distinct biological types of ocular adnexal sebaceous ...purpose: ocular adnexal (oa) sebaceous...

Translational Cancer Mechanisms and Therapy

Distinct Biological Types of Ocular AdnexalSebaceous Carcinoma: HPV-Driven and Virus-Negative Tumors Arise through NonoverlappingMolecular-Genetic AlterationsMichael T. Tetzlaff1,2, Jonathan L. Curry1,3, Jing Ning4, Oded Sagiv5, Thomas L. Kandl5,Bo Peng4, Diana Bell1, Mark Routbort6, Courtney W. Hudgens2, Doina Ivan1,3,Tae-Boom Kim4, Ken Chen4, Agda Karina Eterovic7,8, Kenna Shaw7, Victor G. Prieto1,3,Anna Yemelyanova1, and Bita Esmaeli5

Abstract

Purpose: Ocular adnexal (OA) sebaceous carcinoma is anaggressive malignancy of the eyelid and ocular adnexa thatfrequently recurs and metastasizes, and effective therapiesbeyond surgical excision are lacking. There remains a criticalneed to define the molecular-genetic drivers of the disease tounderstand carcinomagenesis and progression and to devisenovel treatment strategies.

ExperimentalDesign:Wepresent next-generation sequenc-ing of a targeted panel of cancer-associated genes in 42 andwhole transcriptome RNA sequencing from eight OA seba-ceous carcinomas from 29 patients.

Results: We delineate two potentially distinct molecular-genetic subtypes of OA sebaceous carcinoma. The first isdefined by somatic mutations impacting TP53 and/or RB1[20/29 (70%) patients, including 10 patients whose primary

tumors contained coexisting TP53 and RB1 mutations] withfrequent concomitant mutations affecting NOTCH genes.These tumors arise in older patients and show frequentlocal recurrence. The second subtype [9/29 (31%) patients]lacks mutations affecting TP53, RB1, or NOTCH familymembers, but in 44% (4/9) of these tumors, RNA sequenc-ing and in situ hybridization studies confirm transcription-ally active high-risk human papillomavirus. These tumorsarise in younger patients and have not shown localrecurrence.

Conclusions: Together, our findings establish a potentialmolecular-genetic framework by which to understand thedevelopment and progression of OA sebaceous carcinomaand provide keymolecular-genetic insights to direct the designof novel therapeutic interventions.

IntroductionOcular adnexal (OA) sebaceous carcinoma is an aggressive

cancer that accounts for �5% of malignant epithelial eyelidtumors (1–5). Surgical excision remains the principal treat-ment modality. However, OA sebaceous carcinoma has a highpropensity for multifocal intraepithelial and locally infiltra-tive growth that each contribute to frequent local recurrence.Given the tumor's delicate anatomic location on the surfaceof the eye or eyelid, these characteristics make this a chal-lenging tumor to treat (1, 6, 7). Aggressive surgery is oftenrequired but may produce appreciable functional and aes-thetic morbidity, and orbital exenteration is necessary toachieve local control of disease in 13% to 23% of patients(5, 7, 8). Further, regional nodal or distant metastasis occursin 8% to 22% of patients, and up to 22% of patientsdiagnosed with OA sebaceous carcinoma die of the disease(7–9). Nevertheless, systemic therapies for OA sebaceous car-cinoma remain largely ineffective (10). Collectively, theseproperties underscore a critical need to define the completeset of molecular-genetic alterations driving the developmentand progression of OA sebaceous carcinoma to possiblyimprove patient outcomes through the application of rationallydesigned therapeutic strategies in patients with metastatic

1Department of Pathology, The University of Texas MDAnderson Cancer Center,Houston, Texas. 2Department of Translational and Molecular Pathology, TheUniversity of Texas MD Anderson Cancer Center, Houston, Texas. 3Departmentof Dermatology, The University of Texas MD Anderson Cancer Center, Houston,Texas. 4Department of Bioinformatics and Computational Biology, The Univer-sity of Texas MD Anderson Cancer Center, Houston, Texas. 5Orbital Oncologyand Ophthalmic Plastic Surgery, Department of Plastic Surgery, The Universityof Texas MD Anderson Cancer Center, Houston, Texas. 6Department of Hema-topathology, The University of Texas MD Anderson Cancer Center, Houston,Texas. 7Sheikh Khalifa Bin Zayed Al Nahyan Institute for Personalized CancerTherapy, The University of Texas MD Anderson Cancer Center, Houston, Texas.8Department of Systems Biology, The University of Texas MD Anderson CancerCenter, Houston, Texas.

Note: Supplementary data for this article are available at Clinical CancerResearch Online (http://clincancerres.aacrjournals.org/).

Corresponding Authors: Michael T. Tetzlaff, University of Texas MD AndersonCancer Center, Department of Pathology and Translational and MolecularPathology, 1515 Holcombe Blvd, Unit 0085, Houston, TX 77030. Phone:7137922585; Fax: 7137450778; E-mail: [email protected]; and BitaEsmaeli, MD, FACS Orbital Oncology and Ophthalmic Plastic Surgery, Depart-ment of Plastic Surgery, The University of Texas MD Anderson Cancer Center1515 Holcombe Blvd, Unit 1488, Houston, Texas 77030. Phone: 713-792-4457;Fax: 713-794-4662; E-mail: [email protected]

doi: 10.1158/1078-0432.CCR-18-1688

�2018 American Association for Cancer Research.

ClinicalCancerResearch

Clin Cancer Res; 25(4) February 15, 20191280

on June 21, 2020. © 2019 American Association for Cancer Research. clincancerres.aacrjournals.org Downloaded from

Published OnlineFirst November 12, 2018; DOI: 10.1158/1078-0432.CCR-18-1688

http://crossmark.crossref.org/dialog/?doi=10.1158/1078-0432.CCR-18-1688&domain=pdf&date_stamp=2019-2-4

http://clincancerres.aacrjournals.org/

sebaceous carcinoma for whom effective drug treatments arecurrently unavailable.

We and others previously reported that a subset of OA seba-ceous carcinomas contain somatic mutations (11–13) and dif-ferentially expressed miRNAs (14) that culminate in commonactivation of the PI3K pathway, suggesting that PI3K inhibitorsmay be effective in the management of advanced disease (13).However, further mechanistic studies are needed to delineateadditional pathways/mechanisms driving OA sebaceous carcino-ma and identify additional candidates to exploit in targeted orimmune-based therapies to reduce the morbidity and mortalityrelated to this aggressive cancer.

One important risk factor for OA sebaceous carcinoma is long-standing immunosuppression, including that related to solidorgan transplant (15–17) and HIV infection or AIDS (18). Theseassociations not only emphasize the importance of a competentimmune system in the pathogenesis of OA sebaceous carcinoma,but also suggest a potential etiologic relationship to an infectiousagent. The clinical evolution of OA sebaceous carcinoma in thesetting of long-standing conjunctivitis or blepharitis furtherimplicates an infectious origin in some cases.

In this study,we report next-generation sequencingof a targetedpanel of cancer-associated genes in specimens from 42 OA seba-ceous carcinomas (31 primary or locally recurrent tumors from29patients and 11 matched metastases lesions from 8 patients) andwhole transcriptome RNA sequencing on eight of these tumors.We demonstrate nonoverlapping, potentially distinct molecular-genetic subtypes of primary OA sebaceous carcinoma. The firstsubtype (20/29; 69% patients) harborsmutations in TP53 and/orRB1 with concomitantNOTCH family member mutations, arisesin older patients, shows higher nuclear grade and has a greatertendency for local recurrence [all locally recurrent tumors (n¼ 5)harbored mutations in both TP53 and RB1]. The second subtype(9/29; 31% patients) arises in younger patients, shows lower/intermediate nuclear grade, and lacks mutations in TP53, RB1, or

NOTCH. Instead, these tumors harbor transcriptionally activehigh-risk human papillomavirus (HPV) in 44% of cases (four ofnine patient tumors). Together, our findings provide a potentialmolecular-genetic framework by which to understand the path-ogenesis of OA sebaceous carcinoma and further suggest novelopportunities for therapy and/or prevention.

Materials and MethodsTissue samples

The study was performed with an approved protocol fromthe Institutional Review Board (IRB) of The University of TexasMD Anderson Cancer Center (UTMDACC). All aspects of ourresearch were performed in accordance with recognized ethicalguidelines (e.g., Declaration of Helsinki, Belmont Report).Only left-over archival formalin-fixed, paraffin-embedded(FFPE) tissue (beyond that required for routine patient-care)was utilized for this study. We identified patients with OAsebaceous carcinoma diagnosed at our institution during 2007to 2017 for whom sufficient tissue was available for molecularstudies. FFPE tissue blocks and hematoxylin and eosin–stainedsections were retrieved and reviewed. For each tissue sample,the diagnosis of OA sebaceous carcinoma and adequacy of thespecimen for molecular testing were confirmed by at least threepathologists (MTT, DB, DI, VGP, and/or JLC). The diagnosis ofOA sebaceous carcinoma relied on a combination of histo-pathologic assessment for features typical of OA sebaceouscarcinoma (i.e., intracytoplasmic vacuoles that impinge nuclearcontours and intraepithelial extension) together with immu-nohistochemical confirmation of sebaceous differentiation(i.e., Adipophilin positivity) in the tumor cells in most cases(19, 20). We obtained samples from 31 primary and/or locallyrecurrent OA sebaceous carcinomas from 29 patients and 11matched metastatic tumors from eight of those patients. Foreach patient, normal tissue DNA from an uninvolved lymphnode or adjacent skin/conjunctiva was also selected andsequenced to control for germline polymorphisms. Histopath-ologic features (including nuclear grade/differentiation, necro-sis (presence/absence) and intraepithelial spread (presence/absence) were also recorded for each tumor. Histopathologicgrade was defined according to the nuclear morphology andcytoplasmic evidence of sebaceous differentiation and deter-mined according to the predominance (>50%) of the tumor cellpopulation. High-grade tumors exhibited enlarged irregularnuclei with high nuclear:cytoplasmic ratio and only focalevidence sebaceous differentiation (intracytoplasmic vacuolesindenting the nucleus). Intermediate grade tumors exhibitedsimilar enlarged nuclei but with lower nuclear:cytoplasmic ratioand more obvious evidence of sebaceous differentiation. Finally,well-differentiated tumors showed only slightly enlarged nucleiand frank evidence of sebaceous differentiation.

DNA extractionFor each primary, locally recurrent, or metastatic tumor includ-

ed in this analysis, 10 to 20 unstained 5-mm FFPE tissue sectionswere manually microdissected to enrich for tumor cell DNAusing thehematoxylin and eosin–stained slide as a guide. Samplesfrom 23 of the 42 tumors were processed as described in our priorstudy (13). Briefly, we extracted DNA using the PicoPure DNAExtraction Kit (Arcturus) followed by DNA purification usingthe AMPureXP Kit (Agentcourt Biosciences), respectively. The

Translational Relevance

Ocular adnexal (OA) sebaceous carcinoma is an aggressivemalignancy that frequently recurs and/or metastasizes. Itsdelicate anatomic location on the surface of the eye or eyelidand high propensity for multifocal and infiltrative growthmake OA sebaceous carcinoma a challenging tumor to treat.Surgical excision may produce functional and aesthetic mor-bidity, and effective systemic therapies are lacking. Thus, thereis a critical need to define molecular-genetic drivers of OAsebaceous carcinoma to devise rational and efficacious ther-apies. DNA/RNA sequencing on OA sebaceous carcinomasdelineated two potentially distinct molecular-genetic sub-types. The first harbors mutations in TP53 and/or RB1 withconcomitant NOTCH family member mutations, showshigher nuclear grade features, arises in older patients andmorecommonly recurs locally. The second arises in youngerpatients, shows low-to-intermediate grade nuclear features,lacks mutations in TP53, RB1, orNOTCH, but instead harborstranscriptionally active high-risk human papillomavirus(HPV). These findings provide a potential molecular-geneticframework to understand the pathogenesis of OA sebaceouscarcinoma.

Distinct Subtypes of Ocular Adnexal Sebaceous Carcinoma

www.aacrjournals.org Clin Cancer Res; 25(4) February 15, 2019 1281




resulting genomic DNA was quantified by Qubit dsDNA high-sensitivity assay (Life Technologies). For samples from theremaining 19 tumors, DNA was extracted using the Qiagen FFPEExtraction Kit (Qiagen), the resulting genomic DNA was quan-tified by Picogreen (Invitrogen), and quality was evaluated usingthe TapeStation 2200 instrument with Genomic DNA ScreenTapeassay (Agilent).

Next-generation sequencing of a targeted panel of cancer-associated genes

For 23 of the tumors, somatic mutations and copy numbervariations in 409 cancer-related genes (and the methodologiesfor identifying those) have been previously described (13, 21).Briefly, we used the Ion Torrent Comprehensive Cancer Panel(Life Technologies) for target enrichment and sequencingusing Ion Proton sequencer. Sequencing, data analysis, andidentification of somatic genomic aberrations was performedby excluding variants identified in germline DNA fromnormal tissue in paired tumor samples as described previously(13, 21).

For the remaining 19 tumors, somatic mutations were identi-fied using a different sequencing platform (T200.2) based on lackof commercial availability of the previous platform. T200.2 wasinitially developedby the institute for personalized cancer therapy(IPCT) genomic laboratory and currently consists of 323 cancer-related genes (22). Libraries were made using KAPAHyper [KAPAlibrary prep kit (KAPA)] using the "with beads" manufacturerprotocol. Briefly, we performed enzymatic reactions for endrepair, A-tailing, and adaptor ligation and then inserted thebarcode by PCR using KAPA HiFi polymerase (six cycles). Theresulting libraries were purified using the Agencourt AMPure PCRPurification Kit (Agentcourt Biosciences). Following library prep-aration, sample size was analyzed using the TapeStation 2200(Agilent) or Fragment Analyzer (AATI). The KAPA qPCR Quan-tification Kit was used to quantify samples. Biotin-labeled probes(Roche Nimblegen) were designed to capture all exons in 323cancer-related genes, and captures were performed following themanufacture's protocol. Briefly, equimolar amounts of DNAwerepooled (8–16 samples), capture reagents and probes were added,and samples were incubated at 47�C on a thermocycler withheated lid (57�C) for 64 to 74 hours. We recovered targetedregions using streptavidin beads, and the streptavidin-biotin-probe-target complex was washed and then subjected to PCRamplification according to themanufacturer's protocol. The qual-ity of each capture pool was validated using the TapeStation DNAHigh-Sensitivity Kit (Agilent), and enrichment (minimum cutoff50�) was assessed with qPCR using specific primers (RocheNimblegen). The captured libraries were sequenced on a HiSeq4000 (Illumina Inc.) for 2� 100 base paired end reads accordingto the manufacturer's instructions. All regions were covered bymore than 20 reads. For data analysis, we aligned the T200.2target-capture deep-sequencing data to human hg19 using BWA(23) and removed duplicate reads using Picard (24). We calledsingle nucleotide variants and small in-dels using a proprietarycustom analysis pipeline (25), which classified variants as somat-ic, germline, or loss of heterozygosity on the basis of variant allelefrequencies in the tumor and the matched normal tissues. Toassess the potential functional consequences of detected somaticvariants, we compared them with dbSNP, COSMIC (26), andTCGA databases and annotated them using VEP (27), Annovar(28), CanDrA (29), and other programs. A detailed summary of

the mutations in each primary, locally recurrent and metastatictumor canbe found in Supplementary Table S1, andmean samplecoverage is provided in Supplementary Table S2. The rawdatafilesof the targeted sequencing reported in our manuscript will bedeposited in dbGAP in accordance with the consenting rules andguidelines from dbGAP. All other data reported herein will bemade available from the corresponding author(s) upon reason-able request.

RNA isolation, cDNA creation, library preparation, and librarycapture

Eight distinct patient tumors, including two primary tumorswith TP53 mutations, two primary tumors with both TP53 andRB1 mutations, and four tumors wild type for TP53 and RB1(including three primary and one metastatic tumor) were sub-jected to whole transcriptome RNA sequencing. Total RNA wasisolated using the FFPE RNA Purification Kit (Norgen). ExtractedRNA was quantified by Picogreen (Invitrogen), and quality wasassessed using the TapeStation 2200 (Agilent). cDNA was createdfrom10 to 100 ng of total RNAusing theOvation RNA Seq systemV2 Kit (Nugen). cDNA was sheared by sonication using the E220instrument (Covaris). To ensure the proper fragment size, sampleswere checkedon the TapeStation2200using the TapeStationDNAHigh-Sensitivity Kit (Agilent). Libraries were prepared from thecDNA using the KAPA Library Prep Hyper Kit following the "withbeads"manufacturer protocol, andquality controlwas performedas described above. Samples were then quantified using the KAPAqPCR Quantification Kit. Library capture was performed asdescribed above (pooling two to six samples per pool).

RNA sequencingThe captured libraries were sequenced on a HiSeq 2500

(Illumina) on a version 3 TruSeq paired end flowcell accordingto the manufacturer's instructions at a cluster density of 700,000clusters/mm2 to 1,000,000 clusters/mm2. The resulting BCL filescontaining the sequence data were converted into ".fastq.gz" files,and individual libraries within the samples were de-multiplexedusing CASAVA version 1.8.2with nomismatches. All regionswerecovered by more than 20 reads.

RNA sequencing data analysisRaw RNA sequence data were processed by an in-house RNA

seq data analysis pipeline, which, among other tools, uses theSTAR aligner to align raw reads to the hg19 version of the humanreference genome, featureCounts to quantify aligned reads toproduce raw counts, Oncofuse to filter and prioritize fusioncandidate generated by the STAR aligner, and FastQC andQualiMap to evaluate the quality of raw reads and feature counts.

We used VirusFinder, version 2.0, to align reads that did notmap to the human reference genome to a viral database thatcontains viruses of all known classes (32,102 in total; ref. 30),including all classes of human papillomavirus and human poly-omavirus. Relative levels of RNA expression was quantified in"reads per kilobase million (RPKM)" as follows:

RPKM ¼ Number of reads/([gene length/1,000] � total num-ber of reads in million)

In situ hybridization for high-risk HPV subtypesIn situ hybridization to detect expression of RNA from high-

risk HPV subtypes (HPV 16, 18, 31, 33, 35, 45, 52, and 58) wasperformed using RNAscope 2.5 LS Probe-New Target (HPV 8;

Tetzlaff et al.

Clin Cancer Res; 25(4) February 15, 2019 Clinical Cancer Research1282




ACD Biotech; ready to use) according to the manufacturer'sinstructions. Cases with punctate nuclear and cytoplasmicsignals within the tumor cells on the HPV HR preparationswere considered positive for a transcriptionally active high-riskHPV.

ImmunohistochemistryIHCwasperformedusing thepolymeric biotin-free horseradish

peroxide method on the Leica Microsystems Bond stainer usingantibodies to RB (Calbiochem; clone OP66; 1:30), TP53 (Dako;clone DO-7; 1:100), and Adipophilin (Fitzgerald; clone AP125;prediluted). The Bond Refine Polymer Detection Kit (Leica) wasused for detection.

Analysis of relationship between clinicopathologic variablesand somatic mutations in primary tumors

We analyzed whether various clinical and pathologic variables(including sex, age at primary diagnosis, anatomic location,clinical size, clinical stage (AJCC 7th and 8th editions) correlatedwith the presence or absence of mutations in primary tumors. Forcontinuous variables, t test was used to compare the meansbetween two groups. For categorical variables, Fisher exact testwas applied to test whether the distributions of the variables differacross the groups. Statistical analyses were performed using Rversion 3.4.3. Fisher exact test determinedwhether the differencesin the frequency of various mutations (TP53, RB1) were statisti-cally significant between primary tumors and local recurrences (ata significance level of 0.05).

ResultsNext-generation DNA sequencing of a targeted panel of cancer-associated genes identifies distinct subgroups of OA sebaceouscarcinoma

Our cohort included OA sebaceous carcinoma specimens from29 patients, including 16 women and 13men, with a median ageof 68 years (range, 44–93 years) at primary diagnosis. From these29 patients, we sequenced (i) 26 primary OA sebaceous carcino-mas from 26 patients; (ii) five locally recurrent OA sebaceouscarcinoma from five patients (two with and three withoutmatched primary tumor tissue); and (iii) 11 matched metastaticOA sebaceous carcinomas from eight patients whose primarytumor we also sequenced (Table 1).

Results of next-generation sequencing of a targeted panel ofcancer-associated genes in the primary and locally recurrent OAsebaceous carcinomas are summarized in Fig. 1 (detailed infor-mation regarding the specific mutation in each tumor is sum-marized in Supplementary Table S1). The most commonsomatic mutations in these tumors were mutations affectingTP53 (66%; 19/29 patients) and RB1 (48%; 14/29 patients). In13 of the 29 patients (45%), TP53 and RB1mutations coexistedin the same tumor, and Fisher exact test confirmed the signif-icance of the frequency of TP53 and RB1 comutation (P ¼0.0051). Fifteen of the 19 TP53-mutated tumors with sufficienttissue for testing showed variably increased (aberrant) TP53protein expression, whereas 8 of the 10 TP53 wild-type tumorsshowed either absent or variably weak nuclear TP53 proteinexpression by immunohistochemical studies (Figs. 1 and 2;Supplementary Table S3). IHC studies demonstrated absence ofsignificant nuclear RB protein expression in 12/14 RB1-mutanttumors with sufficient tissue for testing, whereas 12/15 RB1

wild-type tumors showed preserved, strong (albeit variable)nuclear RB expression (Figs. 1 and 2; Supplementary Table S3).Fisher exact test confirmed that aberrant TP53 protein expres-sion in TP53mutant tumors (P ¼ 0.001) and loss of RB proteinexpression in RB1 mutant tumors (P < 0.001) were statisticallysignificantly associated (Supplementary Table S3). A correla-tion between loss of RB protein expression and the type orposition of RB1 gene mutations was not identified. Two of thethree TP53 mutant tumors lacking aberrant TP53 proteinexpression were nonsense mutations occurring at amino acidposition 196 (patient 7) and 146 (patient 19), respectively.

Finally, somaticmutations affectingNOTCH family genes werepresent in six tumors from five patients (NOTCH1 mutations infive tumors from four patients,NOTCH2mutation in one tumor),and all tumors with NOTCH family mutations also carried coex-isting TP53 and RB1 mutations. (NOTCH2 mutations are notincluded in Fig. 1 because the figure only includes genes mutatedin more than 1 tumor.)

Analysis of relationships between clinical variables and thepresence or absence of TP53 and/or RB1 mutations in primarytumors revealed that the 17 patients whose primary tumorscontained somatic mutations affecting TP53 and/or RB1 weresignificantly older than the nine patients whose primary tumorsdid not contain mutations affecting these genes (mean age atprimary diagnosis, 70.3 years vs. 58.1 years; P ¼ 0.02; Table 2).Patients with and without mutations affecting TP53 and/or RB1did not differ with respect to sex, anatomic location of the tumor,clinical tumor size, stage according to the 7th edition of theAmerican Joint Committee on Cancer staging manual or muta-tional load.

Table 1. Patient demographics and anatomic distribution of OA sebaceouscarcinomas

Characteristic Value

Patient demographicsTotal number of patients 29SexMale 13Female 16

Median age at diagnosis (range), years 68 (44–93)EthnicityCaucasian 20Hispanic 5African American 2Asian 1American Indian 1

Tumor characteristicsTotal number of OA sebaceous carcinomas tested 42Tumor typePrimary 26Locally recurrent 5a

Metastatic 11b

Location of metastatic tumorParotid lymph node 5c,d

Neck lymph node 2Mandible soft tissue 1d

Skin, subcutaneous tissue 1Lung 1c

Liver 1c

aFrom two patientswith primary tumors included in the study and three patientswithout primary tumors included in this study.bFrom eight patients with primary tumors all included in the study.cOne patient had parotid, lung, and liver metastases included in the study.dOne patient had parotid and soft tissue metastases included in the study.






Somatic mutations affecting TP53 and RB1 are enriched inlocally recurrent tumors

All five (100%) locally recurrent tumors included in our studyharbored mutations in both TP53 and RB1 (Fig. 1). For two ofthese locally recurrent tumors, the primary tumor was alsosequenced; in both cases, the original primary tumor alsoharbored mutations in both TP53 and RB1. In our experience,most paired primary and recurrent or metastatic OA sebaceouscarcinomas do not differ with respect to their TP53 and RB1mutation status. Therefore, under the assumption that theprimary tumors corresponding to the other three locally recur-rent tumors in our study shared similar mutations in TP53 andRB1, TP53 and RB1 mutations coexist in 13 primary tumorsfrom 13 patients; TP53 and RB1 mutations coexist in fiverecurrent tumors from five patients; TP53 and RB mutationsdid not coexist in tumors from 16 patients. Fisher exact test

showed that the enrichment of somatically acquired mutationsaffecting both TP53 and RB1 in recurrent OA sebaceous carci-noma compared with their frequency in primary OA sebaceouscarcinoma overall was statistically significant (P ¼ 0.046),suggesting that abrogation of both TP53 and RB1 tumor sup-pressor function together may confer a more locally aggressiveclinical phenotype in OA sebaceous carcinoma with a greaterpropensity for local recurrence.

OA sebaceous carcinomas with somatic mutations affectingRB1 exhibit poorly differentiated histopathologic features thatmay contribute to local recurrence

To determine if TP3 and/or RB1 mutant tumors exhibit dis-tinctive histopathologic features (that might account for theirunique clinical tendency for local recurrence), we classified eachprimary tumor according to grade/differentiation of the tumor

Figure 1.

Mutational signatures of primary and locally recurrent OA sebaceous carcinomas. A, Heat map shows genes mutated more than once across the cohort ofprimary and locally recurrent OA sebaceous carcinomas. Genes are listed according to relative frequency of mutations. Each OA sebaceous carcinoma(primary or locally recurrent tumor) is represented in a column, and primary tumors are designated by a number whereas locally recurrent tumors aredesignated by a number plus "LR." Tumors are separated according to the presence of TP53 and RB1 mutations. A colored box indicates the presenceof a somatically acquired mutation or transcriptionally active high-risk HPV infection by in situ hybridization; gray boxes indicate the absence of asomatically acquired mutation or transcriptionally active high-risk HPV infection by in situ hybridization; asterisks indicate that HPV RNA was also detectedby RNA sequencing in one of the samples from that patient. B, Immunohistochemical studies for TP53 and RB1 confirm reciprocal relationships between TP53and RB1 mutations and aberrant protein expression of TP3 and RB1, respectively (all images shown as 200� magnification) for patient11, 17, 20, 25, and 27.

Tetzlaff et al.





cells (high, intermediate, or well; see Materials and Methods andSupplementary Fig. S1), necrosis and pagetoid/intraepithelialspread (presence vs. absence for each; see Materials and Methodsand Supplementary Table S3). Although TP53 mutant tumorsshowed more frequent necrosis compared with TP53 wild-typelesions (P ¼ 0.05), RB1-mutated OA sebaceous carcinomasmore commonly showed a predominance of high-grade nuclearmorphology compared with RB1 wild-type tumors (P ¼ 0.03;Supplementary Table S4). Further, TP53/RB1 comutant tumors(patients 1–13)were enriched for high-grade tumor cells comparedwith non-comutated tumors (patients 14–29; P ¼ 0.02). Notsurprisingly, there was a similar nonsignificant trend for high-grade

nuclear features among theprimary tumors that recurred comparedto those that did not (P ¼ 0.06). No additional statisticallysignificant correlations were observed between histopathologicfeatures and the mutational background were observed in thiscohort of tumors (Supplementary Tables S3 and S4).

Whole transcriptome sequencing and in situ hybridizationconfirm high-risk HPV RNA expression in TP53/RB1 wild-typeOA sebaceous carcinoma

Given the dichotomous distribution of TP53 and RB1 muta-tions in our cohort of primary OA sebaceous carcinomas andthe distinct age at primary tumor diagnosis, histopathologicand local recurrence profiles of TP53/RB1 mutant versus TP53/RB1 wild-type tumors, we hypothesized that a viral or bacterialinfection that directly or indirectly abrogates TP53 and RB (ordependent pathways) might drive or promote the developmentof TP53/RB1 wild-type OA sebaceous carcinoma—similar towhat has been described in Merkel cell carcinoma, a cancerwith a similar binary distribution of TP53/RB1 mutations andpolyomavirus infection (31, 32). Thus, we submitted eightOA sebaceous carcinomas from eight patients for whole tran-scriptome RNA sequencing: four primary tumors with TP53mutations (including two tumors with coexisting TP53 and RB1mutations), and four tumors (three primary and one lymphnode metastasis) wild type for TP53 and RB1. RNA sequencingidentified variable levels of high-risk HPV RNA (type 16 andtype 18) in two of the four TP53/RB1 wild-type tumors butnone of the four TP53/RB1mutant tumors. For patient 21, HPVtype 18 RNA sequences were detected at an average of 141RPKM (see Materials and Methods), whereas HPV type 16 RNAsequences were detected at an average of 7 RPKM for patient 22.No additional unique viral or bacterial RNA sequences wereidentified in any of the 8 tumors tested.

To confirm and expand on these findings, we performed in situhybridization to detect RNA expression of high-risk HPV RNA inthe entire cohort of patient tumors. We identified transcription-ally active high-risk HPV RNA in four of the nine patients (44%)with TP53/RB1 wild-type OA sebaceous carcinoma (includingthe two patients inwhomhigh-riskHPVwas previously identifiedin the primary tumor by RNA sequencing), but in none of the20 patients whose tumors harbored somatic TP53 and/or RB1mutations (including the four tumors originally subjected toRNA sequencing; Figs. 1–3). Fisher exact test confirmed thestatistical significance of the dichotomous relationship betweeneither TP53 and/or RB1 mutation versus the presence of tran-scriptionally active HPV (P ¼ 0.0053). Furthermore, patientswith HPV-positive OA sebaceous carcinoma were significantlyyounger (mean age at primary diagnosis, 54 years) than patientswith HPV-negative OA sebaceous carcinoma (mean age at pri-mary diagnosis, 68.3 years; P¼ 0.01) and patients whose primaryOA sebaceous carcinoma contained somatic mutations affectingTP53 and/or RB1 (mean, 70.3 years; P ¼ 0.006).

MetastaticOA sebaceous carcinomas harbor somaticmutationsin TP53 and/or RB1 or infection by transcriptionally activehigh-risk HPV

Results of next-generation sequencing of cancer-associatedgenes in the metastatic OA sebaceous carcinomas (11 metastatictumors from 8 patients) are summarized in Fig. 3, and a pairwisecomparison of somatically acquiredmutations in primary tumorsandpairedmetastases is provided in Supplementary Table S1. The

Figure 2.

Molecular-genetic subtypes of OA sebaceous carcinoma. Left panel showsa tumor carrying both a TP53 and RB1 mutation (A; hematoxylin andeosin; 400�). Right panel shows a tumor wild type for TP53 and RB1(B; hematoxylin and eosin; 400�). IHC studies demonstrate positivity inboth tumors for adipophilin (C, D; 400�). However, the tumor on the leftpanel shows aberrant nuclear accumulation of TP53 protein (E; 400�)compared with no aberrant nuclear accumulation of TP53 protein in theother tumor (F; 400�). There is also loss of RB protein expression(G; 400�) on the left, but mostly preserved RB protein expression onthe right (H; 400�). Finally, whereas the tumor on the left shows nodetectable transcriptionally active high-risk HPV (for the subtypesstudied) by in situ hybridization (I; 400�), in situ hybridization confirmspresence of transcriptionally active high-risk HPV in the tumor cells(nuclear and cytoplasmic signal distribution; J; 400�).






somatically acquiredmutation profile inmetastaticOA sebaceouscarcinoma largely reflected that seen among primary OA seba-ceous carcinoma: six of eight patients (75%) had TP53mutations,and four of eight (50%) hadRB1mutations (all of those coexistedwith TP53 mutations). Finally, somatic mutations affectingNOTCH genes were identified in four metastatic tumors fromfour patients; mutations inNOTCH1were identified in two TP53/RB1mutant tumors, andmutations inNOTCH3were identified intwo TP53 mutant tumors.

Two metastatic OA sebaceous carcinomas from two patientslacked mutations in TP53, RB1, or NOTCH family membersbut harbored transcriptionally active high-risk HPV subtypes.However, neither the relative frequency of TP53 and/or RB1mutations nor the relative frequency of HPV infection differedsignificantly betweenmetastatic andprimary/locally recurrentOAsebaceous carcinomas (P > 0.05 for all comparisons using Fisherexact test).

DiscussionOurfindings in this studyprovide apotentialmolecular-genetic

framework by which to understand the development and pro-gression of primary and locally recurrent OA sebaceous carcino-mas andhave important clinical implications.Wedemonstrated abimodal, potentially distinct distribution of somatically acquiredmutations and transcriptionally active HPV infection that defineat least two distinct molecular-genetic subtypes of primary OAsebaceous carcinoma. Seventy-one percent (22/31) of the primaryand locally recurrent tumors from 69% (20/29) of the patients inthis series had OA sebaceous carcinomas of the first subtype,harboringmutations inTP53 and/orRB1. Fifteen of the 16 tumors(94%) harboring RB1 mutations also harbored TP53 mutations(P¼0.0051), and sixTP53/RB1mutantOAsebaceous carcinomas(from five patients) also carried concomitant NOTCH familymember mutations: NOTCH1 mutation in five tumors and

Table 2. Comparisonof clinical and pathologic variables betweenpatientswith primaryOA sebaceous carcinomaswith TP53 and/orRB1mutations andpatientswithprimary OA sebaceous carcinomas lacking TP53 or RB1 mutations

VariableAll patients(n ¼ 26)

Patients withTP53 and/or RB1mutations (n ¼ 17)

Patients withoutTP53 or RB1mutations (n ¼ 9) P-valuea

Sex, n (%) 1Male 13 (50) 9 (53) 4 (44)Female 13 (50) 8 (47) 5 (56)

Age, mean (SD), years 66.1 (13.1) 70.3 (12.1) 58.1 (11.7) 0.02Tumor location, binary, n (%) 0.69Left 10 (38) 6 (35) 4 (44)Right 16 (62) 11 (65) 5 (56)

Clinical tumor diameter, mean (SD), mm 16.3 (10.1) 17.6 (11.6) 13.7 (6.2) 0.27T category, n (%)b 1T1 1 (4) 1 (6) 0 (0)T2 11 (42) 7 (41) 4 (44)T3 14 (54) 9 (53) 5 (56)

Mutational load (mean; range) 50–22 51–15 50–22 0.369aFor continuous variables (age and clinical tumor diameter), t test was used to compare the means between two groups. For categorical variables (sex, tumorlocation, T category), Fisher exact was used to test whether the distributions of the variables differed across the groups.bPer the seventh edition of the American Joint Committee on Cancer staging manual.

Figure 3.

Mutational signatures of metastaticOA sebaceous carcinomas. Genesmutated more than once are listedaccording to relative frequency ofmutations. Each metastatic OAsebaceous carcinoma is representedin a column, and tumors are groupedaccording to the frequency of TP53andRB1mutations. Red boxes indicatethe presence of a somatically acquiredmutation or transcriptionally activehigh-risk HPV infection by in situhybridization; gray boxes indicate theabsence of a somatically acquiredmutation or transcriptionally activehigh-risk HPV infection by in situhybridization; asterisks indicate thatHPV RNA was also detected by RNAsequencing in one of the samples fromthat patient.

Tetzlaff et al.





NOTCH2mutation in one tumor. In contrast, 29% (9/31) of theprimary and locally recurrent tumors and 31% (9/29) from thepatients in this series hadOA sebaceous carcinomas of the secondsubtype. These lackedmutations in TP53, RB1, orNOTCH and, infour of the nine tumors, harbored infection by transcriptionallyactive high-risk HPV subtypes. The clinical significance of thisdichotomization is underscored by our findings that (i) age atprimary tumor diagnosis was significantly higher for primarytumors with mutations in TP53 and/or RB1 than for primarytumors lacking thesemutations or primary tumors with transcrip-tionally activeHPV infection; (ii) tumors withRB1mutations andTP53/RB1 comutant tumors exhibited higher grade features his-topathologically compared with the RB1wild-type and TP53/RB1non-comutant tumors, respectively; and (iii) all five locally recur-rent tumors harbored mutations in both TP53 and RB1, a statis-tically significant enrichment of this genotype among locallyrecurrent compared to primary OA sebaceous carcinomas (P ¼0.046), indicating a more locally aggressive phenotype amongtumors with this genotype thatmight possibly reflect their enrich-ment for higher nuclear grades. Our demonstration of potentiallydistinct molecular-genetic subtypes of OA sebaceous carcinomawith apparently distinct clinical phenotypes represents a pivotaldevelopment in our understanding of the disease.

Our finding that somatic mutations affecting TP53 were themost commonalteration inOA sebaceous carcinoma is consistentwith prior studies demonstrating TP53 mutations occurred inapproximately 63% (37/59) of tumors (11–13, 33). Prior studieshave shown that TP53 alterations are largely restricted to OAsebaceous carcinomas of the eyelid and further occur in a mutu-ally exclusively fashion with defects in mismatch repair (13, 34),as the latter are predominantly seen in sebaceous carcinomasaway from the eyelid. Together, abrogation of TP53 or of mis-match repair represent distinct pathways in the developmentof sebaceous carcinoma, but as previously described, theselargely occur in accordance with their distinctive anatomicdistributions (34).

Our finding that somatic RB1mutations were the second mostcommon somatically acquired mutation in OA sebaceous carci-noma, occurring in 48%(14/29) of our patients, is consistentwiththe long-established propensity for patients with hereditary ret-inoblastoma to also develop OA sebaceous carcinoma (35–41).More strikingly, 93% (13/14) of the patients in our series whohadRB1 mutation in their primary and/or locally recurrent OA seba-ceous carcinoma also harbored TP53 mutations in the sametumor, and this coalescence of somatic mutations in TP53 andRB1 was enriched in older patients and in locally recurrenttumors, which not only supports a pivotal role for concomitantTP53 and RB1 abrogation in the development of OA sebaceouscarcinoma, but also suggests that this molecular-genetic subtypeconfers a more locally aggressive clinical phenotype, possiblyrelated to their higher nuclear grade features.

The apparent dichotomous distribution of TP53 and RB1mutations across primary or recurrent OA sebaceous carcinomasin this series was similarly striking: 69% (20/29) patients withprimary or recurrent OA sebaceous carcinoma harbored muta-tions in one and/or the other gene [including 45% (13/29) withmutations in both] and occurred in older patients (mean 70.3years) and exhibited higher nuclear grade, whereas 31% (9/29) ofpatients with OA sebaceous carcinoma lackedmutations in eithergene, and these tumors occurred in younger patients (mean 58.1years; P ¼ 0.02) and more commonly showed intermediate and

well-differentiated nuclear grade features. The mutual exclusivityof TP53/RB1 comutant tumors from HPV-positive tumors wasstatistically significant (P ¼ 0.0053). Similar dichotomous dis-tributions have been observed in other tumors, including Merkelcell carcinoma (31, 32) and squamous cell carcinoma of the headand neck (42). In each of these cancers, significant proportions ofcases are caused by infection with an oncogenic virus: Merkel cellcarcinoma polyomavirus (MCPyV) inMerkel cell carcinomas andHPV in many different squamous cell carcinomas. Furthermore,MCPyV-positiveMerkel cell carcinomas contain significantly few-er somatic mutations than their MCPyV-negative counterparts,which harbor a high frequency of ultraviolet light–inducedmuta-tions and are specifically enriched for mutations affecting TP53and RB1 (31, 32). Based on these observations, we hypothesizedthat OA sebaceous carcinoma lacking TP53 or RB1 mutationsmight also contain oncogenic viral sequences that interfere direct-ly or indirectly with TP53- and RB-dependent pathways. RNAsequencing and subsequent in situ hybridization studies con-firmed this prediction, as we identified transcriptionally activehigh-risk HPV infection in 44% (4/9) of the TP53/RB1 wild-typeOA sebaceous carcinomas but none of the TP53/RB1-mutatedtumors. HPV-positive tumors developed in younger patients(mean age at primary diagnosis, 54 years) than HPV-negativetumors (mean age at primary diagnosis, 68.3 years; P¼ 0.01) andtumors with mutations in TP53 and/or RB1 (mean age at primarydiagnosis, 70.3 years; P¼ 0.006) and exhibit similarly distinctivenuclear grade/differentiation histopathologically. The apparentmutually exclusive relationship observed between TP53/RB1mutations and HPV infection most likely reflects the targets ofHPV infection. Integration of HPV viral DNA into the hostgenome results in expression of viral proteins E6 and E7, whichsubsequently bind to and inactivate host cell tumor suppressorproteins TP53 and RB, respectively, interfering with TP53/RB-regulated cellular pathways directing DNA damage repair, apo-ptosis, and proliferation (43). Our findings suggest that thedevelopment of OA sebaceous carcinoma relies exquisitely onthe abrogation of TP53 and/or RB function—either mutationalor viral protein–mediated. Furthermore, our results at least par-tially explain the previously reported predisposition for OAsebaceous carcinoma among patients with long-standing immu-nosuppression, including immunosuppression due to solid organtransplant (15–17) and HIV infection or AIDS (18).

Prior studies have identifiedHPV sequences in 13.1% (14/107)of OA sebaceous carcinomas (11, 44–47), a frequency similar tothe frequency of transcriptionally active high-risk HPV in patientswith primary or locally recurrent tumors in our study: 13.8%(4/29). However, no previous study specifically interrogated therelationship between HPV infection and TP53/RB1 mutationalstatus in OA sebaceous carcinomas. As such, prior efforts haveproduced conflicting results on the relationship between OAsebaceous carcinoma and HPV infection. Some studies (11, 44,46) failed to confirm any relationship with HPV infection across38patient tumors, and a study inwhich a singleHPV-positive casewas identified by PCR for HPV DNA failed to confirm transcrip-tionally active HPV infection by in situ hybridization (48). Thesedifferences are partially attributable to the application of differingtechnologies forHPVdetection, including in situhybridization forHPV DNA or RNA or PCR for HPV DNA. However, to the extentthat aberrant expression of p53 measured by IHC correlateswith TP53 gene mutation, previous findings partially align withour finding of a mutually exclusive relationship between either






TP53/RB1 mutation or HPV infection. Hayashi and colleagues(45) applied immunohistochemistry to detect HPV and TP53proteins to 21 cases of OA sebaceous carcinoma and found thatfive cases were positive for HPV only, five were positive for TP53only, and fourwerenegative for bothHPVandTP53, all consistentwith our findings, although seven tumors in their series werepositive for both HPV and TP53 (45). However, it is unclearwhether these latter TP53-positive tumors actually carried asomatic TP53mutation. Of note, detection of p16 protein expres-sion—typically a surrogate for HPV positivity in HPV-driventumors—has not been shown to predict HPV infection in priorstudies on OA sebaceous carcinoma (48). This is likely becauseHPV-negative OA sebaceous carcinomas harbor a high frequencyinactivating RB1 mutations, and RB1 inactivation itself typicallycorrelates with elevated levels of p16 protein expression (49).Nevertheless, additional mechanisms abrogating TP53 and RBfunction in HPV-negative TP53/RB1 wild-type OA sebaceouscarcinoma remain to be determined and may involve additionalepigenetic controls.

The distinct segregation of NOTCH mutations in tumors withconcomitant TP53 and RB1mutations provides additional expla-nation for the observed dichotomy between OA sebaceous car-cinomas with TP53/RB1mutations and those with high-risk HPVinfection as etiopathogenic. NOTCH genes were mutated in sixprimary or locally recurrent tumors from five patients(NOTCH1 in five tumors and NOTCH2 in one tumor) and inmetastatic tumors from four patients (NOTCH1 in two andNOTCH3 in two). NOTCH mutations occurred exclusively inTP53 mutated tumors, and all but two tumors with NOTCHmutations were TP53/RB1 double mutant tumors. In addition

to binding and inactivating TP53 protein, HPV E6 protein bindsand inactivates MAML1 protein (Mastermind Like Transcrip-tional Coactivator-1), a transcriptional coactivator and keyeffector of NOTCH-mediated transcription (50, 51). Previousstudies have shown similar mutually exclusive relationshipsbetween HPV infection and NOTCH mutations. In a study ofhead and neck squamous cell carcinomas, HPV-negativetumors more commonly carried NOTCH1 mutations whereasHPV-positive tumors more commonly carried a wild-typeNOTCH1 gene (P ¼ 0.031; ref. 52).

Important limitations to this study include the relatively smallsample size (including only four patients with transcriptionallyactive high-risk HPV OA sebaceous carcinoma), the reliance onFFPE tissue resources, and the targeted nature of the DNA-sequencing studies. Five of our patients did not fit into either"type" of OA sebaceous carcinoma. Additional studies (ideallywith fresh tissue) on a larger cohort of patients (includingexpanded whole exome DNA sequencing and whole transcrip-tomeRNA-sequencing studies) are needed to fully understand themolecular landscopae of this rare cancer type.

In summary, we have identified two potentially distinct sub-groups of OA sebaceous carcinoma (Fig. 4). Tumors in the firstsubtype are more common, tend to occur in older patients (meanage at primary diagnosis, 70.3 years), andhave a high frequency ofmutations in TP53 and/or RB1, which occasionally coexist withmutations inNOTCH genes. In the vastmajority of cases withRB1mutation, TP53 is also mutated (P ¼ 0.0051). To the extent thatlocally recurrent tumors in our series consist exclusively of thisgenotype, TP53/RB1 mutated tumors appear to exhibit a morelocally aggressive clinical phenotype including higher grade

Figure 4.

A model for the development of OAsebaceous carcinoma. Type I tumorsdefined by the presence of somaticTP53 and/or RB1 mutations. Theyoccur in younger patients and exhibitincreased frequency of localrecurrence. Type II tumors lackmutations in TP53 or RB1 and insteadshow high frequency infection by HPVhigh risk subtypes, and viral E6 andE7 proteins abrogate TP53 and RBprotein in these tumors. This subtypeof tumor occurs in younger patientsand exhibits less frequent localrecurrence.

Tetzlaff et al.





nuclear features and may warrant more aggressive postoperativesurveillance. Tumors comprising the second subtype of OA seba-ceous carcinoma occur in younger patients (mean age at primarydiagnosis, 58.1 years) and lack mutations in TP53, RB1, orNOTCH family members. Instead, almost half of the tumors inthis subtype harbor infection by transcriptionally active high-riskHPV subtypes. The mean age of patients with these HPV-positivetumors at primary diagnosis is only 54 years. Our findingsposition clinicians treating this aggressive malignancy to deploynext-generation sequencing strategies to identify either genome-matched targeted therapies directed at the clinically actionablemutations or to identify immune-based therapies (possibly incor-porating HPV vaccination strategies) to further improve patientoutcomes.

Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: M.T. Tetzlaff, V.G. Prieto, B. EsmaeliDevelopment of methodology: M.T. Tetzlaff, C.W. Hudgens, A. Yemelyanova,B. Esmaeli

Acquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): M.T. Tetzlaff, O. Sagiv, D. Ivan, A.K. Eterovic,K. Shaw, B. EsmaeliAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): M.T. Tetzlaff, J. Ning, B. Peng, M. Routbort,C.W. Hudgens, T.-B. Kim, K. Chen, A.K. Eterovic, K. Shaw, B. EsmaeliWriting, review, and/or revision of the manuscript: M.T. Tetzlaff, J.L. Curry,J. Ning, O. Sagiv, B. Peng, D. Bell, M. Routbort, D. Ivan, K. Shaw, V.G. Prieto,A. Yemelyanova, B. EsmaeliAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases):M.T. Tetzlaff, T.L. Kandl, A.K. Eterovic, K. Shaw,B. EsmaeliStudy supervision: M.T. Tetzlaff, O. Sagiv, B. Esmaeli

AcknowledgmentsThis study was funded by the Avery Orbital Oncology Fund for Research and

Education. We thank Ms. Kim-Anh Vu for her outstanding help with figuregeneration and graphic design.We thankMs. StephanieDeming formagnificentmedical editing.

The costs of publication of this article were defrayed in part by the pay-ment of page charges. This article must therefore be hereby marked advertise-ment in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Received May 31, 2018; revised July 25, 2018; accepted November 2, 2018;published first November 12, 2018.

References1. Husain A, Blumenschein G, Esmaeli B. Treatment and outcomes for

metastatic sebaceous cell carcinoma of the eyelid. Int J Dermatol 2008;47:276–9.

2. Kass LG, Hornblass A. Sebaceous carcinoma of the ocular adnexa. SurvOphthalmol 1989;33:477–90.

3. Nelson BR, Hamlet KR, Gillard M, Railan D, Johnson TM. Sebaceouscarcinoma. J Am Acad Dermatol 1995;33:1–15; quiz 6–8.

4. Shields JA, Demirci H, Marr BP, Eagle RC Jr, Shields CL. Sebaceouscarcinoma of the ocular region: a review. Surv Ophthalmol 2005;50:103–22.

5. Zurcher M, Hintschich CR, Garner A, Bunce C, Collin JR. Sebaceouscarcinoma of the eyelid: a clinicopathological study. Br J Ophthalmol1998;82:1049–55.

6. Savar A, Oellers P, Myers J, Prieto VG, Torres-Cabala C, Frank SJ, et al.Positive sentinel node in sebaceous carcinoma of the eyelid. Ophthal PlastReconstr Surg 2011;27:e4–6.

7. Shields JA, Demirci H, Marr BP, Eagle RC Jr, Shields CL. Sebaceouscarcinoma of the eyelids: personal experience with 60 cases. Ophthalmol-ogy 2004;111:2151–7.

8. Rao NA, Hidayat AA, McLean IW, Zimmerman LE. Sebaceous carcinomasof the ocular adnexa: a clinicopathologic study of 104 cases, with five-yearfollow-up data. Hum Pathol 1982;13:113–22.

9. Esmaeli B, Nasser QJ, Cruz H, Fellman M, Warneke CL, Ivan D. AmericanJoint Committee on Cancer T category for eyelid sebaceous carcinomacorrelates with nodal metastasis and survival. Ophthalmology 2012;119:1078–82.

10. Jung YH, Woo IS, Kim MY, Han CW, Rha EY. Palliative 5-fluorouraciland cisplatin chemotherapy in recurrent metastatic sebaceous carcino-ma: case report and literature review. Asia Pac J Clin Oncol 2016;12:e189–93.

11. Gonzalez-Fernandez F, Kaltreider SA, Patnaik BD, Retief JD, Bao Y, New-man S, et al. Sebaceous carcinoma. Tumor progression throughmutationalinactivation of p53. Ophthalmology 1998;105:497–506.

12. Hussain RM, Matthews JL, Dubovy SR, Thompson JM, Wang G. UV-independent p53mutations in sebaceous carcinomaof the eyelid.OphthalPlast Reconstr Surg 2014;30:392–5.

13. Tetzlaff MT, Singh RR, Seviour EG, Curry JL, Hudgens CW, Bell D, et al.Next-generation sequencing identifies high frequency of mutations inpotentially clinically actionable genes in sebaceous carcinoma. J Pathol2016;240:84–95.

14. TetzlaffMT,Curry JL, Yin V, Pattanaprichakul P,Manonukul J, UiprasertkulM, et al. Distinct pathways in the pathogenesis of sebaceous carcinomasimplicated by differentially expressed microRNAs. JAMA Ophthalmol2015;133:1109–16.

15. Zavos G, Karidis NP, Tsourouflis G, Bokos J, Diles K, Sotirchos G, et al.Nonmelanoma skin cancer after renal transplantation: a single-centerexperience in 1736 transplantations. Int J Dermatol 2011;50:1496–500.

16. Imko-Walczuk B, Krys A, Lizakowski S, Debska-Slizien A, Rutkowski B,Biernat W, et al. Sebaceous carcinoma in patients receiving long-termimmunosuppresive treatment: case report and literature review. TransplantProc 2014;46:2903–7.

17. Hoss E, Nelson SA, Sharma A. Sebaceous carcinoma in solid organtransplant recipients. Int J Dermatol 2017;56:746–9.

18. Lanoy E, Dores GM, Madeleine MM, Toro JR, Fraumeni JF Jr, Engels EA.Epidemiology of nonkeratinocytic skin cancers among persons with AIDSin the United States. AIDS 2009;23:385–93.

19. Ostler DA, Prieto VG, Reed JA, Deavers MT, Lazar AJ, Ivan D. Adipophilinexpression in sebaceous tumors and other cutaneous lesions with clear cellhistology: an immunohistochemical study of 117 cases. Mod Pathol2010;23:567–73.

20. Tetzlaff MT. Immunohistochemical markers informing the diagnosisof sebaceous carcinoma and its distinction from its mimics: adi-pophilin and factor XIIIa to the rescue? J Cutan Pathol 2018;45:29–32.

21. Singh RR, Patel KP, Routbort MJ, Aldape K, Lu X, Manekia J, et al. Clinicalmassively parallel next-generation sequencing analysis of 409 cancer-related genes for mutations and copy number variations in solid tumours.Br J Cancer 2014;111:2014–23.

22. Chen K, Meric-Bernstam F, Zhao H, Zhang Q, Ezzeddine N, Tang LY, et al.Clinical action ability enhanced through deep targeted sequencing of solidtumors. Clin Chem 2015;61:544–53.

23. Li H, Durbin R. Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 2009;25:1754–60.

24. DePristo MA, Banks E, Poplin R, Garimella KV, Maguire JR, HartlC, et al. A framework for variation discovery and genotypingusing next-generation DNA sequencing data. Nat Genet 2011;43:491–8.

25. ZhouWD,ChenTH, ZhaoH, Eterovic AK,Meric-BernstamF,MillsGB, et al.Bias from removing read duplication in ultra-deep sequencing experi-ments. Bioinformatics 2014;30:1073–80.






26. Bamford S, Dawson E, Forbes S, Clements J, Pettett R, Dogan A, et al. TheCOSMIC (Catalogue of Somatic Mutations in Cancer) database andwebsite. Br J Cancer 2004;91:355–8.

27. McLaren W, Pritchard B, Rios D, Chen YA, Flicek P, CunninghamF. Deriving the consequences of genomic variants with theEnsembl API and SNP Effect Predictor. Bioinformatics 2010;26:2069–70.

28. WangK, LiM,HakonarsonH.ANNOVAR: functional annotationof geneticvariants from high-throughput sequencing data. Nucleic Acids Res2010;38:e164.

29. Mao Y, Chen H, Liang H, Meric-Bernstam F, Mills GB, Chen K. CanDrA:cancer-specific driver missense mutation annotation with optimized fea-tures. PLoS One 2013;8:e77945.

30. Bhaduri A,QuK, LeeCS,Ungewickell A, Khavari PA. Rapid identificationofnon-human sequences in high-throughput sequencing datasets. Bioinfor-matics 2012;28:1174–5.

31. Harms PW, Vats P, Verhaegen ME, Robinson DR, Wu YM, Dha-nasekaran SM, et al. The distinctive mutational spectra of poly-omavirus-negative Merkel cell carcinoma. Cancer Res 2015;75:3720–7.

32. Wong SQ, Waldeck K, Vergara IA, Schroder J, Madore J, Wilmott JS, et al.UV-associated mutations underlie the etiology of MCV-negative Merkelcell carcinomas. Cancer Res 2015;75:5228–34.

33. Kiyosaki K, Nakada C, Hijiya N, Tsukamoto Y, Matsuura K, Nakatsuka K,et al. Analysis of p53 mutations and the expression of p53 and p21WAF1/CIP1 protein in 15 cases of sebaceous carcinoma of the eyelid. InvestOphthalmol Vis Sci 2010;51:7–11.

34. Shalin SC, Sakharpe A, Lyle S, Lev D, Calonje E, Lazar AJ. p53 stainingcorrelates with tumor type and location in sebaceous neoplasms. Am JDermatopathol 2012;34:129–35; quiz 36–8.

35. Boniuk M, Zimmerman LE. Sebaceous carcinoma of the eyelid, eyebrow,caruncle, and orbit. Trans Am Acad Ophthalmol Otolaryngol 1968;72:619–42.

36. DraperGJ, Sanders BM, Kingston JE. Second primary neoplasms in patientswith retinoblastoma. Br J Cancer 1986;53:661–71.

37. Eng C, Li FP, Abramson DH, Ellsworth RM, Wong FL, Goldman MB, et al.Mortality from second tumors among long-term survivors of retinoblas-toma. J Natl Cancer Inst 1993;85:1121–8.

38. Forrest AW. Tumors following radiation about the eye. Trans Am AcadOphthalmol Otolaryngol 1961;65:694–717.

39. Howrey RP, Lipham WJ, Schultz WH, Buckley EG, Dutton JJ, KlintworthGK, et al. Sebaceous gland carcinoma: a subtle second malignancy fol-lowing radiation therapy in patients with bilateral retinoblastoma. Cancer1998;83:767–71.

40. Rundle P, Shields JA, Shields CL, Eagle RC Jr, Singh AD. Sebaceous glandcarcinoma of the eyelid seventeen years after irradiation for bilateralretinoblastoma. Eye (Lond) 1999;13(Pt 1):109–10.

41. Kivela T, Asko-Seljavaara S, Pihkala U, Hovi L, Heikkonen J. Sebaceouscarcinoma of the eyelid associated with retinoblastoma. Ophthalmology2001;108:1124–8.

42. Lechner M, Frampton GM, Fenton T, Feber A, Palmer G, Jay A, et al.Targeted next-generation sequencing of head and neck squamous cellcarcinoma identifies novel genetic alterations in HPVþ andHPV� tumors.Genome Med 2013;5:49–60.

43. Ferenczy A, Franco E. Persistent human papillomavirus infection andcervical neoplasia. Lancet Oncol 2002;3:11–6.

44. Cho KJ, Khang SK, Koh JS, Chung JH, Lee SS. Sebaceous carcinoma of theeyelids: frequent expression of c-erbB-2 oncoprotein. J Korean Med Sci2000;15:545–50.

45. HayashiN, FurihataM,Ohtsuki Y,UenoH. Search for accumulationof p53protein and detection of human papillomavirus genomes in sebaceousgland carcinoma of the eyelid. Virchows Arch 1994;424:503–9.

46. Kwon MJ, Shin HS, Nam ES, Cho SJ, Lee MJ, Lee S, et al. Comparison ofHER2 gene amplification and KRAS alteration in eyelid sebaceous carci-nomas with that in other eyelid tumors. Pathol Res Pract 2015;211:349–55.

47. Liau JY, Liao SL, Hsiao CH, LinMC, ChangHC, Kuo KT.Hypermethylationof theCDKN2Agenepromoter is a frequent epigenetic change inperiocularsebaceous carcinoma and is associated with younger patient age. HumPathol 2014;45:533–9.

48. Stagner AM, Afrogheh AH, Jakobiec FA, Iacob CE, Grossniklaus HE,Deshpande V, et al. p16 expression is not a surrogate marker for high-riskhuman papillomavirus infection in periocular Sebaceous Carcinoma.Am J Ophthalmol 2016;170:168–75.

49. Shapiro GI, Edwards CD, Kobzik L, Godleski J, RichardsW, Sugarbaker DJ,et al. Reciprocal Rb inactivation and p16INK4 expression in primary lungcancers and cell lines. Cancer Res 1995;55:505–9.

50. Tan MJ, White EA, Sowa ME, Harper JW, Aster JC, Howley PM. Cutaneousbeta-human papillomavirus E6 proteins bind Mastermind-like coactiva-tors and repress Notch signaling. Proc Natl Acad Sci U S A 2012;109:E1473–80.

51. Brimer N, Lyons C, Wallberg AE, Vande Pol SB. Cutaneous papillomavirusE6 oncoproteins associate with MAML1 to repress transactivation andNOTCH signaling. Oncogene 2012;31:4639–46.

52. Rettig EM,ChungCH, Bishop JA,Howard JD, SharmaR, Li RJ, et al. CleavedNOTCH1 expression pattern in head and neck squamous cell carcinoma isassociated with NOTCH1 mutation, HPV status, and high-risk features.Cancer Prev Res (Phila) 2015;8:287–95.


Tetzlaff et al.




2019;25:1280-1290. Published OnlineFirst November 12, 2018.Clin Cancer Res Michael T. Tetzlaff, Jonathan L. Curry, Jing Ning, et al. Nonoverlapping Molecular-Genetic AlterationsHPV-Driven and Virus-Negative Tumors Arise through Distinct Biological Types of Ocular Adnexal Sebaceous Carcinoma:

Updated version

10.1158/1078-0432.CCR-18-1688doi:

Access the most recent version of this article at:

Material

Supplementary

http://clincancerres.aacrjournals.org/content/suppl/2018/11/10/1078-0432.CCR-18-1688.DC1

Access the most recent supplemental material at:

Cited articles

http://clincancerres.aacrjournals.org/content/25/4/1280.full#ref-list-1

This article cites 52 articles, 8 of which you can access for free at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://clincancerres.aacrjournals.org/content/25/4/1280To request permission to re-use all or part of this article, use this link



http://clincancerres.aacrjournals.org/lookup/doi/10.1158/1078-0432.CCR-18-1688

http://clincancerres.aacrjournals.org/content/suppl/2018/11/10/1078-0432.CCR-18-1688.DC1

http://clincancerres.aacrjournals.org/content/25/4/1280.full#ref-list-1

http://clincancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://clincancerres.aacrjournals.org/content/25/4/1280


distinct biological types of ocular adnexal sebaceous ...purpose: ocular adnexal (oa) sebaceous...

Documents