The Daily Sign-Out

Lung adenocarcinoma subtyping: practical issue

We have adopted subtyping adenocarcinoma routinely as part of the examination of lung resections along the lines suggested in the recent "comprehensive histologic subtyping" article by Motoi et al. (Am J Surg Pathol 2008;32:810-827).  A recent case exposed an unexpected issue with such subtyping.  The primary tumor measured 8.0 cm and appeared to show an exclusively solid pattern (6 blocks of tumor submitted):

But the lymph node metastases showed an almost exclusively papillary/acinar pattern:

Final stage was IIIA (T3N2).  The primary tumor is EGFR wild-type but showed the KRAS mutation Gly12Cys.  The case demonstrates a significant morphological discrepancy in histological subtypes between the primary tumor and lymph node metastases.  What is the significance of this?  What does this phenomenon suggest about the relationship between the primary tumor and its metastases?

A key question in lung adenocarcinoma currently is the association between histological subtypes and outcomes and mutational profiles. 

I call your attention to the timely and thoughtful editorial by Dr. Sanja Dacic from UPMC, Pittsburgh published in last month's Archives of Pathology and Laboratory Medicine which touches on the issues and problems concerning adenocarcinoma subtyping.  Dr. Dacic's own excellent study published in Modern Pathology last year showed a large morphologic overlap between tumors with/without KRAS and EGFR mutations.  I found this to be an especially provocative quote:

At this point, there are not enough data to prove an independent prognostic significance of detailed histologic subtyping, and it is uncertain whether histologic subtype of adenocarcinoma is an independent predictor of therapy response.

On the research side, I have been retrospectively annotating our cases in several large databases at Rush University Medical Center.  One critical issue in archival specimens is sampling--there is a wide range in the number of sections taken of tumor.  I think it reasonable (and practical) to entirely submit T1 tumors for histological exam.  But currently, there are no guidelines or protocols (or even suggestions!) for what constitutes adequate sampling to be "representative" of tumors in the T2-T3 categories.  The Motoi study only included 101 cases and included stage I to III cases.  Of note, cases were selected on the basis of available frozen tumor not as part of a consecutive series (!).  The range of H&E stained slides reviewed was 1 to 23 with a average of 6.8.

From the clinician's and patient's points-of-view, adenocarcinoma subtype is meaningful if it provides some prognostic and predictive information.  The data are just not out there right now that subtyping provides this.  So should we bother?  Long enough post for a Friday afternoon--I'll give my answer next week.

WCLC 2011 Oral Presentations: (More) Genomics

Continuing posts summarizing my highlights from the oral presentations from WCLC 2011.

The Biomarkers V presentation highlight (O39.01) was a retrospective analysis of the CALGB 30406 study looking at the impact of specific EGFR mutations on outcomes.  CALGB 30406 was a prospective trial in never/light smokers, previously untreated patients with advanced lung adenocarcinoma randomized to receive erlotinib (ERL) or ERL plus carboplatin and paclitaxel (C/P) as first-line therapy.  Patients with EGFR mutations had similar outcomes (RR, PFS, OS) in both arms of the study vs. EGFR-wild type  patients but no significant difference was seen in outcomes in exon 19 deletion vs. L858R mutation, irregardless of treatment arm.  Good evidence to address a certainly relevant question.

Recent studies have shown that NSCLC can be differentiated into "epithelial-like" and "mesenchymal-like" tumors based on gene expression and EMT biomarkers have been associated with erlotinib activity in both in vitro and clinical specimens.  Increasing evidence also suggests that changes in DNA methylation also are associated with the phenotypic EMT transition in lung cancer cells.  MO16.02, presented in the Biomarkers VII session, is a proof-of-principle study by Genentech researchers using an integrated genomics approach combining gene expression profiling and methylated DNA immunopecipitation (meDIP) to define phenotypic subsets of NSCLC by their DNA methylation patterns.  They further developed quantitative methylation-specific PCR assays for 14 markers identified by meDIP and pyrosequencing assays for another 5 markers.  They tested a panel of NSCLC cell lines with defined responses to erlotinib by these assays and showed good correlation with erlotinib activity or EL/ML status.  This study showed the potential for using DNA methylation profiles in targeted therapy profiles.  This seemed like a lot of fancy razzle-dazzle in this context and I'm unclear what it would really add in a practical sense; it seems to me most of the work on DNA methylation patterns in lung cancer has been focused on identifying markers for early diagnosis in serum.

WCLC 2011 Oral Presentations: Squamous Cell Carcinoma

There were several interesting presentations regarding squamous cell lung carcinoma that are worthy of attention.

O06.02 presented in the Biomarkers I session by Selinger et al. was a study of a possible prognostic marker for SQC called Special AT-rich Binding Protein-1 (SATB1).  SATB1 is a global genome organizer that recruits chromatin remodeling proteins for epigenetic regulation of multiple genes in a tissue-specific manner.  SATB1 is normally expressed in normal respiratory ciliated columnar epithelium in bronchi and terminal bronchioles but the authors observed loss of this protein by IHC expression in preinvasive squamous lesions and in NSCLC.

In their cohort of 243 NSCLC patients, the authors found that loss of SATB1 was an independent predictor of decreased cancer-specific survival in SQC patients (HR 2.06, P=0.016).  The authors suppose that loss of SATB1, which can act as a transcriptional activator or repressor, in early lung lesions results in widespread disruption of tissue-specific epigenetic modifications, leading to activation or repression of key cancer genes and promotion of lung tumorigenesis.

This provocative study raises more questions than it can answer but is worthy of notice because of its identification of a possible prognostic marker for SQC (as opposed to ADC) but also the interesting mechanistic questions it raises with regard to changes in tissue-specific epigenetic regulation for the development of or progression of cancer.

 

O037.07 by Savic et al. was presented in the Biomarkers IV session and adds to the growing number of studies on Sex Determining Region Y-box 2 protein (SOX2) in NSCLC.  SOX2 is located on the long arm of chromosome 3, an area that has been previously identified as being associated with loss of 3q in early stages of SCQ carcinogenesis.  This study involved a tissue microarray constructed from 568 patients with stage 1 NSCLC with detailed histopathological and clinical data.  SOX2 gene data was determined by FISH and SOX2 protein expression was assessed by IHC.  This data was correlated then with clinicopathological findings and OS and cancer-specific survival.

Increased SOX2 gene amplification or polysomy was found in 4% of (17/429) evaluable samples but 16/17 (94%) were associated with non-ADC histology.  SOX2 protein expression was found in 48% of NSCLC and was associated with non-ADC histology (P<0.001): SOX2 expression was identified in 78% of SQC specimens (178/227) but only in 15% of ADC (29/191).  Further, SOX2 protein expression was significantly associated with high grade tumors, patients >50 years, and male gender.  Finally, they noted a significant association between increased SOX2 gene number and protein expression (P<0.001).

While these are some intriguing findings, this presentation doesn’t quite deliver. Unfortunately, the authors did not report any correlation data with survival (and this is a glaring deficiency) nor do they offer any speculation on the discrepancy between the few cases that show increased SOX2 gene copy number and the frequent observation of SOX2 protein expression in SQC.  Further work on SOX2 definitely seems warrented, especially because of this association with SQC histology and its location in a chromosome region frequently lost in early stages of squamous carcinogenesis.

 

WCLC 2011 Oral Presentations: Genomics

There were 3 presentations on genomics from the Biomarkers II session that I thought noteworthy for extending the understanding of genotype driver mutations in lung ADC.

O016.01 is an update from the U.S. Lung Cancer Mutation Consortium (LCMC) on 911 patients registered to date who have undergone comprehensive testing for currently known driver mutations in NSCLC.  KRAS, EGFR, HER2, BRAF, PIK3CA, AKT1, and NRAS have been testing using multiplexed assays, whilst ALK rearrangements and MET amplification have been assessed using FISH (which are reported in a separate abstract from the conference).

Multiplexed assays have been completed in 420 patients and a genetic alteration has been identified in >50%.  All patients had advanced stage (3B or 4) ADC.  The median age of this population is 61 years and 59% are women.  There was no significant difference in the age between those patients with tumors associated with identifiable mutations and those having tumors without mutations.  Neither did gender or smoking history show any significant association with mutation status.

This is very interesting early data from this project that contrasts the clinical observations of driver mutations being associated with female gender and never (or light) smoking history.  This data also supports mutation testing for all patients with lung ADC, at least in advanced stage patients who might be considering erlotinib therapy or enrollment into a clinical trial testing other “targeted” agents.

O016.05 was a presentation of another high profile project, the Lung Cancer Mutation Analysis Project (LC-MAP) from Memorial Sloan Kettering Cancer Center.  The authors presented their 2-year results of comprehesive mutation analysis of 1231 NSCLC (ADC=896, other non-SQC=49).  Evidence of driver mutations was found in 56% of tumors as follows: KRAS 30%, EGFR 20%, ALK-EML4 7%, HER2 5%, PIK3CA 1%, BRAF 1%, MEK1 0.2%, AKT1 0.1%.

This study provides some preliminary benchmarking data to discuss with clinicians, although I would provide a caveat that there is likely to be a referral bias (since this is a single-institutional study from one of the world’s most reknown cancer centers) that overestimates the frequency of driver mutations in ADC.

Finally, O016.06 is another presentation from the MSKCC that examined the prognostic impact of driver mutations in lung ADC with respect to smoking history.  There is important background that puts this study in context, namely, the demonstration of an independent dose-dependent relationship between smoking history and survival in patients with advanced stage NSCLC with never-smokers living 50% longer than smokers (Janjigian et al., Cancer 2010).  The authors reviewed 301 consecutive never smokers and 373 consecutive smokers with lung ADC who underwent testing for EGFR and KRAS mutations and ALK rearrangements between May 2009 and May 2010.

They found statistically significant differences in the frequency of these mutations between never smokers and smokers: KRAS mutations in 4% of never smokers vs. 43% smokers (P<0.0001); EGFR mutations in 37% of never smokers vs. 14% smokers (P<0.0001); ALK rearrangements in 12% of never smokers vs. 2% smokers (P<0.0001).  By multivariate analysis, there were significant survival differences between never smokers with EGFR mutations vs. KRAS mutations (HR for death 4.2, favoring EGFR mutations) and between never smokers with EGFR mutations vs. ALK rearrangements (HR for death 0.31, favoring ALK rearrangements).  But, within a given genotype, there were no significant survival differences between never smokers vs. smokers.

This is very interesting data suggesting that the prognostic impact of smoking is driven by the type of mutation (at least in those patients in whose tumor has a detectable driver mutation).  Integration of smoking history with the type of mutation detected should refine estimation of the overall prognosis of patients with lung ADC.

 

WCLC 2011 Oral Presentations: Tumor microenvironment

So I'm finally finished sorting through the abstracts from the 14th World Conference on Lung Cancer that I recently attended and this begins a series of posts of what I found of interest from the oral abstract presentations.

There were two presentations that focused on the immune cell response in the tumor microenvironment that were of particular interest to me in that I have a current project looking at CD3+ lymphocyte responses in resected NSCLC through my ongoing work with the Thoracic Oncology Research Group at Rush University Medical Center.

MO 22.01 from session VIII examined tissue microarrays constructed from 479 patients with stage 1 adenocarcinoma and a median follow-up of 4.3 years.  They looked at a wide variety of different immune markers by immunohistochemistry and assessed localization (tumor vs. stromal) of immune cell infiltration, the grade of that immune cell infiltrate (low vs. high), and expression (sum of intensity of staining plus semi-quantitative assessment of distribution).

Results: The 5-yr relapse free survival (RFS) was 82% for the entire cohort.  There was no independent prognostic significance for any subpopulation of immune cells.  However, the ratio between stromal FOXP3+ (Treg cells) and CD3+ (total T cells) and tumor expression of IL7R and IL12Rbeta2 were found to be significant factors in predicting recurrence by univariate analysis.  Tumor expression of IL12Rbeta2 and stromal FOXP3:CD3 were prognostic even in tumors less than 2 cm.  By multivariate analysis, gender, stage (1A/1B), lymphatic invasion, tumor IL12Rbeta2 and stromal FOXP3:CD3 were independent predictors associated with recurrence.

This study highlights the importance of localization of immune cell response but also underscores the challenge for future studies to use more refined quantitative assessments of immune cell subpopulations and relative expression of IHC staining intesity (such as utilizing digital imaging and image analysis). 

MO 22.04, also from session VIII, presented interesting complementary data to the above presentation. This study examined 196 resected stage 1-3A NSCLC using IHC for infiltrating CD8+ and FOXP3+ cells.  They counted cells expressing these markers in tumor and stromal regions in five "randomly selected" high-power-fields and reported their data as the number of cells per mm² for each region.  They used median cell count as the cutoff to define patient subgroups.

Results:  Increased tumor islet CD8+ and stromal FOXP3+ infiltration were associated with better prognosis, whilst increased stromal CD8+ and tumor islet FOXP3+ infiltration were associated with worse prognosis.  Further, high CD8+ tumor islet/stromal infiltration was associated with better prognosis, whilst high FOXP3+ tumor islet/stromal infiltration was associated with worse prognosis.

Again, this study reinforces the crucial importance of identifying the microlocalization of immune cell subpopulations.  But I note several critical issues with methodology in this study that are worth considering for future studies.  First, the investigators did not clarify their methodology as to how high-power-fields were "randomly selected."  Further, their use of the term "tumor islet" suggests a non-random selection of such foci (tumor budding?) for assessment.  As a pathologist, I can safely say that this is quite unlikely "random"--which would undoubtedly lead to bias towards areas of the tumor that show higher concentrations of positively staining cells.  In fact, our Rush group recently began re-evaluating EMT markers assessed by IHC in a group of NSCLC patients who underwent surgical resection to specifically assess and grade distribution and extent of staining at the tumor-stromal interface rather than the global expression of such staining.  The data are being analyzed currently.  Moreover, I have been assessing the lymphocytic response just by H&E and note an association between foci of tumor budding and intensity of lymphocytic response at the tumor invasive front.  Second, manual counting of "events" just doesn't make it for this type of study and (again!) I must reiterate the need to use image analysis for this type of study.

ASCO issues Provisional Clinical Opinion on EGFR testing in Non-Small-Cell Lung Cancer

Hot off the electronic press is the PCO issued by ASCO recommending that patients with advanced stage NSCLC should have their tumors tested for EGFR mutation if they are being considered candidates for EGFR-TKI therapy.  Read here  for the abstract or Download ASCO EGFR testing in NSCLC for more deets.

Provisional Clinical Opinion
On the basis of the results of five phase III randomized controlled trials, patients with
NSCLC who are being considered for first-line therapy with an EGFR TKI (patients who have
not previously received chemotherapy or an EGFR TKI) should have their tumor tested for
EGFR mutations to determine whether an EGFR TKI or chemotherapy is the appropriate
first-line therapy.

If you haven't been following this topic, the paper nicely summarizes the clinical trials forming the basis for this recommendation.

Of especial interest to pathologists is the issue of adequate tissue for testing in patients with advanced stage disease:

the ad hoc panel commented that there is often a limited amount of tissue for mutation testing and, therefore, endorses a relevant recommendation in ASCO’s stage IV NSCLC guideline update: “In order to obtain tissue formore accurate histologic classification or for investigational purposes, theUpdate Committee supports reasonable efforts to obtain more tissue than what is contained in a routine cytology specimen.”  Properly fixed material from cytology cell block preparations is generally required for analysis, as opposed to cytology smear preparations.

Finally, note that this PCO only addresses mutation analysis for activating mutations.  Giving recent literature showing concommitant EGFR T790M resistance mutation in 30%-35% of patients at the time of diagnosis, I think we should take this into consideration when looking at testing pathways.  This PCO specifically does not recommend EGFR testing by FISH or IHC at this time.  By the way, testing issues are discussed in the companion journal, Journal of Oncology Practice.  I'll hunt down the link and post later.

Aberrant Wnt1 and beta-catenin expression in NSCLC

This month's April 2011 Journal of Thoracic Oncology has an interesting paper by Xu and colleagues (abstract) that examines Wnt1 expression in non-small cell lung cancer (NSCLC) in relation to downstream Wnt signaling molecules, including beta-catenin, and correlates different marker expression with traditional clinicopathological parameters.

This is an immunohistochemical study of a tissue microarray composed of 262 NSCLC resected specimens.    The authors define aberrant beta-catenin expression as: 1) decreased membranous pattern in less than 70% of tumor cells, 2) cytoplasmic pattern of expression, or 3) nuclear pattern of expression.  As expected, the majority of patients are stage 1 or 2 but nearly 36% of the study population are never-smokers--over 50% in the adenocarcinoma histology group (!).  The authors found that Wnt1 overexpression was significantly associated with never-smoker status and ADC histology; aberrant beta-catenin expression was associated with SQC histology.  While each Wnt1 overexpression and aberrant beta-catenin expression were independently associated with decreased OS regardless of TNM stage, patients that demonstrated both had worse OS than other combinations of Wnt1/beta-catenin expression.  Further, Wnt1/beta-catenin expression was also an independent factor adjusted by stage, pleural invasion, lymphatic invasion, ECOG PS or stage, age, gender, histology, and p53 expression.

A few comments about this paper.  One significant caveat to the findings is the over-representation of never smokers (35%) in this Asian study population, especially nearly 54% of never smokers in the ADC group, compared to a more typical American population in a community hospital setting (about 10%).  Overexpression of Wnt1 was significantly associated with never-smoking and as well as ADC histology, but, somewhat paradoxically, Wnt1 overexpression was associated with significantly shorter overall survival.  Unfortunately, the authors do not break this further down into Wnt1 overexpression and OS with respect to histological type--perhaps some SQCs showing Wnt1 overexpression and dismal OS are skewing the data here since one would expect never-smokers with ADC to have a better prognosis in general.  Another is the usual issues with using TMAs to represent complex heterogenous tumors.  The authors do not sufficiently describe from where they target obtaining cores for constructing the TMA--are they targeting the invasive margin, the central tumor, wherever?  what about mixed histology tumors?  These are critical methodological issues--especially when extrapolating the data concerning marker expression.  One admirable feature of this paper is the clear definition, supported by appropriate photomicrographs, of "aberrant" beta-catenin expression.  As I am currently working on projects that are similarly assessing beta-catenin expression in NSCLC, I found this very helpful in assessing this paper's findings.

Optimal panel for differentiating NSCLC in small specimens

There has been a plethora of papers, abstracts, and marketing materials in the last 12 to 18 months regarding the "optimal" panel for differentiating adenocarcinoma (ADC) from squamous cell carcinoma (SQC) in non-small cell lung cancer (NSCLC).  The December 2010 American Journal of Surgical Pathology features a paper from Terry, Leung, Laskin et al. that evaluates an optimal panel and, perhaps, more importantly, a testing algorithm for differentiating ADC from SQC in small biopsy specimens.

Caveat: one of the authors is the Medical Director/Chief Pathologist for a commercial pathology reference lab specializing in IHC diagnostics (PhenoPath Laboratories).

This study makes several practical points that can be nicely summarized as bullets:

• For ADC the most sensitive marker and best negative predictor is CK7: 93% sensitivity, 91% NPV.
• For ADC the most specific marker and best positive predictor is Napsin A: 94% specificity, 90% PPV.
• TTF-1 has similar specificity as Napsin A for ADC.
• For SQC the most sensitive marker and best negative predictor is p63: 84% sensitivity, 86% NPV.
• For SQC the most spepcific marker and best positive predictor is NTKR2 (a neurotrophic tyrosine kinase receptor, rarely used outside of research labs) but CK5/6 has a nearly identical specificity and PPV: about 95% specificity, about 95% PPV.
• No single marker is BOTH highly sensitive and specific for either ADC or SQC.

The authors identified a 6-marker panel that includes p63, TTF-1, CK5/6, CK7, Napsin A, and mucicarmine has demonstrating the best AUC in ROC analysis.  Since a 6-marker panel may not be feasible (or even possible) due to various factors with small specimens with limited diagnostic tumor material, the authors develped a binary scheme using a sequential addition of markers (Fig. 3, 1809).

The authors do point out, however, that the "best" sequential order may vary by laboratory according to the actual senstitivities and specificities acheived in individual labs and this is important to remember if your lab does not perform all of the stains.

I think one important caution, however, is to note that this study was performed using tissue microarrays derived from resected lung cancer specimens and did not study or validate the performance of these markers on "real" small biopsy specimens.  The authors claim that the results are "similar" to earlier studies using actual patient biopsy material.  Given their careful statement above regarding the "best" sequential order, I think that this is somewhat disingenuous elision over important details regarding significant differences between resected tumors and small biopsy specimens.  Notwithstanding this major issue, there are some good quantitative data here that warrants your consideration in evaluating your own panels at least for comparing your own sensitivity and specificity for individual markers.

A Consensus Statement of EGFR Mutation Testing in NSCLC

The International Association for the Study of Lung Cancer (IASLC) and European Thoracic Oncology Platform recently published a consensus statement regarding EGFR mutation testng in NSCLC.  There was agreement that the decision to request EGFR mutation testing should be made by the treating physician.  The worskshop group underscored that the most important issue is cooperation and communication between the various disciplines with regard to tumor specimen handling.  The group did not reach a consensus on a preferred methodology but did stress the need for good laboratory practice with regard to validation and standardization of whatever methodology is used.  You can read more in the October 2010 Journal of Thoracic Oncology.

This is interesting because of the fact that a number of institutions already are routinely referring NSCLC tumors for molecular testing, usually incorporating a panel that includes EGFR, KRAS, and ALK-EML4.  Moreover, routine EGFR mutation testing for NSCLC is being aggessively marketing by various companies.  I suppose that these "consensus" recommendations do not truly reflect what is already happening in practice, at least at "academic" medical centers in the U.S.  Perhaps someone in the U.K. or Europe might comment.

Association between EGFR mutations and ERCC1 expression in NSCLC

Gandara and colleagues from UC Davis Cancer Center have published ahead-of-print on October 21, 2010 in Journal of Thoracic Oncology an interesting article examining the association between EGFR activating mutations and ERCC1 gene expression in NSCLC.

NSCLC tumors that show EGFR activating mutations are also more likely to show low expression of ERCC1 mRNA levels according to this study.  This might explain (at least in part) the clinical observation that suggest enhanced efficacy of platinum-based chemotherapy in patients with EGFR-mutant NSCLCs.

The authors performed microdissection of tumors from 1207 patients with NSCLC and analyzed EGFR mutation by allele-specific by PCR and ERCC1 mRNA expression by RT-PCR.

Median ERCC1 expression was histology-related: adenocarcinoma (ADC) showed a lower median expression (1.68 (0.22-11.31) versus squamous cell carcinoma (SQC) (2.42 (0.51-14.28) (p<0.001).  The presence of EGFR mutations was highly associated with low ERCC1 expression (p<0.001).

It would be interesting if future or additional work focused specifically on ADC as a subgroup and perhaps the authors might consider re-analyzing their data and presenting this data.  I suggest this as a practical concern since we are not routinely testing SQC for EGFR mutation.  In addition, the next step is to examine the relationship between EGFR mutation and low versus high ERCC1 expression and response to combined EGFR-TKI and platinum-based chemotherapy, especially in advanced stage disease.