A Qualitative Transcriptional Signature for the Histological Reclassification of Lung Squamous Cell carcinomas and Adenocarcinomas

doi:10.21203/rs.2.9782/v1

Download PDF

Research article

A Qualitative Transcriptional Signature for the Histological Reclassification of Lung Squamous Cell carcinomas and Adenocarcinomas

https://doi.org/10.21203/rs.2.9782/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 21 Nov, 2019

Read the published version in BMC Genomics →

You are reading this older preprint version

Read the latest preprint version →

Background

Targeted therapy for non-small cell lung cancer is histology dependent. However, histological classification by routine pathological assessment with hematoxylin-eosin staining and immunostaining for poorly differentiated tumors, particularly those from small biopsies, is still challenging. Additionally, the effectiveness of immunomarkers is limited by technical inconsistencies of immunostaining and lack of standardization for staining interpretation.

Results

Using gene expression profiles of pathologically-determined lung squamous cell carcinomas and adenocarcinomas, denoted as pSCC and pADC respectively, we developed a qualitative transcriptional signature, based on the within-sample relative gene expression orderings (REOs) of gene pairs, to distinguish ADC from SCC. The signature consists of two genes, KRT5 and AGR2, which has the stable REO pattern of KRT5>AGR2 in pSCC and KRT5<AGR2 in pADC. In an integrated dataset for frozen tissues, the signature reclassified 8.51% of the 188 pSCC patients as ADC and 4.22% of the 805 pADC patients as SCC. Similar results were observed in the clinical challenging cases, including FFPE specimens, mixed tumors, small biopsy specimens and poorly differentiated specimens. The survival analyses showed that the pADC patients reclassified as SCC had significantly shorter overall survival than the signature-confirmed pADC patients (log-rank p = 0.0123, HR = 1.89), consisting with the knowledge that SCC patients suffer poor prognoses than ADC patients. The proliferative activity, subtype-specific marker genes and consensus clustering analyses also supported the correctness of our signature.

Conclusions

The qualitative transcriptional signature, independent of the subjective diagnosis of pathologists, would be an effective auxiliary tool in distinguishing ADC from SCC.

Epigenetics & Genomics

non-small cell lung cancer

histological subtype

pathological assessment

relative gene expression orderings

qualitative transcriptional signature

Lung cancer is the most frequent cause of cancer-related deaths worldwide. Non-small cell lung cancer (NSCLC) represents around 80% of lung cancers [ 1], with two major histological subtypes: adenocarcinoma (ADC) and squamous cell carcinoma (SCC) [ 2]. Despite sharing many biological features, ADC and SCC differ in their cell of origin, location within the lung and tumor progression [ 1, 3], suggesting that they are distinct diseases that develop through differential molecular mechanisms. Consequently, some therapy regimens for NSCLC are histology dependent. For example, compared with SCC patients, ADC patients have a higher response rate to treatment of the epidermal growth factor receptor (EGFR) tyrosine kinase inhibitor [ 4-6]. The angiogenesis inhibitor bevacizumab is approved for non-squamous patients but forbidden to SCC patients due to the high rate of life-threatening pulmonary hemorrhag [ 4, 7]. Similarly, another chemotherapy drug, pemetrexed, also has been demonstrated efficacy for ADC or non-squamous patients [ 8]. These discrepancies in tumor biology and response to drug treatment highlight the importance to distinguish ADC from SCC accurately.

Microscopic morphological features observed from hematoxylin-eosin (HE) staining are currently the golden standard for the lung cancer histological classification. In general, if there is an adequate tumor specimen and the tumor is well or moderately differentiated, imaging technique is sufficient to determine ADC or SCC [ 1]. However, the histological classification for the poorly differentiated specimens or the small biopsy specimens, which account for about 70% of the initial lung cancer diagnoses [ 9], is still a challenge. Therefore, immunohistochemistry (IHC) detection of subtype-specific markers has been proposed for assisting the histological classification of NSCLC [ 10, 11]. Before the recommendations of the WHO 2015 Classification of lung cancer, most of the poorly differentiated NSCLC cases without morphologic evidence of glandular or squamous differentiation are assigned to the large cell carcinoma (LCC) subtype [ 12, 13]. However, Rekhtman et al. have reported that, except LCC with neuroendocrine features (LCNEC), a neuroendocrine subtype of lung cancer, most LCC should be classified as ADC or SCC [ 12]. Currently, many LCCs identified according to previous criteria can be reclassified as ADC or SCC subtype based on their immunomarkers [ 12, 13]. However, even with the auxiliary of immunomarkers, there is still a certain percentage of misclassified cases because of the subjective diagnoses of HE staining or immunostaining results by pathologists using varying pathological criteria or interpretations [ 14]. Additionally, based on the combinations of SCC and ADC immunomarkers, such as TTF-1 and p63 [

                 11
              </a>], there is still about 10% samples could not be classified as they are positive or
           negative of two immunomarkers [<a href="#_ENREF_15">
              15</a>].</p>

Therefore, in recent years, considerable efforts have been devoted to extracting signatures based on gene expression profiles to stratify ADC and SCC [ 1, 16]. However, most of the reported transcriptional signatures, such as the 42-gene signature [ 1], are based on risk scores summarized from the quantitative expression measurements of the signature genes, which lack robustness for clinical applications due to large measurement batch effects [ 17] and quality uncertainties of clinical samples [ 18-20].

Fortunately, the within-sample relative expression orderings (REOs) of genes, which are the qualitative transcriptional characteristics of samples, are robust against to experimental batch effects and disease signatures based on REOs can be directly applied to samples at the individualized level [ 21-26]. Besides, we have reported that the within-sample REOs of genes are highly robust against to partial RNA degradation during specimen storage and preparation [ 18], varied proportions of the tumor cells in tumor tissues [ 19], and low-input RNA specimens [ 20]. Therefore, it is worthwhile to apply the within-sample REOs to find a robust qualitative signature for distinguishing ADC from SCC.

In this study, we developed a REOs-based qualitative signature for individualized NSCLC histological reclassification. We performed the survival analyses, proliferative activity analyses, subtype-marker genes expressions and consensus clustering analyses to provide evidences that the signature could rectify some misclassifications of histological subtypes by routine pathological assessments. Especially, the sample reclassifications by the signature were validated in various specimen types, including the frozen tissue specimens, formalin fixed paraffin-embedded (FFPE) tissue specimens, small biopsy specimens, mixed tumor specimens with high varied proportions of tumor cells and poorly differentiated tumor (LCC) specimens. Therefore, this signature would be an effective auxiliary tool for precise diagnoses of lung squamous cell carcinomas and adenocarcinomas.

Identification of the signature for distinguishing ADC from SCC

Figure 1 describes the flowchart of this study. First, using the GSE30219 dataset, we extracted 14,533 DE genes between the 61 pSCC samples and the 14 normal controls, and 10,474 DE genes between the 85 pADC samples and the 14 normal controls (SAM, FDR < 0.05). Interestingly, we found 295 genes that were DE genes in both the pSCC and pADC samples but with opposite dysregulated directions in the two types of samples when compared with the normal controls, and defined them as the subtype-opposite genes. Similarly, we found 481 subtype-opposite genes in the GSE18842 dataset. Notably, all the 148 overlapped subtype-opposite genes had consistently dysregulated directions in both pSCC and pADC samples, compared with the normal controls, respectively. The results indicated that these subtype-opposite genes were highly reproducible across different datasets. Given that a dataset may usually capture only a part of all DE genes due to insufficient statistical power [ 27, 28], we integrated the subtype-opposite genes from the two datasets, and finally obtained 495 subtype-opposite genes. Then, we utilized the subtype-opposite genes to develop a qualitative transcriptional signature for distinguishing ADC from SCC. In the training data integrated from two microarray datasets (GSE30219 and GSE18842), from all the gene pairs consisting of the subtype-opposite genes, we extracted 61,602 gene pairs with potentially subtype-opposite REO patterns (Ea> Eb in pSCC or equally Eb > Ea in pADC) occurring significantly more frequently in pSCC samples than in pADC samples (Fisher’s exact test, FDR < 0.05). Next, we calculated the C-score of each subtype-opposite gene pair in the training data. The top 50 subtype-opposite gene pairs were displayed in Supplementary Table 2. Finally, using forward selection procedure, the gene pair (KRT5 and AGR2) with the highest C-score (98.43%) was selected as the signature for distinguishing ADC from SCC, as no any other gene pair could be added to increase C-score. The classification rule of the signature is that a sample was reclassified as SCC if the mRNA expression level of KRT5 was higher than that of AGR2; otherwise ADC. According to the classification rule, two of the 61 pSCC samples in the GSE30219 dataset and one of the 31 pSCC samples in the GSE18842 dataset were reclassified as ADC and all the 99 pADC samples in the two datasets were confirmed by the signature.

Furthermore, the IHC analyses showed that Krt5 protein was mainly expressed in pSCC samples, while Agr2 protein was mainly expressed in pADC samples (Fig. 2A). For the 80 pSCC samples, 43 (53.75%) samples were stained strongly positive of Krt5 protein, while only 8 (10.00%) samples were stained strongly positive of Agr2 protein (Fig. 2B). In contrary, for the 96 pADC samples, 63 (65.63%) samples were stained strongly positive of Agr2 protein, while only 7 (7.29%) samples were stained strongly positive of Krt5 protein (Fig. 2C). The representative IHC staining of the two proteins were represented in Figure 2D and 2E. The results provided the biological evidences of the signature in distinguishing ADC from SCC. However, the IHC analyses also showed that 2 (2.5%) pSCC and 6 (6.25%) pADC samples were strongly positive of both Agr2 and Krt5 protein, and 13 (16.25%) pSCC and 12 (12.50%) pADC samples were completely negative of both Krt5 and Agr2 protein, suggesting the limitation of IHC of immunomarkers in distinguishing ADC from SCC.

Validation of the signature

Firstly, based on the knowledge that SCC patients suffer poorer prognoses than ADC patients [ 29], we evaluated the correctness of the reclassification by our signature through survival analyses. For this purpose, we integrated 8 datasets recording survival information of patients treated with curative surgery resection only. In the integrated dataset, the signature reclassified a total 16 (8.51%) of the 188 pSCC samples as ADC and a total 34 (4.22%) of the 805 pADC samples as SCC, respectively (Table S3). The survival analyses showed that the 34 pADC patients reclassified as SCC had significantly shorter OS than the remained 771 signature-confirmed pADC patients (log-rank p = 0.0123, HR = 1.89, 95% CI = 1.14-3.14, Fig. 3A), whereas the 16 pSCC patients reclassified as ADC showed longer OS than the 172 signature-confirmed pSCC patients but without significantly difference (log-rank p = 0.4122, HR = 1.46, 95% CI = 0.59-3.61, Fig. 3B). Multivariate Cox analysis showed that the pADC patients reclassified as SCC also had marginally significantly shorter OS than the signature-confirmed pADC patients, after adjusting for data centers and clinical parameters, including stage, age and gender (p = 0.0458, HR = 1.72, 95% CI = 1.01-2.93, Table S4). Notably, the 206 SCC patients reclassified by our signature had significantly shorter OS than the 787 ADC patients reclassified by the signature (log-rank p= 0.0001, HR = 1.61, 95% CI = 1.26-2.06, Fig. 3C), which was more significant than the OS difference between the original pSCC and pADC groups (log-rank p= 0.0037, HR = 1.46, 95% CI = 1.13-1.89, Fig. 3D). The OS between the two histological subtypes reclassified by our signature remained significantly different after adjusting for data centers and clinical parameters (p = 0.0018, HR = 1.56, 95% CI = 1.18-2.06, Table S5). The above results suggested that the signature could rectify some misclassifications by routine pathological assessment which confused the survival difference between the two histological subtypes. Besides, in the GSE50081 dataset with the highest reclassification rate (12.35%) in the integrated dataset, we analyzed the proliferative activities of the reclassified samples by calculating their proliferation scores. The results showed that the 15 pADC samples reclassified as SCC had significantly higher proliferation scores than the signature-confirmed pADC samples (Student’s t-test, p = 0.0310, Fig. 3E), indicating that the pADC samples reclassified as SCC are more proliferative than the signature-confirmed pADC samples which may cause worse prognoses. While the 6 pSCC samples reclassified as ADC had lower proliferation scores than the signature-confirmed pSCC samples though the difference was not significant possibly due to the small sample size (Student’s t-test, p = 0.1829, Fig. 3E). Next, we also performed differential expression analyses for the subtype-specific marker genes. The result showed that the 6 pSCC samples reclassified as ADC had significantly increased mRNA expression of an ADC marker gene (NAPSA: p< 0.0001, Fig. 3F), and decreased mRNA expressions of the two SCC marker genes (KRT5: p< 0.0001; TP63: p< 0.0001, Fig. 3F), respectively, when compared with the signature-confirmed pSCC samples. In contrast, the 15 pADC samples reclassified as SCC had significantly increased mRNA expression of the two SCC marker genes (KRT5: p< 0.0001; TP63: p< 0.0001, Fig. 3F), and decreased mRNA expressions of an ADC marker gene (NAPSA: p< 0.0001, Fig. 3F) than the signature-confirmed pADC samples. The mRNA expressions of three neuroendocrine marker genes were in very low level in all the samples. Finally, based on the mRNA expression measurements of top 1000 most variable genes in the GSE50081 dataset, the samples were optimally classified into two subgroups (k=2) using consensus clustering (Fig. S1). The clustering result showed that 5 of the 6 pSCC samples reclassified as ADC were clustered with the signature-confirmed ADC samples and 10 of the 15 pADC samples reclassified as SCC were clustered with the signature-confirmed SCC samples (Fig. S1).

Our previous study has demonstrated that REOs of gene pairs were highly stable in FFPE specimens with partial RNA degradation [ 18]. Here, we applied the signature to the GSE44170 dataset derived from FFPE specimens, and found three (7.89%) of the 38 pSCC samples were reclassified as ADC (Table S3). The three reclassified samples had lower proliferation scores than the signature-confirmed pSCC samples though the difference was not significant possibly due to the small sample size (Student’s t-test, p = 0.3896, Fig. 4A). Moreover, the three reclassified samples had (marginally) significantly decreased mRNA expression of KRT5 (p = 0.0765, Fig. 4B) and TP63 (p = 0.0033, Fig. 4B), respectively, than the signature-confirmed pSCC samples. Consensus clustering was not performed as the dataset contains mainly one subtype.

In another research, we also demonstrated that the REOs of gene pairs were robust against tumor cell variations in mixed tumor specimens [ 19]. Thus, we applied the signature to mixed tumor samples with 10% ~ 100% tumor cell in the TCGA datasets. In the TCGA-SCC dataset, the signature reclassified 82 (16.43%) of the 499 pSCC samples as ADC (Table S3), which were not significantly related with tumor cell proportions (Spearman's rank correlation, p = 0.8886). The 82 reclassified samples had significantly lower proliferation scores than the signature-confirmed pSCC samples (Student’s t-test, p< 0.0001, Fig. 4C). And, these reclassified samples had significantly increased mRNA expression of NAPSA (p< 0.0001, Fig. 4D) and decreased mRNA expressions of KRT5 (p < 0.0001, Fig. 4D) and TP63 (p < 0.0001, Fig. 4D), respectively, when compared with the signature-confirmed pSCC samples. Similarly, in the TCGA-ADC dataset, 12 (2.41%) of the 498 pADC samples were reclassified as SCC (Table S3), which were also not related with tumor cell proportions (Spearman's rank correlation, p = 0.9331). The proliferation scores of the 12 reclassified samples were significantly higher than that of the signature-confirmed pADC samples (Student’s t-test, p = 0.0486, Fig. 4E). Comparing with the signature-confirmed pADC samples, the 12 reclassified samples had significantly increased mRNA expressions of KRT5 (p < 0.0001, Fig. 4F) and TP63 (p< 0.0001, Fig. 4F), and decreased mRNA expression of NAPSA (p = 0.0126, Fig. 4F).

Previously, we have reported that the REOs of gene pairs were also robust to low-input RNA specimens. Thus, we next tested the performance of the signature in the GSE58661 dataset for small biopsy specimens, and found that the signature reclassified 4 (11.11%) of the 36 pSCC samples as ADC and 2 (4.76%) of the 42 pADC samples as SCC (Table S3). The four pSCC samples reclassified as ADC had marginally lower proliferation scores than the signature-confirmed pSCC (Student’s t-test, p = 0.0576, Fig. 5A). Although the two pADC samples reclassified as SCC had lower proliferation scores, they had (marginally) significantly increased mRNA expression of KRT5 (p< 0.0001, Fig. 5B) and TP63 (p< 0.0001, Fig. 5B) than the signature-confirmed pADC samples, indicating the correctness of the reclassification for the two samples. In contrast, the 4 pSCC samples reclassified as ADC had significantly increased mRNA expression of NAPSA (p < 0.0001, Fig. 5B), and decreased mRNA expressions of KRT5 (p< 0.0001, Fig. 5B) and TP63 (p< 0.0001, Fig. 5B), respectively, than the signature-confirmed pSCC samples. Additionally, consensus clustering based on the top 1000 most variable genes showed that all the 4 pSCC samples reclassified as ADC by the signature were clustered with the signature-confirmed ADC samples (Fig. S2).

Finally, we also evaluated the performance of the signature for poorly differentiated specimens. In the GSE94601 dataset, the 4 and 19 poorly differentially samples initially assigned to the LCC subtype were reclassified to pSCC and pADC with the positive expressions of SCC markers (Krt5/P40) and ADC markers (NAPSA/TTF1), respectively. The signature confirmed the histological reclassification of all the 19 pADC and reclassified 2 of the 4 pSCC samples as ADC (Table S3). The proliferation scores of the two reclassified samples were 8.82 and 8.83, respectively, which were similar with the median proliferation scores (8.89) of the 19 signature-confirmed pADC samples but lower than that (9.67 and 9.06) of the two signature-confirmed pSCC samples though the difference was not significant possibly due to the small sample size (Student’s t-test, p=0.3232, Fig. 5C). Additionally, we also performed differential expression analysis for the 44 proliferation-related genes [ 30] and found 20 genes that were differentially expressed in the two reclassified samples (RankProd algorithm, p < 0.05), and all the 20 genes were down-regulated (Fig. 5D), when compared with the two signature-confirmed pSCC samples. The results provide tentative evidences for the correctness of the reclassification by our signature. Consensus clustering was not performed as the small sample size of pSCC.

Taken together, the above results indicated that the signature could accurately distinguish ADC from SCC, especially for the clinical challenging cases.

In this study, we developed a robust qualitative transcriptional signature consisting of two genes (KRT5 and AGR2), which can accurately distinguish ADC from SCC and is independent of the subjective experiences of pathologists. It is known that Krt5 protein is expressed primarily in basal keratinocytes in the epidermis, whose overexpression is the unique characteristic of SCC [ 31]. Another gene in the signature, Agr2, is known as an adenocarcinoma antigen, which promotes cell migration and metastasis [ 32, 33]. Our immunohistochemistry results demonstrated that krt5 protein was preferentially expressed in pSCC, while Agr2 protein was preferentially expressed in pADC, providing biological evidences of the signature in distinguishing ADC from SCC. However, it is note that, there were 8 (4.55%) samples with strongly positive of the two proteins and 25(14.20%) samples with completely negative of the two proteins, indicating that a certain percentage of patients could not be accurately classified by the immunohistochemical assessment. The limitation was also observed in the commonly used immunomarkers for ADC and SCC [ 15].

In all the datasets analyzed in this study, our signature could reclassify a total 13.99% of the 765 pSCC patients as ADC and a total 3.52% of the 1,346 pADC patients as SCC. Notably, there were large differences in the reclassification rates among cohorts. It has been reported that different pathologists had discordant histological classifications for about one third of the cases because of their subjective diagnoses using varying criteria or interpretations, which might be the major factor causing different misclassification rates among cohorts [ 14]. Therefore, we first supported the reclassifications of our signature by the survival and proliferative activity analyses based on the knowledge that SCC suffers more proliferative activity than ADC, which may cause worse prognoses [ 29, 34]. Next, subtype-specific marker genes and consensus clustering analyses also provided the transcriptional evidences for supporting the histological classifications by our signature. These results together suggested that our signature could rectify some misclassifications of NSCLC histological subtypes by current pathological assessment, especially for the clinical challenging cases, such as FFPE specimens with partial RNA degradation, mixed tumors with varied tumor cells proportions, small biopsy specimens with low-input mRNA specimens and poorly differentiated specimens, indicating the necessity for molecular classification.

Previously, several gene expression signatures to stratify ADC and SCC have been reported [ 1, 16, 35]. We additionally compared the performance of our signature with a previously reported 42-gene signature for stratifying ADC and SCC, which has more advantages than the other reported signatures [ 1]. In the 12 datasets tested in this study, the total C-score of our signature was 92.72%, slightly higher than that (92. 06%) of the 42-gene signature. More importantly, the survival analyses showed that the OS difference between the ADC and SCC subtypes reclassified by our signature (log-rank p = 0.0001, HR = 1.61, 95% CI = 1.26-2.06, Fig. 3C) was more significant than that between the two histological subtypes reclassified by the 42-gene signature (log-rank p = 0.0029, HR = 1.47, 95% CI = 1.14-1.89, Fig. S3). The above results indicated that our signature could perform better than the 42-gene signature in rectifying some misclassifications by routine pathological assessment. The proliferation scores and gene expression patterns were not compared due to the small sample size of the inconsistent samples reclassified by the two signatures in each dataset. Moreover, comparing with the 42-gene signature, whose classification is dependent on the samples in its training dataset, our signature is much more convenient as it only needs detection of two genes within an individual and is independent of the other samples.

However, a limitation of this study is that most of the samples we used were diagnosed according to WHO 2004 criteria because few samples diagnosed according to WHO 2015 criteria could be found in public data. It is known that the current criteria auxiliary with immunostaining could reclassify a subset of cases diagnosed with the WHO 2004 criteria [ 36]. However, the choice of immunomarkers [ 36] and the subjective diagnosis of immunostaining results by pathologists have effects on the histological classification [ 37]. For the 23 poorly differentiated samples in the GSE94601 dataset, which were classified according to WHO 2015 criteria, our signature confirmed all the 19 pADC samples and reclassified two of the four pSCC samples as ADC. We provided tentatively evidences for supporting the reclassifications by our signature through estimating the proliferative activities of the two reclassified samples, based on the phenomenon that SCC has higher proliferative activity with poorer survival than ADC. The reclassifications by our signature for the samples diagnosed according to WHO 2015 criteria should be further evaluated. Another limitation of this study is that the current signature can only distinguish ADC and SCC. We additionally applied the signature to the other subtypes of lung cancer (GSE60644), including three small cell lung cancer and nine large cell neuroendocrine carcinoma. The result showed that all these samples were reclassified into ADC group, indicating that the signature might have a better performance in distinguishing non-SCC from SCC, which should be instructive for treatment as most histology-dependent therapies for NSCLC are directed at ADC or non-squamous histological types [ 38]. Therefore, the current signature should be an auxiliary tool for the poorly differentiated NSCLC, which is a challenging for clinical diagnosis. It is noted that, the REO approach can handle more than 2 output classes [ 25] but currently we cannot find sufficient samples for the other NSCLC subtypes, which merits exploration in the future.

Conclusions

The qualitative signature, which is independent of the subjective diagnoses of pathologists, would be an effective auxiliary tool for the histological classification of NSCLC, especially for the clinical challenging cases. Notably, it is worthwhile to design a customized microarray or RNA-sequencing panel to measure all possible qualitative transcriptional signatures together (“a panel for all”) for diagnosis, histological classification, prognosis and drug evaluation of lung cancer, which could preserve precious tissue for the other molecular testing such as EGFR mutation detection.

The public data sources and data pre-processing

The 12 public gene expression datasets of NSCLC tissues (Table 1) were collected from Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/), ArrayExpress (https://www.ebi.ac.uk/arrayexpress/) and The Cancer Genome Atlas (TCGA, http://cancergenome.nih.gov/).

The training data for extracting a REOs-based signature was integrated from two microarray datasets (GSE30219 and GSE18842), including 99 pathologically-determined ADC (denoted as pADC) samples, 92 pathologically-determined SCC (denoted as pSCC) samples and 59 normal controls.

To test the predictive performance of the signature, 8 datasets (Table 1) recording survival information were integrated for survival analyses, including 188 pSCC and 805 pADC samples of stage I-III patients treated with curative surgery resection only. The clinical information of the 8 datasets was displayed in Supplementary Table 1. The signature was also tested in FFPE specimens with partial RNA degradation (GSE44170), mixed tumors with varied proportions of tumor cells (TCGA-SCC, TCGA-ADC), small biopsy specimens with low-input RNA (GSE58661) and poorly differentiated specimens (GSE94601). Notably, the GSE94601 dataset includes 4 poorly differentiated pSCC and 19 poorly differentiated pADC samples, which were improperly assigned to LCC subtype before and reclassified by the authors using SCC and ADC immunomarkers according to the recommendations of the WHO 2015 criteria [ 39]. Except these reclassified LCC samples, histological classifications of the other tumors in the above datasets were diagnosed according to the WHO 2004 criteria [ 1].

For data generated by the Affymetrix microarray platforms, the Robust Multi-array Average algorithm [ 40] was used for preprocessing the raw data and the other datasets generated by the Illumina platforms, the originally processed data (series matrix files) were used. Probe IDs were mapped to Gene IDs using the corresponding platform files. For each sample, the expression measurements of all probes corresponding to the same Gene ID were averaged to obtain a single measurement. Probes that did not match any Gene ID or matched multiple Gene IDs were deleted. For TCGA data (SCC/ADC), the normalized count values of level 3 gene expression data derived from Illumina HiSeqV2 were extracted as gene expression measurements.

Developing the signature for distinguishing ADC from SCC

First, Significance Analysis of Microarrays (SAM) algorithm [

                 41
              </a>] was used to identify differentially expressed genes (DE genes) in pSCC and pADC
           respectively, when compared with normal controls. The <em>p </em>values were adjusted using the Benjamini-Hochberg procedure for multiple testing to control the false discovery rate (FDR) [<a href="#_ENREF_42">
              42</a>]. We selected the genes which were differentially expressed in both pSCC and pADC
           but with opposite dysregulated directions when compared with normal controls, and
           defined them as subtype-opposite genes. </p>

Then, for a pair of genes (a and b) both derived from the subtype-opposite genes, Fisher’s exact test was used to evaluate whether the frequency of pSCC samples with a specific REO pattern (Ea > Eb), where Ea and Eb represent the expression levels of gene a and b, respectively, was significantly higher than the frequency in pADC samples. The significant gene pairs were defined as subtype-opposite gene pairs.

Finally, for each subtype-opposite gene pair, we calculated the concordance score (C-score, Formula 1) of its histological classification classified by this gene pair and routine pathological assessment. We chose the gene pair with the highest C-score as a seed and performed forward selection procedure to add another subtype-opposite gene pairs with the majority vote rule to construct a signature that achieved the highest C-score.

C-score=C/M*100% [ 1]

where C is the number of samples which were consistently classified by the signature and their original histological classification, and M is the total number of samples in the dataset.

Immunohistochemical analysis of lung cancer tissue microarrays

The slides of human lung cancer tissue microarray (#LC2161, Alenabio, Xi'an, China) used in this study were dewaxed with xylene and rehydrated through gradient ethanol into water. Antigens were unmasked by 10 mM citric acid buffer. Endogenous peroxidase activity was blocked by 3% hydrogen peroxidase. After further blocking for non-specific binding, the slides were incubated with specific anti-Krt5 (#AF0194, Affinity) or anti-Agr2 (ab76473, Abcam) primary antibody and then with secondary antibody conjugated with HRP. The slides were developed with DAB substrates, counterstained with haematoxylin, dehydrated and mounted with neutral balsam. Each tissue sample of the microarray was then imaged under miscroscope Carl Zeiss GmbH with same settings. The protein expression was quantified and scored basing on a multiplicative index of the average staining intensity (1-4) and the extent of staining (1-4 in the cores, yielding a 16-point staining index that ranged from 1 (no staining) to 16 (extensive, strong staining). The staining intensity was scored as follows: 1, negative; 2, weak; 3, moderate; and 4, strong. The extent of staining was scored as follows: 1, less than 10% positive cells; 2, 10% to 30%; 3, 30% to 70%; and 4, more than 70%. Score of staining index less than 5 was considered low expression, score of 5-10 considered medium expression and score of 11-16 considered high expression. Fisher’s exact test was used to compare the frequencies of the Krt5 and Agr2 proteins expressions between 96 pADC samples and 80 pSCC samples.

Survival analyses

To avoid the bias of patient follow-up duration among different datasets, the overall survival (OS) of patients were truncated at 60 months. Survival curves were estimated using the Kaplan-Meier method and were compared using the log-rank test [ 43]. A multivariate Cox proportional-hazards regression model was used to evaluate independent associations between predictive factors and patient survival after adjusting for data centers and clinical parameters, including stage, age and gender. Hazard ratios (HRs) and 95% confidence intervals (CIs) were generated using Cox proportional hazards model.

Proliferation score calculation

The proliferation score for each sample was calculated by averaging the expressions of the 44 genes in a proliferation signature [ 30] in a cohort. A higher proliferation score indicates more proliferative activity. Student’s t-test was used to test the difference of proliferation scores between two groups.

Differential expression and consensus clustering analyses

The RankProd algorithm [ 44], an effective method for small sample sizes, was used to estimate whether the subtype-specific marker genes were differentially expressed between reclassified samples and signature-confirmed samples. The subtype-specific marker genes include SCC marker genes (KRT5, TP63) [ 11], ADC marker genes (NAPSA, TTF1) [ 11] and neuroendocrine marker genes (CD56, SYP, CHGA) [ 45].

Consensusclustering was performed using the ConsensusClusterPlus package according to the Ward method for hierarchical clustering [ 46]. The samples were clustered using top 1000 most variable genes across the pADC and pSCC samples in a cohort, which were determined by median absolute deviation. To identify the optimum number of clusters, k corresponding to the first downwards inflection in cumulative distribution function (CDF) was used.

All statistical analyses were performed using the R 2.15.3 (http://www.r-project.org/).

NSCLC: Non-small cell lung cancer; ADC: adenocarcinoma; SCC: squamous cell carcinoma; EGFR: epidermal growth factor receptor; LCC: large cell carcinoma; NSCLC-NOS: NSCLC-not otherwise specified; REOs: relative expression orderings; pSCC: pathologically-determined SCC; pADC: pathologically-determined ADC; FFPE: formalin fixed paraffin-embedded; SAM: Significance Analysis of Microarrays; DE genes: differentially expressed genes; FDR: false discovery rate; OS: overall survival; HRs: Hazard ratios; CIs: confidence intervals; FC: fold change

Notes

Xin Li and Gengen Shi contributed equally to this work.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Availability of data and material

The datasets generated and/or analysed during the current study are available in the Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/), ArrayExpress (https://www.ebi.ac.uk/arrayexpress/) and The Cancer Genome Atlas (TCGA, http://cancergenome.nih.gov/).

Competing interests

The authors declare that they have no competing interests.

Funding

This work was supported by National Natural Science Foundation of China (Grant numbers: 61701143, 81372213 and 81572935), the Joint Scientific and Technology Innovation Fund of Fujian Province (Grant number: 2016Y9044) and the Fundamental Research Funds for the Provincial Universities (2017JCZX48).

Authors' contributions

LSQ conceived and designed the experiments, wrote the manuscript, ZG conceived the idea, XL, GGS designed and performed the experiments, QSC performed immunohistochemical analysis, WBJ, YXL analyzed the data, YL, SNZ preprocessed the data, FH provided and processed the data, XXL collected the data. All authors approved the final version.

Acknowledgements

Not applicable.

1. Girard L, Rodriguez-Canales J, Behrens C, Thompson DM, Botros IW, Tang H, Xie Y, Rekhtman N, Travis WD, Wistuba, II et al: An Expression Signature as an Aid to the Histologic Classification of Non-Small Cell Lung Cancer. Clin Cancer Res 2016, 22(19):4880-4889.

2. Tian S: Classification and survival prediction for early-stage lung adenocarcinoma and squamous cell carcinoma patients. Oncol Lett 2017, 14(5):5464-5470.

3. Sun S, Schiller JH, Gazdar AF: Lung cancer in never smokers--a different disease. Nat Rev Cancer 2007, 7(10):778-790.

4. Bishop JA, Benjamin H, Cholakh H, Chajut A, Clark DP, Westra WH: Accurate classification of non-small cell lung carcinoma using a novel microRNA-based approach. Clin Cancer Res 2010, 16(2):610-619.

5. Montezuma D, Azevedo R, Lopes P, Vieira R, Cunha AL, Henrique R: A panel of four immunohistochemical markers (CK7, CK20, TTF-1, and p63) allows accurate diagnosis of primary and metastatic lung carcinoma on biopsy specimens. Virchows Arch 2013, 463(6):749-754.

6. Lindeman NI, Cagle PT, Beasley MB, Chitale DA, Dacic S, Giaccone G, Jenkins RB, Kwiatkowski DJ, Saldivar JS, Squire J et al: Molecular testing guideline for selection of lung cancer patients for EGFR and ALK tyrosine kinase inhibitors: guideline from the College of American Pathologists, International Association for the Study of Lung Cancer, and Association for Molecular Pathology. J Mol Diagn 2013, 15(4):415-453.

7. Johnson DH, Fehrenbacher L, Novotny WF, Herbst RS, Nemunaitis JJ, Jablons DM, Langer CJ, DeVore RF, 3rd, Gaudreault J, Damico LA et al: Randomized phase II trial comparing bevacizumab plus carboplatin and paclitaxel with carboplatin and paclitaxel alone in previously untreated locally advanced or metastatic non-small-cell lung cancer. J Clin Oncol 2004, 22(11):2184-2191.

8. Klein R, Wielage R, Muehlenbein C, Liepa AM, Babineaux S, Lawson A, Schwartzberg L: Cost-effectiveness of pemetrexed as first-line maintenance therapy for advanced nonsquamous non-small cell lung cancer. J Thorac Oncol 2010, 5(8):1263-1272.

9. Travis WD, Brambilla E, Noguchi M, Nicholson AG, Geisinger K, Yatabe Y, Ishikawa Y, Wistuba I, Flieder DB, Franklin W et al: Diagnosis of lung cancer in small biopsies and cytology: implications of the 2011 International Association for the Study of Lung Cancer/American Thoracic Society/European Respiratory Society classification. Arch Pathol Lab Med 2013, 137(5):668-684.

10. Zhan C, Yan L, Wang L, Sun Y, Wang X, Lin Z, Zhang Y, Shi Y, Jiang W, Wang Q: Identification of immunohistochemical markers for distinguishing lung adenocarcinoma from squamous cell carcinoma. J Thorac Dis 2015, 7(8):1398-1405.

11. Kim MJ, Shin HC, Shin KC, Ro JY: Best immunohistochemical panel in distinguishing adenocarcinoma from squamous cell carcinoma of lung: tissue microarray assay in resected lung cancer specimens. Ann Diagn Pathol 2013, 17(1):85-90.

12. Rekhtman N, Tafe LJ, Chaft JE, Wang L, Arcila ME, Colanta A, Moreira AL, Zakowski MF, Travis WD, Sima CS et al: Distinct profile of driver mutations and clinical features in immunomarker-defined subsets of pulmonary large-cell carcinoma. Mod Pathol 2013, 26(4):511-522.

13. Clinical Lung Cancer Genome P, Network Genomic M: A genomics-based classification of human lung tumors. Sci Transl Med 2013, 5(209):209ra153.

14. Stang A, Pohlabeln H, Muller KM, Jahn I, Giersiepen K, Jockel KH: Diagnostic agreement in the histopathological evaluation of lung cancer tissue in a population-based case-control study. Lung Cancer 2006, 52(1):29-36.

15. Osmani L, Askin F, Gabrielson E, Li QK: Current WHO guidelines and the critical role of immunohistochemical markers in the subclassification of non-small cell lung carcinoma (NSCLC): Moving from targeted therapy to immunotherapy. Seminars in cancer biology 2018, 52(Pt 1):103-109.

16. Zhang A, Wang C, Wang S, Li L, Liu Z, Tian S: Visualization-aided classification ensembles discriminate lung adenocarcinoma and squamous cell carcinoma samples using their gene expression profiles. PLoS One 2014, 9(10):e110052.

17. Guan Q, Yan H, Chen Y, Zheng B, Cai H, He J, Song K, Guo Y, Ao L, Liu H et al: Quantitative or qualitative transcriptional diagnostic signatures? A case study for colorectal cancer. BMC Genomics 2018, 19(1):99.

18. Chen R, Guan Q, Cheng J, He J, Liu H, Cai H, Hong G, Zhang J, Li N, Ao L et al: Robust transcriptional tumor signatures applicable to both formalin-fixed paraffin-embedded and fresh-frozen samples. Oncotarget 2017, 8(4):6652-6662.

19. Cheng J, Guo Y, Gao Q, Li H, Yan H, Li M, Cai H, Zheng W, Li X, Jiang W et al: Circumvent the uncertainty in the applications of transcriptional signatures to tumor tissues sampled from different tumor sites. Oncotarget 2017, 8(18):30265-30275.

20. Liu H, Li Y, He J, Guan Q, Chen R, Yan H, Zheng W, Song K, Cai H, Guo Y et al: Robust transcriptional signatures for low-input RNA samples based on relative expression orderings. BMC Genomics 2017, 18(1):913.

21. Eddy JA, Sung J, Geman D, Price ND: Relative expression analysis for molecular cancer diagnosis and prognosis. Technol Cancer Res Treat 2010, 9(2):149-159.

22. Qi L, Chen L, Li Y, Qin Y, Pan R, Zhao W, Gu Y, Wang H, Wang R, Chen X et al: Critical limitations of prognostic signatures based on risk scores summarized from gene expression levels: a case study for resected stage I non-small-cell lung cancer. Brief Bioinform 2016, 17(2):233-242.

23. Qi L, Li T, Shi G, Wang J, Li X, Zhang S, Chen L, Qin Y, Gu Y, Zhao W et al: An individualized gene expression signature for prediction of lung adenocarcinoma metastases. Mol Oncol 2017, 11(11):1630-1645.

24. Qi L, Li Y, Qin Y, Shi G, Li T, Wang J, Chen L, Gu Y, Zhao W, Guo Z: An individualised signature for predicting response with concordant survival benefit for lung adenocarcinoma patients receiving platinum-based chemotherapy. British journal of cancer 2016, 115(12):1513-1519.

25. Song K, Guo Y, Wang X, Cai H, Zheng W, Li N, Song X, Ao L, Guo Z, Zhao W: Transcriptional signatures for coupled predictions of stage II and III colorectal cancer metastasis and fluorouracil-based adjuvant chemotherapy benefit. FASEB J 2018:fj201800222RRR.

26. Peng F, Wang R, Zhang Y, Zhao Z, Zhou W, Chang Z, Liang H, Zhao W, Qi L, Guo Z et al: Differential expression analysis at the individual level reveals a lncRNA prognostic signature for lung adenocarcinoma. Mol Cancer 2017, 16(1):98.

27. Wang D, Cheng L, Wang M, Wu R, Li P, Li B, Zhang Y, Gu Y, Zhao W, Wang C et al: Extensive increase of microarray signals in cancers calls for novel normalization assumptions. Comput Biol Chem 2011, 35(3):126-130.

28. Zhang M, Zhang L, Zou J, Yao C, Xiao H, Liu Q, Wang J, Wang D, Wang C, Guo Z: Evaluating reproducibility of differential expression discoveries in microarray studies by considering correlated molecular changes. Bioinformatics 2009, 25(13):1662-1668.

29. Gyorffy B, Surowiak P, Budczies J, Lanczky A: Online survival analysis software to assess the prognostic value of biomarkers using transcriptomic data in non-small-cell lung cancer. PLoS One 2013, 8(12):e82241.

30. Whitfield ML, George LK, Grant GD, Perou CM: Common markers of proliferation. Nat Rev Cancer 2006, 6(2):99-106.

31. Khayyata S, Yun S, Pasha T, Jian B, McGrath C, Yu G, Gupta P, Baloch Z: Value of P63 and CK5/6 in distinguishing squamous cell carcinoma from adenocarcinoma in lung fine-needle aspiration specimens. Diagn Cytopathol 2009, 37(3):178-183.

32. Wang Z, Hao Y, Lowe AW: The adenocarcinoma-associated antigen, AGR2, promotes tumor growth, cell migration, and cellular transformation. Cancer Res 2008, 68(2):492-497.

33. Brychtova V, Vojtesek B, Hrstka R: Anterior gradient 2: a novel player in tumor cell biology. Cancer Lett 2011, 304(1):1-7.

34. Cooke DT, Nguyen DV, Yang Y, Chen SL, Yu C, Calhoun RF: Survival comparison of adenosquamous, squamous cell, and adenocarcinoma of the lung after lobectomy. Ann Thorac Surg 2010, 90(3):943-948.

35. Charkiewicz R, Niklinski J, Claesen J, Sulewska A, Kozlowski M, Michalska-Falkowska A, Reszec J, Moniuszko M, Naumnik W, Niklinska W: Gene Expression Signature Differentiates Histology But Not Progression Status of Early-Stage NSCLC. Transl Oncol 2017, 10(3):450-458.

36. Micke P, Mattsson JS, Djureinovic D, Nodin B, Jirstrom K, Tran L, Jonsson P, Planck M, Botling J, Brunnstrom H: The Impact of the Fourth Edition of the WHO Classification of Lung Tumours on Histological Classification of Resected Pulmonary NSCCs. J Thorac Oncol 2016, 11(6):862-872.

37. Lewis DR, Check DP, Caporaso NE, Travis WD, Devesa SS: US lung cancer trends by histologic type. Cancer 2014, 120(18):2883-2892.

38. Oliver TG, Patel J, Akerley W: Squamous non-small cell lung cancer as a distinct clinical entity. Am J Clin Oncol 2015, 38(2):220-226.

39. Marx A, Chan JK, Coindre JM, Detterbeck F, Girard N, Harris NL, Jaffe ES, Kurrer MO, Marom EM, Moreira AL et al: The 2015 World Health Organization Classification of Tumors of the Thymus: Continuity and Changes. J Thorac Oncol 2015, 10(10):1383-1395.

40. Irizarry RA, Hobbs B, Collin F, Beazer-Barclay YD, Antonellis KJ, Scherf U, Speed TP: Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics 2003, 4(2):249-264.

41. Tusher VG, Tibshirani R, Chu G: Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 2001, 98(9):5116-5121.

42. Benjamini Y, Hochberg Y: Controlling the false discovery rate: a practical and powerful approach to multiple testing. Journal of the royal statistical society 1995, 57: 289-300.

43. Bland JM, Altman DG: The logrank test. BMJ 2004, 328(7447):1073.

44. Hong F, Breitling R, McEntee CW, Wittner BS, Nemhauser JL, Chory J: RankProd: a bioconductor package for detecting differentially expressed genes in meta-analysis. Bioinformatics 2006, 22(22):2825-2827.

45. Karlsson A, Brunnstrom H, Micke P, Veerla S, Mattsson J, La Fleur L, Botling J, Jonsson M, Reutersward C, Planck M et al: Gene Expression Profiling of Large Cell Lung Cancer Links Transcriptional Phenotypes to the New Histological WHO 2015 Classification. J Thorac Oncol 2017, 12(8):1257-1267.

46. Wilkerson MD, Hayes DN: ConsensusClusterPlus: a class discovery tool with confidence assessments and item tracking. Bioinformatics 2010, 26(12):1572-1573.

Table 1. The datasets analyzed in this study

Types	Data Source	Database	Platform	pSCC	pADC	Normal
Train (frozen)	GSE30219	GEO	Affy. Plus 2	61	85	14
	GSE18842	GEO	Affy. Plus 2	31	14	45
	Total	--	--	92	99	59
Integrated data (frozen)	E-MTAB-2435*	ArrayExpress	Affy. Plus 2	63	0	--
	GSE42127*	GEO	Illu. WG V3.0	32	90	--
	GSE50081*	GEO	Affy. Plus 2	43	127	--
	GSE37745*	GEO	Affy. Plus 2	24	40	--
	GSE31210*	GEO	Affy. Plus 2	0	204	--
	GSE31546*	GEO	Affy. Plus 2	0	13	--
	GSE14814*	GEO	Affy. U133A	26	32	--
	GSE68465*	GEO	Affy. U133A	0	299	--
	Total	--	--	188	805	--
FFPE	GSE44170	GEO	Affy. U133A	38	0	--
mixed	TCGA	TCGA	Illu. HiSeqV2	499	498	--
biopsy	GSE58661	GEO	Affy. 2.0	36	42	--
frozen	GSE94601	GEO	Illu. HT V4.0	4^#	19^#	--

pADC = pathologically-determined ADC, pSCC = pathologically-determined SCC; Affy. Plus 2 = Affymetrix Plus 2, Affy. U133A = Affymetrix U133A, Affy. 2.0 = Rosetta/Merck Human RSTA Custom Affymetrix 2.0, Illu. WG V3.0 = Illumina HumanWG-6 V3.0, Illu. HT V3.0 = Illumina HumanHT-12 V3.0, Illu. HiSeqV2=Illumina HiSeqV2, Illumina HT V4.0 = Illumina HumanHT-12 V4.0;
* the data records the survival information of patients treated with curative surgery resection only;
# the 4 pSCCs and 19 pADCs samples are poorly differentiated which were improperly assigned to LCC subtype before and reclassified by the authors using SCC and ADC immunomarkers.

supplement1.docx

Download PDF

Journal Publication

published 21 Nov, 2019

Read the published version in BMC Genomics →

Editorial decision: Major revision
04 Jul, 2019
Review #2 received at journal
26 Jun, 2019
Review #1 received at journal
07 Jun, 2019
Reviewer #2 agreed at journal
26 May, 2019
Reviewers invited by journal
23 May, 2019
Reviewer #1 agreed at journal
23 May, 2019
Submission checks completed at journal
22 May, 2019
Editor invited by journal
19 May, 2019
Editor assigned by journal
19 May, 2019
First submitted to journal
02 May, 2019

You are reading this older preprint version

Read the latest preprint version →

A Qualitative Transcriptional Signature for the Histological Reclassification of Lung Squamous Cell carcinomas and Adenocarcinomas

Status:

Journal Publication

Version 1

Abstract

Figures

Background

Results

Identification of the signature for distinguishing ADC from SCC

Validation of the signature

Discussion

Conclusions

Methods

The public data sources and data pre-processing

Developing the signature for distinguishing ADC from SCC

C-score=C/M*100% [ 1]

Immunohistochemical analysis of lung cancer tissue microarrays

Survival analyses

Proliferation score calculation

Differential expression and consensus clustering analyses

Abbreviations

Declarations

Notes

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Availability of data and material

Competing interests

Funding

Authors' contributions

Acknowledgements

Not applicable.

References

Tables

Supplementary Files

Status:

Journal Publication

Version 1