ACI, acinar; LEP, lepidic; MIP, micropapillary; PAP, papillary; SOL, solid; VUS, variant of unknown significance.

Finally, we investigated known LUAD targetable alterations with existing therapies by histologic subtype (Figure 5C). MIP/SOL tumors had the lowest frequency of targetable LUAD alterations, compared with ACI/PAP and LEP tumors (27% vs. 36% vs. 41%). Targetable EGFR alterations (exon 19 deletion or L858R mutation) were more frequently identified in LEP tumors than ACI/PAP or MIP/SOL tumors (23% vs. 14% vs. 5%; p=0.001). In contrast, targetable BRAF-V600E mutations were more frequently identified in MIP/SOL than ACI/PAP or LEP tumors (5% vs. 1.2% vs. 1.1%; p=0.046). Interestingly, there were no differences in KRAS-G12C mutations across subtypes.

Discussion

This is the first study to use broad-panel NGS to characterize the predominant histologic subtypes of invasive LUAD and investigate the association between mutational load, CIN, oncogenic pathways, mutational signatures, and tumor recurrence within subtypes. TMB, FGA, copy number amplifications, WGD, and NPA were associated with increasing subtype invasiveness. Among the mutational signatures evaluated, SBS2 and SBS13 were associated with the highest risk of recurrence in all subtypes. LEP tumors were most likely to have a targetable EGFR alteration, whereas MIP/SOL tumors were more likely to have a targetable BRAF alteration. Finally, we observed that alterations in the cell cycle and PI3K pathways for ACI/PAP tumors and in the PI3K pathway for MIP/SOL tumors were independently associated with tumor recurrence.

We identified a strong association between clinicopathologic characteristics, aggressive genomic profile, architectural grade, and tumor recurrence. Increasing smoking pack-years, which results in the accumulation of somatic mutations and increased CIN,^{26, 27} was associated with subtype invasiveness. This is evidenced by statistically significantly higher summary metrics, such as TMB and FGA, among MIP/SOL tumors. Among patients with NSCLC treated with immunotherapy, higher TMB has been associated with improved benefit.²⁸ In contrast, higher FGA has been associated with reduced response to immunotherapy in patients with melanoma,^{29, 30} and the highest FGA was associated with the least benefit from immunotherapy in patients with NSCLC.²⁸ A recent study identified a statistically significant relationship between SOL-predominant histologic subtype and increased expression of programmed death-ligand 1, as further evidence of the immunoresistant microenvironment of these tumors.³¹ Moreover, consistent with their more-aggressive profile, MIP/SOL tumors were statistically significantly associated with pathologic markers of invasiveness, such as LVI, VPI, and tumor STAS. Our group has previously identified tumor STAS as an independent predictor of recurrence among patients with LUAD undergoing complete resection.³²

CIN is associated with the development of metastases and decreased recurrence-free survival in LUAD in general.³³ However, we have shown for the first time that CIN is significantly higher in MIP/SOL-predominant tumors than in other LUAD histologic subtypes, as demonstrated by higher FGA, higher fraction of somatic copy number amplifications, and higher rate of WGD. Interestingly, higher copy number amplification has been associated with increased rates of WGD in LUAD³⁴ and late-stage NSCLC.³⁵ Across all tumors, we observed amplification of MDM2, TERT, NKX2-1, EGFR, MYC, ERBB2, and MET, which has been previously described.³⁶ WGD is an early event in LUAD evolution and is associated with increased subclonal mutations and subclonal copy number alterations.^{33, 37} Accumulating evidence suggests a preceding oncogenic driver mutation (e.g., TP53) establishes a permissive environment for the proliferation of cancer cells, which subsequently undergo genome doubling after errors in cell division.^{38, 39} WGD is associated with poor prognosis across multiple tumors,³⁸ providing further evidence of the aggressive biology of MIP/SOL tumors.

Somatic mutations in cancer genomes result from both exogenous and endogenous mutational processes that occur before and during tumorigenesis.⁴⁰ Whereas mutational-signature analysis classically uses whole-genome or whole-exome sequencing, we have previously shown the feasibility of performing somatic-mutational-signature analysis using MSK-IMPACT.²⁰ In this study, we observed an association between mutational signatures attributed to APOBEC activity enzymes (SBS2 and SBS13) and increasing subtype invasiveness. APOBEC enzymes are involved in fueling tumor diversity, subclonal evolution, and therapeutic resistance.⁴¹ We found SBS2 and SBS13 to be associated with the highest risk of recurrence across all histologic subtypes. Interestingly, combined high TMB and APOBEC mutational signature was recently reported to predict response to immunotherapy in an NSCLC cohort.⁴²

Analysis of alterations of oncogenic signaling pathways revealed a higher NPA among MIP/SOL tumors, with 3 cell growth and proliferation pathways⁷ (p53, Wnt, and Myc) significantly altered. The RTK/RAS pathway was the most and least altered in LEP and MIP/SOL tumors, respectively. Certain targetable LUAD driver mutations in the RTK/RAS pathway, such as EGFR, MET, and BRAF, are almost exclusively clonal and occur early in tumor evolution.³³ These observations help explain the robust initial responses observed across multiple sites of disease following targeted therapy for these alterations.^43-45 The RTK/RAS pathway was least altered in MIP/SOL tumors, suggesting that a higher frequency of subclonal events occur during tumor evolution, which is exemplified by the higher frequency of alterations in known LUAD subclonal genes, such as NF1, ATM, SMARCA4, and SMAD4. Fewer clonal mutations in MIP/SOL tumors may be the result of higher CIN and WGD, which are thought to limit the efficacy of therapeutic strategies aimed at clonal alterations.³³

Although ACI/PAP-predominant tumors account for 70% of LUADs and are associated with an intermediate risk of postresection recurrence,^{2, 3} studies investigating these tumors are lacking. In ACI/PAP tumors, alterations in the cell cycle and PI3K oncogenic pathways and SBS2 and SBS13 mutational signatures were independently associated with recurrence. These observations support consideration of future studies examining increased surveillance and/or adjuvant therapy following resection for patients with ACI/PAP tumors with these alterations. Three-year CIR was 37% among patients with MIP/SOL-predominant tumors, and PI3K pathway alterations were also independently associated with recurrence. Alterations in the PI3K pathway, and its gene-transcription targets, have recently been associated with poor survival in pan-cancer analyses.⁴⁶

Limitations of our study include that NGS was performed using single-region sampling of the primary tumor. As previously noted, intratumoral heterogeneity is intrinsic to LUAD,³³ and single-region sampling may not accurately capture the complexity of the disease, such as its clonal architecture.⁴⁷ Our study focused on the analysis of nonmucinous invasive LUAD, as this group represents more than 70%-90% of surgically resected tumors, whereas mucinous LUAD represents approximately 2%-10%.¹ Mucinous LUAD tumors are associated with a more aggressive clinical presentation and poor prognosis and are characterized by a distinct molecular profile.^48-50 We used the predominant histologic subtype of invasive non-mucinous LUAD for our analyses, yet minor components of other subtypes may also play a role in prognosis.⁵¹ However, current pathologic standards are to report the predominant histologic subtype, with other subtype percentages noted.¹ Future studies will investigate the utility of high-throughput single-cell sequencing to obtain a more comprehensive understanding of the genetic and epigenetic profiles of cancer cells within each histologic pattern.⁵² Finally, in our mutational-signature analysis, we did not examine doublet-base substitutions or insertions or deletions, which have been reported to contribute to the overall mutational burden.²⁴

Our findings provide new insights into the tumor genomic underpinnings associated with the aggressive biology of and the therapeutic challenges associated with MIP/SOL tumors. We show that MIP/SOL tumors have a higher TMB, increased CIN, more APOBEC mutational signatures, more oncogenic pathway alterations, and the lowest frequency of targetable LUAD alterations, compared with other LUAD histologic subtypes. Collectively, these studies offer an important first step in unravelling the previously unknown biologic reasons of how tumor morphologic appearance is linked to clinical outcomes in LUAD.

Abbreviations

ACI

acinar

CIN

chromosomal instability

CIR

cumulative incidence of recurrence

CT

computed tomography

DLCO

diffusing capacity of the lungs for carbon monoxide

FEV1

forced expiratory volume in 1 second

FGA

fraction of genome altered

HR

homologous recombination

IQR

interquartile range

LEP

lepidic

LUAD

lung adenocarcinoma

PAP

papillary

PET

positron emission tomography

MIP

micropapillary

MSK-IMPACT

Memorial Sloan Kettering–Integrated Mutation Profiling of Actionable Cancer Targets

NGS

next-generation sequencing

NPA

number of pathways altered

NSCLC

non-small cell lung cancer

SBS

single-base substitution

SHR

subhazard ratio

SOL

solid

STAS

spread through air spaces

SUVmax

maximum standardized uptake value

TMB

tumor mutational burden

VPI

visceral pleural invasion

VUS

variant of unknown significance

WGD

whole-genome doubling

WT

wild-type