A Focal Region on 8p Is Preferentially Amplified in SqCC
We compared the 8p chromosome arm of 161 microdissected NSCLC tumors—103 AC and 58 SqCC (Table S1
, sample set 1a)—by tiling resolution array CGH 
. After hybridization experiments, genomic profiles were normalized and subjected to a smoothing algorithm in order to computationally define regions of copy number gain and loss along the entire length of Chromosome arm 8p 
. Individual samples were then grouped by their corresponding cell type, and probes were aggregated into regions on the basis of similar copy number status. The resulting frequency of alteration for each region along the arm was compared between cell types using the Fisher exact test to identify regions of copy number disparity, and the resulting p
-values were corrected for multiple comparisons with a cut-off of ≤0.01 considered significant (Methods
). Although the telomeric portion of 8p was frequently lost in both AC and SqCC cell types, two regions spanning a total of 5.65 Mbp at 8p12–8p11.21 were found to be frequently gained specifically in SqCC (; Table S2
). Copy number increase of focal regions at 8p12–p11.21 was found in up to 40% of SqCC tumors, while DNA loss was the most prevalent event in AC (~39%). In addition, high-level amplification (log2
ratio >0.6) was present in ~12% of SqCC samples (seven out of 58) demonstrating the preferential selection for this alteration in tumors of this cell lineage. The increased incidence of 8p amplification in comparison to previous reports is attributed to analyzing the cell types as distinct groups, as opposed to combining all the NSCLC cell types as a single entity. In addition, the small sample sizes of previous studies may also have limited the detection of specific disruptions unique to each cell type to just extremely high frequency events, such as the gain of Chromosome 3q 
. These results indicate that gain/amplification of 8p12–8p11.21 is restricted to SqCC and occurs far more frequently than previously thought, highlighting the importance of considering cell lineage in genomic studies of malignancies from the same tissue site.
Chromosome 8p amplification in NSCLC is restricted to the SqCC lineage.
BRF2 Gene Expression Drives Selection of the 8p Amplicon in Lung SqCC
The cell type dependent pattern of 8p amplification raised the possibility that a lineage-specific oncogene may be driving the preferential selection of this amplicon in SqCC. Such a gene should display five fundamental properties each translating into its own testable hypothesis. First, increased expression would be restricted to SqCC tumors mirroring the specificity of DNA amplification (hypothesis 1). Second, as the target of the amplicon, expression would be higher in SqCC tumors with gain/amplification than those without (hypothesis 2). Third, expression should be significantly higher in SqCC tumors than normal bronchial epithelial cells; that is, the gene should be activated in cancerous and not normal tissue (hypothesis 3). Fourth, the gene should have oncogenic potential and provide a growth and/or survival advantage to cells when overexpressed (hypothesis 4). Lastly, if necessary for initiating tumorigenesis, amplification should occur early in tumor development and therefore be present in lung SqCC precursor lesions (hypothesis 5).
To test the first hypothesis, we generated gene expression microarray profiles for a subset of 47 tumors (34 AC, 13 SqCC) with sufficient amounts of material that were also analyzed by array CGH in order to integrate genetic and gene expression information (Table S1
, sample set 1b). In total, 62 probes corresponding to 44 unique genes mapped to within the alteration boundaries (Table S2
). To identify lineage-restricted genes, we compared the expression levels for all probes between the AC and SqCC samples. Since we predicted candidate genes to be overexpressed in SqCC, a one-tailed Mann-Whitney U test was used with Benjamini-Hochberg corrected p
-values ≤0.01 considered significant. Ten unique genes meeting these criteria were uncovered from this analysis that showed a clear distinction in expression levels between the AC and SqCC tumors (; Table S3
After identifying these SqCC specific genes, we next aimed to ensure that amplification is responsible for their differential expression, as these will be candidate targets driving amplicon selection (hypothesis 2). For this purpose, we utilized two complementary approaches. First, a nonparametric Spearman correlation coefficient was calculated for each gene using Z-transformed copy number ratios and log10
gene expression ratios (Methods
). Five of the ten genes (LSM1
, and WHSC1L1
) had a correlation coefficient of >0.75 and a corrected p
-value (representing the statistical significance of a positive correlation) of <0.01 and were further considered as candidates (Table S3
for all values). The second approach involved the comparison of expression levels between SqCC tumors with gene dosage increase (gain/amplification) and those with neutral copy number status (Methods
). Of the five genes with a positive association between copy number and expression, only three (LSM1
, and ASH2L
) also showed significantly elevated transcript levels specifically in SqCC samples with gain or amplification and were therefore determined to be regulated by copy number (Table S3
). qRT-PCR analysis of BRF2
(the most likely target gene, see below) confirmed the microarray results (Figure S1
; Table S4
). Importantly, none of these genes demonstrated a correlation between copy number and expression in AC, reinforcing the specificity of this alteration to SqCC.
In addition to demonstrating a linkage between expression and amplification, a candidate oncogene should only be expressed at elevated levels in cancerous, and not normal, tissues 
. Therefore, to test the third hypothesis, we analyzed the RNA levels of these three genes in an independent panel of 53 SqCC lung tumors and 67 samples of exfoliated bronchial cells from cancer-free individuals generated using the Affymetrix U133 Plus 2 platform (Table S1
, sample sets 2 and 3). Strikingly, only BRF2
was aberrantly expressed (>2-fold, p
) in cancerous tissues identifying it as the sole gene passing the three main criteria of a candidate lineage-specific oncogene described above (; Table S3
). To further confirm these observations, a third, independent sample set consisting of 118 NSCLC tumors and 39 non-neoplastic lung tissues (Table S1
, sample set 4) was analyzed for BRF2
expression by qRT-PCR (Methods
). Consistent with the microarray results, expression of BRF2
in primary tumors was significantly higher than that in the non-neoplasia tissues (p
<0.001) with overexpression more common in SqCC than AC (p
0.03), supporting our findings.
BRF2 is a lineage-specific oncogene targeted by amplification in SqCC.
Taken together, results from testing the first three hypotheses clearly demonstrate that BRF2
is the driver gene of the 8p amplicon and identify it as a candidate lineage-specific oncogene in SqCC. Previous studies investigating this amplified region in NSCLC have proposed FGFR1
as potential oncogenes 
. However, we ruled out FGFR1
as a possible target as it was not differentially expressed between AC and SqCC, and as such, was excluded from further analysis. This conclusion is in agreement with a study by Tonon et al. that suggested WHSC1L1
as the more likely amplification target in NSCLC 
. Although we demonstrated that WHSC1L1
expression was restricted to SqCC and correlated with increased gene dosage, it was not significantly higher in samples with gain/amplification or different between normal and cancerous cells (p
0.12, fold change
1.3), and therefore, also discounted.
BRF2 Contributes to SqCC Tumorigenesis by Regulating Cell Growth and Proliferation through the Increase of Polymerase III Activity
encodes a subunit of a transcription initiation complex responsible for RNA polymerase III (Pol III)-mediated transcription 
. Pol III transcribes a limited set of genes that encode nontranslated RNAs including 5S rRNA, tRNA, 7SL RNA, and U6 RNA, which are essential for protein synthesis and RNA processing 
. Because these processes are fundamental determinates of the capacity of a cell to grow, increased activity of Pol III is often observed during cancer development 
. Indeed, transformed cells express elevated levels of Pol III transcripts, and inhibition of these transcripts limits cell growth and proliferation 
. It has been proposed that deregulation of Pol III in transformed cells can occur through three different mechanisms: release from cellular repressors, direct activation by oncogenes, and overexpression of transcription factors 
. In normal cells where growth is tightly controlled, tumor suppressors including RB, p53, and PTEN repress Pol III transcription 
. Inactivation of these genes or activation of oncogenes such as MYC
reverse this process 
. Interestingly, the majority of these genes are mutated in lung cancer, representing a potential mechanism of increasing Pol III activity, and subsequently, cell growth potential during tumorigenesis. Transcription factors, however, are often the limiting components of Pol III-mediated transcription and elevated levels of these components have been observed in numerous cancer types 
. Recently, the overexpression of another Pol III transcription factor BRF1
has been shown to increase Pol III-mediated transcription, resulting in the transformation of cells in vitro
and tumor formation in vivo 
. A study by Marshall et al. was the first to implicate Pol III deregulation as a causative factor in cancer formation 
; however, no studies have been reported to date of activating mutations in Pol III subunits or associated transcription factors in tumors. Therefore, we hypothesized that the amplification and overexpression of BRF2
may contribute to lung SqCC tumorigenesis by contributing to increased cell growth and proliferation, representing a novel alternative mechanism of increasing Pol III transcription in cancer.
To test this hypothesis (hypothesis 4), we performed complementary loss and gain of function in vitro experiments using lung cancer cell lines and immortalized HBEC lines, respectively. Twenty NSCLC cell lines (16 AC and 4 SqCC) previously analyzed by array CGH were assayed for BRF2
expression by qRT-PCR (Methods
). Mirroring the findings from the clinical tumor specimens, BRF2
expression was strongly correlated with gene dosage with the two cell lines with amplification (HCC95 and H520) displaying the highest transcript levels (Figure S2
; Table S5
). In addition, both these lines were derived from SqCC samples and no AC cell lines contained amplification, re-enforcing the lineage specificity of BRF2
activation. To determine the effect of BRF2
overexpression on BRF2 protein levels, three cell lines were selected for Western blot analysis: a SqCC with amplification (H520), an AC with neutral copy number (H1395), and an AC with loss (H2347) (). Consistent with a role in tumorigenesis, high protein levels were only found in H520.
BRF2 activation contributes to cell growth and proliferation.
To determine if increased BRF2 levels lead to higher Pol III activity, we performed Northern blot analysis to assess the expression of Pol III-mediated transcripts. BRF2 is specifically involved in transcription from type 3 (gene external) Pol III promoters, which are responsible for the expression of snRNA genes, two of the best characterized of which are U6 and 7SK 
. This finding is in contrast to type 1 and type 2 Pol III promoters (gene internal), which require BRF1 for transcription and are involved in the expression of 5S rRNA and tRNA genes, respectively 
. Thus, we hypothesized that BRF2 activation would lead to an increase only in type 3 transcripts and not those regulated by BRF1. As expected, assessment of transcript levels in the lung cancer lines showed drastically higher levels of both U6 and 7SK relative to 5S loading control in H520 cells compared to H1395 and H2347 cells, confirming increased BRF2-dependent Pol III transcription upon BRF2 activation (). Furthermore, U6 levels were decreased upon knockdown of BRF2
expression in H520 cells using an shRNA construct (Figure S3
). These results match those of a recent study, which showed that BRF2 protein levels correlate with U6 promoter activity 
. These data suggest that increased BRF2 levels are sufficient to increase Pol III activity, demonstrating the downstream mechanistic effect of gene amplification.
To assess the functional significance of BRF2
amplification and overexpression on SqCC development, RNAi-mediated knockdown was performed in H520 cells. Expression of two different shRNAs) targeting BRF2
substantially reduced transcript levels () and significantly decreased cell proliferation compared to a negative vector control (). In addition, knockdown of BRF2
expression significantly reduced the ability of these cells to grow in an anchorage-independent manner as measured by colony formation in soft agar (). Similar results on cell proliferation were observed with BRF2
siRNA pool transfection of H520 cells (). In contrast, siRNA knockdown of an AC cell line without BRF2
amplification and overexpression, H1395, did not diminish cell proliferation and resulted in an increase in proliferation relative to transfection with a nontargeting control siRNA pool (). How knockdown of BRF2
could lead to an increase in cell proliferation in this context remains unclear and warrants further investigation. Lastly, to further confirm the specificity of this effect to cell lines with amplification, we also performed knockdown experiments in two SqCC cell lines (HCC15 and HCC2450) without BRF2
amplification (Figure S4
). As expected, no significant decrease in proliferation was seen in HCC15 or HCC2450 upon BRF2 inhibition. These results demonstrate a crucial role for BRF2
in contributing to the sustained cellular proliferation and survival of SqCC tumors with gene activation and highlight its cell type specific oncogenic potential in lung cancer.
To further validate its tumorigenic properties, we performed complimentary experiments by overexpressing BRF2
by stable transduction of immortalized HBEC lines (Figure S5
) and measured cell growth compared to vector-expressing controls. HBEC lines are immortalized without the use of viral oncoproteins, have minimal genetic changes, and do not exhibit a transformed phenotype 
. In addition, since they express epithelial markers and morphology and can differentiate into mature airway cells, they represent an attractive model for testing the importance of specific gene alteration found in the initiation of epithelium-derived lung cancer 
. Strikingly, the introduction of BRF2
alone resulted in a modest but significant increase in cellular growth and saturation density, further supporting a tumorigenic role for this gene (). Furthermore, as p53 is inactivated in ~50% of NSCLC tumors, and is known to repress Pol III-mediated transcription, we sought to investigate the impact of BRF2
overexpression in conjunction with p53 silencing on HBEC growth. Interestingly, the combination of these two alterations enhanced cell growth greater than each alteration alone (p
), suggesting a synergistic role for these alterations in promoting proliferation (). Taken together, our results demonstrate that BRF2
overexpression plays a key role in regulating cell growth and proliferation, confirming the functional significance of BRF2
gene amplification in SqCC.
BRF2 Activation Is an Early Event in SqCC Development
The cell type restricted pattern of activation coupled with its transformation potential strongly implicates BRF2
as a lineage-specific oncogene in lung SqCC. SqCC carcinogenesis is thought to be a multistep process that involves the transformation of normal mucosa though a continuous range of precursor lesions up to CIS before invasive cancer and finally metastasis 
. However, since most studies focus on clinically evident tumors, little is known about the molecular events preceding the development of lung cancer and the underlying basis of carcinogenesis. Unlike low grade dysplastic lung lesions that rarely progress, the majority of CIS cases will become invasive cancer 
. Therefore, we hypothesized that critical alterations necessary for disease progression would be evident in preinvasive CIS lesions and persist in invasive tumors. To determine if BRF2
activation occurs early in SqCC development (hypothesis 5), we analyzed gene dosage in a panel of 20 CIS lesions (Table S1
, sample set 5) obtained by autofluorescence bronchoscopy (Methods
). Remarkably, array CGH revealed BRF2
copy number increases in the majority of CIS cases () with 35% (seven out of 20) demonstrating high-level amplification (log2
ratio >0.8; ). WHSC1L1
were only amplified five times (five out of 20) and once (one out of 20), respectively, further excluding these genes as primary driver genes of the amplicon (). To confirm that amplification results in increased expression of BRF2
in preinvasive lesions, we performed immunohistochemistry (IHC) on a CIS sample (CIS2) with amplification (). As expected, BRF2 expression was elevated in CIS epithelia in this sample in comparison to normal epithelia from the same patient (). Strong BRF2 expression was also observed in additional CIS cases with lower levels in earlier stages of neoplastic progression (mild, moderate, and severe dysplasia) and little or no staining in benign lesions (hyperplasia and metaplasia), confirming that gene activation is an early event in SqCC development (). Interestingly, the only benign lesion in which BRF2 expression was observed was obtained from a patient that had also developed CIS (). The high frequency of activation in preinvasive lesions suggests that BRF2 plays a critical role in the development of SqCC through the increase of cell growth potential. Since patient survival can be significantly improved if the lesions are detected and treated at their preinvasive stage, the identification of genes involved in the development of CIS and invasive SqCC is of vital clinical importance 
. Our finding that BRF2
is a lineage-specific oncogene amplified early in SqCC development, and not expressed in normal lung tissue, represents a critical step in understanding the development of SqCC, and represents a promising target for therapeutic intervention.
Amplification and overexpression of BRF2 in preinvasive SqCC lesions.
BRF2 expression in SqCC precancerous stages.
Increased RNA Processing Is Associated with BRF2 Overexpression
To identify other genes and functions that may be associated with BRF2
-mediated initiation of tumorigenesis, we performed significance analysis of microarrays (SAM) on a panel of 111 NSCLC tumors (Table S1
, sample set 2), followed by gene enrichment analysis using ingenuity pathway assist (IPA) (Methods
). This analysis revealed 86 genes, which were significantly increased (78) or decreased (8) (false discovery rate <5%) in tumors with the highest BRF2
expression (Table S6
). IPA analysis revealed enrichment for genes with diverse biological functions including RNA post-transcriptional modification, gene expression, cell cycle, and cancer (Table S7
). The identification of RNA post-transcriptional modification as the most significantly affected function (p
, the two significance values refer to a range of specific subfunctions) was significant, as this is one of the main roles of Pol III-related transcripts as stated above. The genes related to this function, which are increased in expression, include FBL
, and CPSF3
, and are involved in the modification, polyadenylation, and processing of both mRNA and rRNA. Since these are fundamental processes necessary for proper protein production and therefore cell growth, upregulation of these components may be associated with the increased proliferative capacity of SqCC cells upon BRF2
activation. However, the exact nature of this association is currently unknown and future studies will be needed to understand the mechanism responsible for BRF2-induced cell growth in SqCC.
Interestingly, as shown above, BRF2 activation leads to increased transcription from type 3 Pol III promoters that are involved in the transcription of snRNA genes including U6 and 7SK 
. snRNAs are responsible for a range of regulatory functions, including the alteration of gene expression and a potential role for snRNAs in the genomic instability of cancer that has been proposed 
. In particular, U6 snRNA forms the catalytic core of the spliceosome 
. The spliceosome performs the splicing of precursor mRNA in eukaryotic cells, removing introns and joining exons. This process is tightly regulated during growth and development and aberrant splicing has been linked to numerous human diseases, including cancer 
. In fact, many oncogenes demonstrate alternative splicing patterns associated with neoplasia, and splicing regulatory factor expression levels have been shown to increase during cancer progression. Strikingly, many of the genes we identified as being associated with increased BRF2
expression, including SNRPA
, are known to interact with snRNAs including U6 in the spliceosome complex. In addition, SNAPC5
, which encodes a member of the snRNA-activating complex that is required in conjunction with BRF2 to initiate transcription from snRNA promoters 
, was also found to be increased in samples with high BRF2 expression. Taken together, our data suggest that BRF2-mediated increase of U6 as well as other splicing regulatory factors may contribute to oncogenesis in SqCC with 8p amplification. Future studies of the role BRF2 overexpression plays in spliceosome function will yield insight into this potential function, and its role in the neoplastic transformation of lung epithelium to SqCC.
Association of BRF2 with Clinical-Pathological and Genomic Features
Lastly, to investigate the potential clinical significance of BRF2 activation in patients with lung SqCC and better characterize this subgroup of tumors, we next sought to determine the association between 8p amplification and clinical-pathological and genetic features. For this purpose, we expanded our sample set to include 92 SqCC tumors with well-annotated clinical information that were analyzed by tiling-path array CGH. Overall, increase of BRF2
copy number was found in 43% of the expanded dataset, in concordance with the original frequency of alteration. No associations were found between age, gender, smoking status, or stage and BRF2
amplification in our dataset. Furthermore, no significant associations between the level of BRF2
expression and patient survival were seen in two independent datasets (Figure S6
). However, SqCC tumors with and without BRF2
activation showed a unique genome-wide spectrum of DNA amplifications, suggesting that different genetic pathways may be involved in their development (Table S8
). Assessment of other clinical and genetic features—for example response to therapy and mutation events—will be necessary in the future to further explore the characteristics of SqCC patients harboring BRF2 amplification.