[2111] MicroRNA Profiling Complements Microscopic/Mutational Analysis by Defining Alternative

Pathways of Carcinogenesis

Joseph F Annunziata, Anna Banizs, Christina M Narick, Sara Jackson, Jan F Silverman, Sydney
Finkelstein. Allegheny General Hospital, Pittsburgh, PA; Interpace Diagnostics, Pittsburgh, PA

Background: Thyroid follicular cell neoplasia is biologically heterogeneous ranging from benign to
indolent malignancy to high grade cancer. Mutation genotyping is effective at predicting tumor
aggressiveness. Aggressive disease can be seen in a significant minority of cases where common driver
mutations are undetected. We performed a correlative analysis of microRNA profiling with specific
patterns of thyroid follicular cell nodules using a large clinical case database (n=5,210) seeking
discriminating information.
Design: Data was retrospectively mined from a large clinical testing cohort of FNA thyroid nodule
cytology. Cytology was based on pathologist Bethesda diagnostic classification. Mutation analysis
targeted common mutations (BRAF, ras, PIK3CA, PAX8/PPAR and RET/PTC translocations) by next
generation (Illumina). RNA expression classifier utilized a 10 miRNA panel consisting of 5 miRs with
increased expression in cancer and 5 miRs with hyperexpression in benign states.
Results: Mutation analysis defined six groups for correlative miR profiling: BRAF+ (aggressive), N/H/K
ras+ (indolent to aggressive), PAX8/PPARgamma translocation+ (indolent to aggressive) and No
detectable mutation yet aggressive. BRAF+ compared to ras+ groups showed distinct profiling
differences for 7 of 10 individual miRs. No detectable mutation/aggressive subset closely matched
BRAF+ supporting common molecular pathway involvement (BRAF phenotype). Among the different
ras+ nodules, NRAS and HRAS showed a nearly matching miR profile for all 10 miRs. KRAS profiling
differences were detected for 5 miRs in keeping with potentially differing biology of this form of ras
mutation. PAX8/PPARgamma displayed profiling differences for 8 miR distinct from both BRAF and ras
groups. Cytology diagnoses were significant for B5 and B6 in BRAF+ cases but otherwise molecular
patterns were not predicted by microscopic classification.
Conclusions: Oncogene driver mutations can predict biological aggressiveness including benign versus
malignant status in a proportion of thyroid follicular nodules that are indeterminate after cytology
evaluation. BRAF mutated nodules display a unique miR expression profile distinct from other common
mutations. This BRAF profile is present in mutation negative aggressive disease suggesting BRAF
mutational heterogeneity across the nodule with sampling variation. An approach combining RNA
expression profiling to define distinct molecular pathways and mutational genotype is well suited to
complement microscopic assessment of thyroid neoplasia.
Category: Techniques (including Ultrastructure)

Session: Poster Session II #303, Monday Afternoon

Date/Time: Monday, March 6, 2017 - 1:00 pm
Room: Exhibit Hall 1
[1285] Clinical and Biological Significance of Common Mutational Genotypes of Thyroid Follicular
Neoplasia

Joseph F Annunziata, Anna Banizs, Christina M Narick, Sara Jackson, Jan F Silverman, Sydney
Finkelstein. Allegheny General Hospital, Pittsburgh, PA; Interpace Diagnostics, Pittsburgh, PA

Background: Somatically acquired oncogene mutational change is a fundamental to thyroid neoplasia.

While certain mutations confer highly predictable degrees of biological aggressiveness, other mutation
types are less predictable. We report on our large experience thyroid nodule mutational change
correlated with cytology and RNA expression level determination to better understand neoplastic
progression according to specific mutation type.
Design: 5,210 indeterminate thyroid nodule needle aspirates underwent mutational analysis for
common mutations (BRAF, HRAS, KRAS, NRAS, PIK3CA, PAX8/PPAR and RET/PTC translocation) by next
generation sequencing (Illumina). Cytology diagnosis was based on pathology report diagnoses
(Bethesda diagnostic categories). As an independent predictor of neoplastic behavior, a panel of 10
microRNAs was developed. Mutation type was correlated with cytology diagnosis and microRNA
prediction of clinical aggressiveness.
Results: Oncogenic mutations were detected in 1100/5,210 (21.1%) of the cases. The majority were
point mutations (95.3%) and the remainder were translocations (4.7%). Virtually all nodules displayed a
single genotype for this mutation panel. RAS genes accounted for most point mutations (n=736)
distributed as NRAS (57.0%), HRAS (26.7%) and KRAS (16.3%). Mutations were seen in across cytology
categories, the relative distribution varied according to specific genotype. B-V was dominated by BRAF
V600E (40.9%) but included all genotypes. B-III and B-IV manifested all genotypes with relatively greater
content of ras gene mutations (16.7% and 15.6 %) and all BRAF point mutations outside of V600E.
MicroRNA (miR) expression classifier yielded a four level quantitative measure of increasing malignancy
risk (very low 99+% NPV, low 94% NPV, moderate 74% PPV, high 99+% PPV). Individual mutational
genotypes within specific oncogenes displayed unique malignancy risk profiles correlating with cytology.
Conclusions: Cytology prediction of malignancy risk is challenging in the earlier stages of neoplastic
progression. RAS gene mutations, the most common oncogene alteration, are diverse and not
equivalent. While certain mutations are highly predictive of cancer, the majority of individual mutation
genotypes show biological heterogeneity. Combining RNA expression profiling with mutation
determination leads to more information to base risk assessment.
Category: Head and Neck Pathology

Session: Poster Session IV #173, Tuesday Afternoon

Date/Time: Tuesday, March 7, 2017 - 1:00 pm
Room: Exhibit Hall 1
[1822] Biological Aggressiveness of Thyroid Neoplasia Depends on the Strength of the Genetic
Mutation and Associated Cellular Interaction

Anna Banizs, Joseph F Annunziata, Christina M Narick, Sara Jackson, Jan F Silverman, Sydney
Finkelstein. Allegheny General Hospital, Pittsburgh, PA; Interpace Diagnostics, Pittsburgh, PA

Background: Neoplasia of the thyroid can harbor a variety of genetic mutations displaying a range of
biological aggressiveness across the benign to malignant spectrum. To better understand the
mechanistic basis for this differential aggressiveness we analyzed a large database of thyroid nodule
aspirates (n=3341) by cytology, mutational and microRNA (miRNA) classifier analysis.
Design: Cytology diagnosis was based on cytology reports submitted using Bethesda Diagnostic
Categories (BDC- I to VI). Separate needle passes were used for molecular testing. Mutational analysis
encompassed common mutations (BRAF, RAS, PIK3CA, PAX8/PPARg and RET/PTC translocations) on next
generation sequencing (NGS) platform (Illumina). Variant sequence content (%) on NGS was quantified
for each detectable oncogene point mutation. The miRNA classifier utilized a 10 miRNA panel trained on
257 thyroid reactive, benign, malignant. Diagnostic modalities underwent statistical comparison using
ANOVA on Ranks.
Results: miRNA classifier results yielded a quantitative measure across the benign/malignant continuum
assigned to four cancer risk categories: very low (n=830) with 99+% NPV, low (n=1962) with 94% NPV,
moderate (n=372) with 74% PPV and high (n=177) with 99+% PPV. NGS mutational analysis, using the
miRNA classifier as the gold standard, showed clear differences between strong (BRAFV600E),
intermediate (i.e. other BRAFs) and weak (RAS) driver mutations. Differential properties were also
demonstrated within the RAS gene, with NRAS (n=279) and HRAS (n=125) statistically stronger than
KRAS (n=79, p <0.05). Mutation variant sequence percentage varied from non-detectable to 100%.
Strong driver mutations tended to be clonally expanded and associated with higher risk miRNA classifier
status even with low percentage of mutated copy (<5%).
Conclusions: Acquired mutations in early stages of follicular neoplastic progression are heterogeneously
distributed and clonally expanded to varying degree. Classifier status is causally related to mutation type
however each nodule appears to consist of a mixture of mutated and non-mutated cells exhibiting
similar cellular morphology. Strong driver mutations represented in very low copy numbers appear able
to recruit non-mutated cells more effectively compared to weak mutations supporting a critical role for
intracellular communication likely mediated through exosomes.
Category: Pathobiology (including Pan-genomic/Pan-proteomic approaches to cancer)

Session: Poster Session IV #277, Tuesday Afternoon

Date/Time: Tuesday, March 7, 2017 - 1:00 pm
Room: Exhibit Hall 1
[339] The Role of miRNA Expression Analysis in Needle Aspirate Cytology of Indeterminate Thyroid
Nodules

Anna Banizs, Joseph F Annunziata, Christina M Narick, Sara Jackson, Jan F Silverman, Sydney
Finkelstein. Allegheny General Hospital, Pittsburgh, PA; Interpace Diagnostics, Pittsburgh, PA

Background: Fine needle aspirate indeterminate cytologic diagnosis of thyroid nodule is a management
challenge. Three approaches have been used: 1) cytology (Bethesda Diagnostic Categories BDC-I to VI),
2) search for accumulated mutational change and 3) RNA expression classifier status based on panels of
messengerRNA (mRNA)/microRNA (miRNA). Using a large database (n=3341), we compared each
modality for diagnostic performance and to better understand their role in cancer risk assessment.
Design: Cytology diagnosis was based on submitted cytology reports sorted into BDC-I to VI. Separate
needle passes were used for molecular testing on two distinct platforms. Mutational analysis
encompassed common mutations (BRAF, RAS, PIK3CA, PAX8/PPAR and RET/PTC translocations) on next
generation sequencing (NGS; Illumina). Classifier utilized a 10 miRNA panel trained on 257 thyroid
reactive, benign, malignant specimens. Three modalities underwent statistical comparison using ANOVA
on Ranks and Rank Sum Test.
Results: miRNA classifier yielded a quantitative measure across the benign/malignant continuum
assigned to four cancer risk categories: very low (n=830) with 99+% NPV, low (n=1962) with 94% NPV,
moderate (n=372) with 74% PPV and high (n=177) with 99+% PPV. BDC-V (n=119) showed significant
differences in malignant prediction by miRNA classifier (p<0.05) compared to BDC-III (n=1734) and BDC-
IV (n=637). Cytology classification showed no statistical difference in benign vs. malignant prediction
between BDC III and IV (p =0.138). BRAFV600E point mutation was present 31.93% (BDC V) vs 2.88%
(BDC III) and 3.92% (BDC IV). Mutation analysis, using miRNA classifier as gold standard, revealed clear
differences between specific mutational genotypes with strong driver (BRAFV600E) and weak drivers
(RAS and others). Presence of multiple mutations within one specimen was rare, representing 1.27% of
total mutated cases.
Conclusions: Cytology can separate BDC IV and V; however, it appears to be less discriminating in earlier
neoplasia stages (BDC III and IV). miRNA classifiers demonstrate the ability to assess risk in earlier stages
of neoplastic changes. Therefore, combined diagnostic approach (cytology/mutation/classifier) predicts
more reliably the neoplastic potential of thyroid nodules. Oncogene mutation types, while diverse, tend
to be mutually exclusive conferring varying degrees of driver mutation strength.
Category: Cytopathology

Session: Poster Session VI #61, Wednesday Afternoon

Date/Time: Wednesday, March 8, 2017 - 1:00 pm
Room: Exhibit Hall 1