Genomics of lung cancer

Multi-platform investigation of non-small cell lung cancer has the capability of discovering clinically important biological pathways involved in cancer development, progression, prognosis and response to treatment in patients with lung cancer. These two complementary publications from this productive research group describe genomic analyses of the two commonest subtypes of lung cancer: lung adenocarcinoma (LUAD) and lung squamous cell carcinoma (LUSC). The investigators analyse primary whole exome sequencing (WES) data from resected early stage lung cancers, supplemented by combinatorial comparative analyses of available landmark data: The Cancer Genome Atlas (TCGA) (1,2).

In addition to describing the types and frequency of mutations, they provide a subset analysis of clinical relevance to survival and response to adjuvant therapy, as well as correlation with a panel of selected immune response biomarkers. The latter analyses are of topical interest given emerging evidence of clinical benefit from immunotherapies designed to address the immune checkpoints that tumours hijack to suppress host anti-cancer immunity.

The results are summarised in Table 1, which also provides a narrative comparison of the two lung cancer subtypes.

Table 1. Summary of major findings in LUAD (Kadara et al. 2016) and LUSC (Choi et al. 2016).

Clinico-biological correlates	LUAD	LUSC
Methods
WES	108 early stage (I–III) cases	108 early stage (I–III) cases
Sequencing depth average	221X	189X
Tumour content	≥30%	≥30%
Ever smokers	86%	100% (50% current)
Median follow-up time, months (range)
To survival	50.6 [1–172]	42.8
To recurrence	35 [1–162]	28.9
PD-L1, PD-1, CD3, CD4, CD8, CD45RO, CD57, granzyme B, FOXP2, CD68 by immunohistochemistry	92 cases	91 cases
TCGA supplementary data set	387 cases	178 cases
Mutations
Average coding mutations	243	209
Substitutions	Higher rates in smokers More C > A, C > T and C > A common	Most C > A
Significantly mutated genes	28	14
Most common	TP53	TP53
Known mutated genes	TP53, KEAP1, STK11, NF1, ATM, KRAS, EGFR, PIK3CA, BRAF, SMARCA4, SETD2, RBM10, U2AF1	TP53, MLL2, PIK3CA, NFE2L2, KEAP1, PTEN, GRM8, FBXW7, RB1, CDKN2A
Other mutated genes	VCAN, ROBO2, BAZ2B, FOLH1, COLI2AI, HEPACAM2, TFHDE, UBA6, INHBA, SPATA18, ZNF479, EPRS, NFATC2, LRRIQ3, ALS2CRI1	CDH8, ADCY8, PTPRT, CALCR
Mutually exclusive		NFE2L2 and KEAP1
LOH
Median events per tumour	4	21.5
Overall rate across the genome	10.6%	42.6%
CNV
CN gain	MCL1, TERT, EGFR, CDK6, MYC, MUC5AC, AKT1, ERBB2, BCL2L1	3q (SOX2, PIK3CA, TP63), MYC, BCL2L1, MCL1, CDK6, JAK3, AKT1, FGFR1, WHSC1L1
CN loss	SETD2, APC, PRDM1, TSC1, CDKN2A, TP53, STK11, SMARCA4	3P (SETD2, VHL), CDKN2A, PTCH1, APC
Prognostic significance
Poor RFS	SETD2 mutations	MLL2 (regardless of TP53 status) in those without adjuvant treatment
Poor RFS in KRAS mutated tumours	Concurrent STK11, ATM mutations
Poor response to adjuvant treatment
Mutations	EGFR, KEAP1, PIK3CA	FBXW7, KEAP1 (especially TP53 mutated tumors)
CNVs	Focal gains chr14 (AKT1)
Immunological correlations
Immune markers	Increased in smokers and those with relatively high mutation burdens, C > A transversions, KRAS and TP53 mutations	Overall upregulated immune response seen in CDKN2A mutated tumours
Down-regulated CD4+/CD8-Tcells (muted immune response)	STK11 mutations
Tumoural PD-L1	Most elevated immune marker in smokers
High	TP53 mutations
Low	PIK3CA	ADCY8, PIK3CA (also associated with downregulated peri-tumoural PD-1 expression)
Upregulated CD57 and Granzyme B (augmented NK cell infiltration)	TP53 or KEAP1	–
Upregulated CD45ro	–	PIK3CA

WES, whole exome sequencing; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; CNV, copy number variation; LOH, loss of heterozygosity; RFS recurrence free survival.

Briefly, in LUAD they found an average of 243 coding mutations, with 28 genes associated with an increased mutation burden based on a genome-wide threshold of P<2.4×10^-6. Copy number variations (CNVs) were also identified, with gains in putative oncogenes and losses in tumour suppressor genes, as expected. Correlation between the primary data set and the publicly available TCGA data is reassuring as to face-validity. The observations between genomic changes and immune profiles by immunohistochemistry (IHC) and some clinical parameters (prognosis and response to adjuvant treatment) are interesting, and add to the accumulating knowledge of potential actionable biomarkers in this cancer phenotype.

In early stage LUSC, there were a mean of 209 exonic mutations, with 14 reaching statistically significance. Recurrent CNVs were also reported including gain of chromosome 3q, reaching a frequency of 70.4% in this sample-set. Mutations (apart from CDKN2 and MLL2) and CNVs appeared consistent between this cohort and TCGA dataset. Again there were correlations between genomic changes and immunoprofiling by IHC and certain clinical characteristics as summarised in Table 1. In contrast to LUAD where a high mutational burden corresponded with elevated immune markers including PD-L1, only CD57 expression was correlated in LUSC.

Strengths of these studies include technology platforms, depth of sequencing, and demonstration of general concordance with TCGA datasets. Many known genes were replicated and some new candidates have emerged. Functional correlation is of value in helping to decipher the meaning of genomic landscape aberrations in these cancers.

The relationship to prognosis and response to adjuvant therapy is interesting, but due to the relatively small sample size and absence of control data, it is mainly hypothesis generating, rather than a definitive demonstration of prognostic or predictive power. Also, since not every case was analysed for immune molecules, there is a possibility of selection bias, so it would be informative to know whether the cases were consecutive or a convenience sample, to judge the risk of such bias. Nonetheless, these data provide a rational basis for validation studies.

In summary, these companion papers report high quality data in modestly sized lung cancer subsets and add to the pivotal data generated by TCGA. They report some new gene candidates, and potentially useful biomarkers predictive of response to therapy. Furthermore, there were correlations with immune IHC biomarkers, some of which are known to differentially affect responses to emerging immunotherapies. The future addition of multi-omic comparison, e.g., epigenomics, RNA-Seq, proteomic, as well as multi-region sampling would add further value to these data helping to better understand the increasingly recognised complexities of intra- and inter-tumoural genomic heterogeneity (3-5).