INSTITUTO METRÓPOLE DIGITAL
PROGRAMA DE PÓS-GRADUAÇÃO EM BIOINFORMÁTICA
VANDECLÉCIO LIRA DA SILVA
BIOINFORMÁTICA APLICADA NA IDENTIFICAÇÃO DE GENES DE CÂNCER/TESTÍCULO E SUA ASSOCIAÇÃO COM PROGNÓSTICO EM UMA
ANÁLISE PAN-CÂNCER
NATAL - RN 2019
VANDECLÉCIO LIRA DA SILVA
BIOINFORMÁTICA APLICADA NA IDENTIFICAÇÃO DE GENES DE CÂNCER/TESTÍCULO E SUA ASSOCIAÇÃO COM PROGNÓSTICO EM UMA
ANÁLISE PAN-CÂNCER
Defesa de Doutorado apresentada ao Programa de Pós-Graduação em Bioinformática da Universidade Federal do Rio Grande do Norte.
Área de concentração: Bioinformática
Linha de Pesquisa: Desenvolvimento de Produtos e Processos
Orientador: Prof. Dr. Sandro José de Souza
Co-orientadora: Prof. Dr. Lucymara Fassarella Agnez Lima.
NATAL-RN 2019
da Silva, Vandeclecio Lira.
Bioinformática aplicada na identificação de genes de
câncer/testículo e sua associação com prognóstico em uma análise pan-câncer / Vandeclecio Lira da Silva. - 2019.
47 f.: il.
Tese (doutorado) - Universidade Federal do Rio Grande do Norte, Instituto Metrópole Digital, Programa de Pós-Graduação em Bioinformática, Natal, RN, 2020.
Orientador: Prof. Dr. Sandro José de Souza.
Coorientadora: Profa. Dra. Lucymara Fassarella Agnez Lima. 1. Bioinformática - Tese. 2. Genes de câncer - Tese. 3. Genes de testículos Tese. 4. Biomarcadores Tese. 5. Antígenos -Tese. I. Souza, Sandro José de. II. Lima, Lucymara Fassarella Agnez. III. Título.
RN/UF/BCZM CDU 004:616.6 Catalogação de Publicação na Fonte. UFRN - Biblioteca Central Zila Mamede
DEDICATÓRIA
Dedico este trabalho, aos meus pais, Valdetário Paulino da Silva e Célia Maria Lira da Silva, que depositaram seu amor e dedicação para que eu pudesse chegar até aqui.
Em especial a minha, para sempre namorada, Kiwia Dayanne, que agora me acompanha nessa nova etapa da minha vida.
AGRADECIMENTOS
Agradeço a Deus, pois tudo que pedi a Ele, me foi concedido. Agradeço aos meus pais, irmão e familiares por todo apoio que me deram.
Em especial ao meu orientador, Sandro José de Souza, responsável por toda minha trajetória acadêmica. Agradeço todos os conselhos e incentivos acadêmicos proporcionados.
O presente trabalho foi realizado com apoio da Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES), através da rede de Biologia Sistêmica do Câncer. A todos meus colegas do BioME, que estiveram presentes nessa etapa da minha formação
RESUMO
Os genes de câncer/testículo (CT) são excelentes candidatos para o desenvolvimento de ferramentas imunoterapêuticos associadas ao câncer, devido à sua expressão restrita em tecidos normais e à capacidade de provocar uma resposta imune quando expressa em células tumorais. Neste sentido, no presente estudo realizamos uma análise em escala genômica para os genes de CT, identificando 745 putativos genes de CT. Ao compararmos com outros conjuntos de genes foram identificados novos genes de CT. Realizamos a integração de várias bases de dados de expressão gênica de tecidos normais e de tumores, para identificação desses genes. A integração de dados clínicos e de infiltração de células CD8+ no tumor, nos proporcionou a identificação de dezenas de genes de CT associados com prognóstico bom ou ruim. Para aqueles genes de CT relacionados ao bom prognóstico, mostramos em pacientes com câncer uma relação direta entre a expressão gênica do CT e um sinal de infiltração de células CD8+ para alguns tipos de tumores, especialmente melanoma. Adicionalmente, nesta tese contextualizamos a bioinformática em um cenário de big data.
ABSTRACT
Cancer/testis (CT) genes are excellent candidates for cancer immunotherapies because of their restrict expression in normal tissues and the capacity to elicit an immune response when expressed in tumor cells. In this study, we provide a genome-wide screen for CT genes with the identification of 745 putative CT genes. Comparison with a set of known CT genes shows that 201 new CT genes were identified. Integration of gene expression and clinical data led us to identify dozens of CT genes associated with either good or poor prognosis. For the CT genes related to good prognosis, we show that there is a direct relationship between CT gene expression and a signal for CD8+ cells infiltration for some tumor types, especially melanoma. In addition, we contextualized bioinformatics in a big data scenario.
LISTAS DE ABREVIAÇÕES
CG Cancer/Germline
CT Cancer/Testis
GTEx Genotype-Tissue Expression
HBM Human Body Map
HPA Human Protein Atlas
MHC Major Histocompatibility Complex NGS Next Generation Sequencing
PGH Projeto Genoma Humano
POP Procedimento Operacional Padrão
TCGA The Cancer Genome Atlas
LISTA DE APÊNDICES
APÊNDICE A - Trabalhos publicados
SUMÁRIO
1. INTRODUÇÃO ... 11
1.1. Sobre a bioinformática ... 11
1.2. NGS ... 11
1.3. Bioinformática na Era do Big Data ... 11
2. TCGA ... 12 2.1. Expressão Gênica ... 13 2.2. Dados clínicos ... 13 2.3. Dados de Mutação ... 13 2.3.1. mydbmutation ... 13 3. CÂNCER ... 14 4. IMUNOTERAPIA ... 15 5. TRABALHO PRINCIPAL ... 17 6. DISCUSSÃO ... 30 7. CONCLUSÃO ... 31 8. REFERÊNCIAS ... 32
APÊNDICE A - Trabalhos publicados ... 36
1. INTRODUÇÃO
1.1. Sobre a bioinformática
A bioinformática é uma ciência multidisciplinar caracterizada pela utilização de métodos computacionais e estatísticos para solucionar problemas da área biológica. Por ser uma ciência multidisciplinar a bioinformática pode ser aplicada para diversas finalidades.
Diversos projetos científicos em escala mundial, só obtiveram êxito com o auxílio da bioinformática. Por exemplo, o sequenciamento do primeiro genoma humano, disputado entre o consórcio Projeto Genoma Humano e a empresa Celera Genomics (LANDER et al. 2001; VENTER et al. 2001), que consolidou a bioinformática da maneira que a enxergamos hoje em dia.
Nesse período de descobertas de genomas, o Brasil também teve seu protagonismo, realizando o primeiro sequenciamento do genoma da bactéria Xylella fastidiosa (SIMPSON et al. 2000).
1.2. NGS
Anos depois da publicação do genoma humano, a bioinformática ganha força novamente com o surgimento das plataformas de sequenciamento, chamadas de Next-Generation Sequencing (NGS) ou sequenciadores de segunda geração. O ponto forte dos NGS está na sua capacidade de produzir uma grande quantidade de dados em pouco tempo, o que influenciou muito no crescimento da bioinformática.
Os NGS podem ser utilizados para diversas finalidades, como por exemplo, o sequenciamento de genomas complexos, exoma, identificação de variações no DNA, transcriptoma, metagenômica entre outras (GOODWIN et al. 2016).
O sucesso dos NGS deve-se aos sistemas computacionais embutidos nas próprias máquinas (EID et al. 2009; HARISMENDY et al. 2009; GOLPARIAN et al. 2018; MARGULIES et al. 2005; QUAIL et al. 2012), possibilitando o processamento das amostras de forma paralela, resultando numa maior produção de dados e uma redução no custo do sequenciamento.
1.3. Bioinformática na Era do Big Data
Nos últimos anos, a bioinformática vem se inserindo em um novo contexto, a área de “Big Data”. Este é um termo utilizado na área de Tecnologia da Informação (TI) para descrever
um grande volume de dados, tanto estruturados quanto não estruturados. Muitas técnicas computacionais, como algoritmos de inteligência artificial, técnicas de mineração de dados e aprendizagem de máquinas, são comumente aplicados no Big Data para se extrair informação ou conhecimento (KANTARDZIC, 2011).
A bioinformática adequa-se muito bem às três principais características do Big Data (Volume, Velocidade e Variedade). Podemos observá-las no grande volume de dados produzidos pelos NGS que cresce exponencialmente; na necessidade de análises com resultados rápido, tendo em vista, que muito dos dados investigados pela bioinformática são de pacientes em tratamento de alguma doença. E por fim, tratando-se da biologia, existe uma variedade enorme de dados que podem ser produzidos sobre um assunto específico, por exemplo, sobre o câncer.
O TCGA é um bom exemplo de base de dados utilizado pela bioinformática em um contexto de Big Data. Neste banco de dados podemos destacar sua grande variabilidade de dados, que possibilita estudar o câncer por diferentes pontos de vista, observando diversos fatores que atuam sobre ele. Outro ponto a se destacar, sobre os dados do TCGA é a possibilidade de aplicação de métodos computacionais como Data Mining e Machine Learning.
2. TCGA
O projeto The Cancer Genome Atlas (TCGA) iniciado em 2006, reuniu pesquisadores de diversas instituições em conjunto com o Instituto Nacional do Câncer (National Cancer Institute) e o Instituto Nacional de Pesquisa do Genoma Humano (National Human Genome Research Institute), com o objetivo de obter uma compreensão abrangente das alterações e funcionamento dos principais tipos de câncer (Cancer Genome Atlas Research Network, 2008). Atualmente o TCGA contém mais de 20.000 casos de pacientes, abrangendo 33 tipos de câncer. Desde sua criação o TCGA gerou mais de 2,5 petabytes de dados genômicos, epigenômicos, transcriptômicos e proteômicos, através de análises como as de mutação, CNV, expressão e relatórios clínicos. A maioria destes dados são disponibilizados de forma gratuita para comunidade científica, contribuindo para o avanço das pesquisas na área.
2.1. Expressão Gênica
O termo expressão gênica foi estabelecido para mensurar a quantidade de transcritos que estariam sendo produzido em uma célula ou conjunto de células, podendo ser oriundos de genes codificadores de proteinas ou de outras classes de genes.
A informação do DNA um organismo é comum a todas as células do individuo, sendo a expressão dessas informações responsavel para que células de um determinado técido sejam diferentes de outros técidos.
Os dados de expressão gênica são muito utilizados para analizar células cancerigenes, com essa informação é possível estudar mais sobre o câncer, por exemplo, em diferentes situações como em metástase ou pós-tratamento.
2.2. Dados clínicos
Uma das grandes questões ausentes quando se avalia uma doença são as informações clínicas dos pacientes envolvidos e embora pareçam ser informações triviais tem uma enorme importância científica, por isso é preciso ter-se muita atenção no momento em que são gerados. Como exemplo cito a análise de sobrevida, comumente aplicada a esses dados, tendo com objetivo de comparar eventos de morte em grupos de indivíduos ao longo do tempo, sendo representado pelo gráfico de Kaplan-Meier (Carvalho et al. 2011), ver figura 4 do artigo principal da tese.
2.3. Dados de Mutação
Nesta seção dou destaque aos dados das mutações observadas na molécula de DNA de cada paciente e predizer seu impacto.
Mutação são um dos principais fatores genéticos envolvidos na tumorigenesis. As mutações são alterações que ocorrem no DNA, podendo ocasionar impactos nas funções da célula, sendo comumente estudadas em doenças como o câncer.
2.3.1. mydbmutation
Os dados de mutação oriundo do banco de dados do TCGA são muito utilizados por nosso grupo (DE SOUZA et al. 2014; PUTNAM et al. 2016; SRIVATSAN et al. 2019), por isso, criou-se uma base de dados chamada de ‘dbmutation’, inicialmente anotada na versão do genoma hg19, com 16 tipos de tumores do TCGA e com dados de outras fontes, posteriormente foi atualizado para a versão dbmutation2015. Tendo em vista o aumento do número de
amostras no TCGA e mudança da versão do genoma e chamadores de mutação utilizados pelo TCGA, ficou sobre minha responsabilidade a atualização deste banco de mutação (mydbmutation2018), sobre o qual foi gerado um arquivo de Procedimento Operacional Padrão (POP), que descrevem passo a passo a execução do pipeline de processamento e elaboração do banco mydbmutation2018, contido no APÊNDICE B.
3. CÂNCER
O câncer é uma das doenças mais estudadas no mundo e pode ser definida como um conjunto de doenças que ocorre devido a uma desordem celular caracterizada pelo acúmulo progressivo de células diferentes do tecido onde se encontram. Essa multiplicação excessiva não é compensada pelo processo de morte celular normal, levando à formação de uma massa celular. Essas células cancerosas causam danos e invadem outros tecidos do corpo (LOWITZ and CASCIATO, 2012). Estima-se que, em 2018, tenham ocorrido cerca de 18,1 milhões de casos novos e 9,6 milhões de morte por câncer em todo o mundo (World Health Organization, 2019).
O câncer é uma doença de ordem multifatorial, desencadeada por alterações genéticas, epigenéticas e/ou ambientais, a qual apresenta elevada heterogeneidade tanto a nível histológico, quanto genômico (ANDERSON et al., 2014; PAQUET et al., 2014).
O processo de desenvolvimento tumoral, conhecido como carcinogênese, ocorre devido à interação de fatores externos e internos ao indivíduo (Institute of Medicine, 2002). Existem diversos fatores de risco que podem aumentar as chances do surgimento do câncer, como a exposição excessiva ao sol, uso de cigarros e bebidas, hábitos alimentares, doenças virais, entre outros. Nestes casos é possível adotar medidas preventivas para se evitar o surgimento do câncer, porém, fatores aleatórios que ocorrem em nosso genoma, como rearranjos genômicos e mutações aleatórias no DNA, nos impossibilitam de tomar ações preventivas. Desse modo, o câncer é considerado uma doença complexa (INCA, 2018), dependente de mutações germinativas e somáticas. A carcinogênese pode durar anos, ocorrendo em diferentes etapas e levando à formação do tumor (DARK, 2013).
O sucesso da carcinogênese depende das capacidades adquiridas pelas células cancerosas para prevalecerem no tecido. Nesse processo, elas devem ser capazes de gerar mais mutações somáticas por instabilidades genômicas, além de resistir à morte celular, apresentar a energia celular desregulada e a sustentação da sinalização proliferativa, conseguindo também escapar da ação de genes supressores de crescimento celular, evitar a destruição imune,
habilitar a imortalidade replicativa, levar à inflamação promotora tumoral, ativar invasão e metástase, além de induzir a angiogênese (HANAHAN and WEINBERG, 2011).
Para evitar que erros no DNA se propaguem, nossas células possuem um mecanismo chamado de “sistema de reparo”, composto por genes cuja função é corrigir erros ocorridos durante momentos de replicação, transcrição ou pós- transcricionais. O sistema de reparo do DNA é essencial para estabilidade celular. Mutações em genes de reparo podem influenciar na evolução e crescimento do tumor, como demonstrado em um de nossos trabalhos (SRIVATSAN et al. 2019)
4. IMUNOTERAPIA
Nos últimos anos, os tratamentos para o câncer baseado em imunoterapia vêm ganhando destaque. Este método consiste em utilizar o sistema imune do próprio paciente para combater as células tumorais.
O sistema imune é composto por células e moléculas responsáveis por dar uma resposta coletiva e coordenada a entrada de substâncias estranhas, denominada de resposta imune (ABBAS et al. 2015).
Antígenos são substâncias estranhas que induzem as respostas imunes específicas ou são reconhecidas pelos linfócitos ou anticorpos (Rezaei and KESHAVARZ-FATHI, 2018; YADDANAPUDI, MITCHELL and EATON, 2013; SIMPSON et al. 2005). Os tumores também podem induzir uma resposta do sistema imo, através de antígenos tumorais.
Anticorpos, moléculas do complexo maior de histocompatibilidade (MHC) e receptores de antígeno da célula T, são as três classes de moléculas usadas pelo sistema imune adaptativo.
Quando o tumor produz antígenos tumorais, estes podem ser processados e seus peptídeos reconhecidos pelas moléculas do MHC classe I, que irá expor o peptídeo a superfície celular para serem reconhecidas pelos linfócitos T CD8+ citotóxico (CTL), produzindo uma resposta imune contra o tumor, ilustrado na Figura1.
Conceitos relacionados à resposta imune ao tumor não é algo novo, Macfarlane Burnet na década de 1950, propôs o conceito de vigilância imune do câncer, afirmando que uma função fisiológica do sistema imune é a de reconhecer e destruir clones de células transformadas antes que se transformem em tumores e também de eliminar tumores depois de formados (ABBAS et al., 2015).
Os tumores induzidos por vírus oncogênicos geralmente induzem fortes respostas imunes, devido as proteínas virais serem antígenos estranhos, porém, os tumores espontâneos induzem uma resposta mais fraca, devido as células tumorais serem muitos semelhantes a células normais. Além do mais, mutações no tumor podem reduzir a capacidade de estimular fortes respostas imunes (LAKINS et al. 2018).
A imunoterapia tem se tornado uma abordagem eficaz para o tratamento de muitos tipos de câncer (LYON et al. 2017). O principal desafio para desenvolvimento de vacinas está na identificação de bons alvos terapêuticos (BANCHEREAU and PALUCKA 2018; HU et al. 2018).
Os cancer-germline (CG) genes, tem se demonstrado excelentes candidatos para alvos imunoterapêuticos. Os CG são genes com expressão predominante no tecido normal de tecidos
Figura 1: Representação simplificada do mecanismo de reconhecimento dos antígenos tumorais pelas células CD8+. Nesta figura vemos uma célula cancerígena expressando antígenos tumorais que serão degradadas pelo proteossoma, gerando os peptídeos que são reconhecidos pelas moléculas MHC classe 1, que irão conduzir os peptídeos até a superfície da célula, sendo reconhecidos pelas células CD8+ do sistema imune, através da ligação com o TCR.
germinativos e também expressos em tumores (VIGNERON, 2015; AKERS et al. 2010; SIMPSON et al. 2005; KOSLOWSKI et al. 2004). Os cancer/testis (CT) genes, são a classe mais promissora do CG, com vários ensaios clínicos em andamento (KRISHNADAS et al. 2015). Os CT genes apresentam expressão exclusiva ou predominantemente em tecido normal de testículo, mas também são expressos em uma variedade de tipos de tumores (SCANLAN et al. 2004).
Respostas imune foram observadas para vários CT antígenos em pacientes de vários tipos de tumores (WANG et al. 2016; AKCAKANAT et al. 2004; GNJATIC et al. 2003), tornando-os bons candidatos para o desenvolvimento de vacinas contra o câncer ou em combinação com terapias citotóxicas convencionais (AYYOUB, ZIPPELIUS, et al. 2003; AYYOUB, RIMOLDI, et al. 2003; AYYOUB, BREHM, et al. 2003; TSUJI et al. 2009; GJERSTORFF et al. 2015).
Neste contexto encontra-se o trabalho principal da minha tese, nós realizamos uma análise com todos os genes humanos para identificar potenciais CT genes, resultando na identificação de 745 CTs putativos, dentre estes, 201 não classificados previamente. Integramos dados de expressão gênica, dados clínicos e infiltração de células CD8+, associando-os a prognostico bom ou ruim.
Oncotarget
92966
www.impactjournals.com/oncotarget
Genome-wide identification of cancer/testis genes and their
association with prognosis in a pan-cancer analysis
Vandeclecio Lira da Silva1,2,3,*, André Faustino Fonseca1,2,3,*, Marbella Fonseca1,
Thayna Emilia da Silva1, Ana Carolina Coelho1,3, José Eduardo Kroll1,3,4, Jorge
Estefano Santana de Souza3,5, Beatriz Stransky3,6 Gustavo Antonio de Souza1,3 and
Sandro José de Souza1,3
1Instituto do Cérebro, UFRN, Natal, Brazil
2Ph.D. Program in Bioinformatics, UFRN, Natal, Brazil
3Bioinformatics Multidisciplinary Environment (BioME), Digital Metropolis Institute, UFRN, Natal, Brazil 4Instituto de Bioinformática e Biotecnologia, Natal, Brazil
5Instituto Metrópole Digital, UFRN, Natal, Brazil
6Departmento de Engenharia Biomédica, UFRN, Natal, Brazil *These authors have contributed equally to this work
Correspondence to: Sandro José de Souza, email: [email protected] Keywords: cancer/testis, cancer antigens, biomarkers, prognosis
Received: June 03, 2017 Accepted: August 17, 2017 Published: October 10, 2017
Copyright: Silva et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License 3.0
(CC BY 3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
ABSTRACT
Cancer/testis (CT) genes are excellent candidates for cancer immunotherapies because of their restrict expression in normal tissues and the capacity to elicit an immune response when expressed in tumor cells. In this study, we provide a genome-wide screen for CT genes with the identification of 745 putative CT genes. Comparison with a set of known CT genes shows that 201 new CT genes were identified. Integration of gene expression and clinical data led us to identify dozens of CT genes associated with either good or poor prognosis. For the CT genes related to good prognosis, we show that there is a direct relationship between CT gene expression and a signal for CD8+ cells infiltration for some tumor types, especially melanoma.
INTRODUCTION
Genes with a restricted pattern of expression in normal tissues and expression in tumors cells are excellent candidates for biomarkers and therapeutic targets. Among these genes, cancer/testis (CT) genes are the most promising with several clinical trials under way [1]. These CT genes are exclusive or predominantly expressed in testis among normal tissues but are also expressed in a variety of tumor types [2]. Few authors have suggested that these genes may be considered as cancer-germline (CG) genes, as they can also present regular expression in ovary and placenta [3, 4]. When protein products of CT genes elicit an immune response, they are called CT antigens [3]. Methodologies, like SEREX and protein arrays, have identified several CT genes as CT antigens.
Cellular and humoral immune responses have been observed for many CT antigens in patients bearing a variety of tumor types [5-7], making them good candidate targets for cancer immunotherapy, like cancer vaccination, adoptive T-cell transfer with chimeric T-cell receptors or in combination with conventional cytotoxic therapies [8-12].
Although there is a lack of consensual classification of CT genes, authors have divided them into two distinct groups [13]. The first one is organized in multigene families generally located on the X chromosome, where they comprise approximately 10% of the genes on that chromosome (CT-X genes) and show heterogeneous expression in cancer tissues that increase during tumor evolution and can elicit immune responses in cancer patients. The other group comprises the non-X CT genes since they are located on autosomal chromosomes. This last class does not appear to exist as
Oncotarget
92967
www.impactjournals.com/oncotarget
antigen, with more than 200 members, were gathered into a database (http://www.cta.lncc.br/). This CT Database provides information about CT genes including names and aliases, genomic location among others, but has not incorporated data from new technologies and contains less than 250 CT genes [15]. Moreover, several other genome-wide analyses have been reported. Hofmann et al. [14], for example, used a combination of CAGE, MPSS and RT-PCR to perform a survey of cancer/testis genes. They identified more than 30 candidate CT genes. More recently, our group characterized the human surfaceome and identified 14 putative CT genes coding for cell surface proteins [16]. Using public microarray data, two genes FMR1NB and TMEM31 were characterized as CT genes coding for cell surface proteins, which render them excellent candidates for targeted therapies [16]. One of them, FMR1NB, was found to elicit immune response in sarcoma patients [17]. While the project reported here was underway, Wang et al. [5] reported a systematic identification of CT genes in 19 tumor types.
The development of next-generation sequencing (NGS) technologies catalyzed a series of projects whose primary objective was to genetically characterize a cohort of cancer patients and associate that information with clinical data. The most successful of these projects was “The Cancer Genome Atlas” (TCGA) that integrated information for more than 11,000 patients from a variety of tumor types. NGS technologies have also allowed a better characterization of the human transcriptome derived from healthy cells and tissues. Projects like the Human Body Map (HBM) (GEO accession: GSE30611), the Genotype-Tissue Expression (GTEx) [18] and the Human Protein Atlas (HPA) [19] have deep sequenced the human transcriptome of dozens of cell types.
While the genome-wide screens performed so far were necessary for a better characterization of the universe of CT genes, most of them were executed at a time in which these NGS technologies were not widely available. Yao [20] have recently used TCGA data to explore CT genes. They, however, restricted their analysis to a subset of known CT genes. On the other hand, Wang et al. [5] used TCGA data to perform a genome-wide screen for CT genes. In the present paper, we have integrated information derived from both HBM/GTEx/HPA and TCGA to provide a complete genome survey of CT genes with the characterization of 201 new putative CT genes. By using mass-spectrometry data, it is shown that several of our putative CT genes exist at the protein level in some tumor types. Finally, CT genes associated with either a good or poor outcome are identified. For some CT genes associated to good prognosis, an association to CD8+ cells infiltration is observed. We propose that CT genes whose expression is related to a good outcome are excellent candidates for immunotherapeutic approaches.
Identification of genes predominantly (or exclusively) expressed in testis
Assuming that CT genes are predominant or exclusively expressed in testis when compared to the remaining tissues, we used RNA-Seq data from three different sources to identify genes with expression bias to the testis. Absolute transcript level in each tissue was converted to a proportional score (transcript level in a tissue divided by the sum of levels in all tissues), and a threshold of at least 0.9 was used to select genes preferentially or exclusively expressed in testis. A threshold of 0.9 would imply that at least 90% of all expression of that gene in all analyzed tissues was derived from testis. HBM (GEO accession: GSE30611), GTEx [18] and the HPA [19] datasets, all reporting RNA-Seq data from a variety of normal tissues, were used to select genes preferentially (or exclusively) expressed in testis. The resulting selection, detailed in Figure 1A, contained a set of 1,103 non-redundant genes based on the three sources used in our analysis (Figure 1B and Supplementary Table 1 for a complete list of testis-biased genes). A gene ontology analysis in this gene set showed, as expected, a strong bias towards biological processes related to germline cells (Figure 1C).
To make available a set of putative CT genes identified using more stringent criteria, we modified our pipeline to select genes with a proportion score of at least 0.99. This selection generated a list of 793 genes with an exclusive expression in testis. While we have used the set of 1,103 genes in the remaining analysis, one can identify the more restricted set of testis-specific genes in Supplementary Table 1.
Identification of putative CT genes
The expression pattern in tumors of the 1,103 genes predominantly expressed in testis was evaluated using the TCGA dataset. RNA-Seq data from 6,221 tumor samples were collected from the TCGA data repository comprising 15 tumor types. To identify genes significantly expressed in a given tumor, we used statistics provided by TCGA itself. A gene was considered a putative CT gene if it had a level of expression (cutoff threshold of RSEM >1) in at least 10% of all informative samples for a given tumor. These two criteria identified 745 putative CT genes (201 as new CT genes by comparing to the CT Database and data from Wang et al. [5]) significantly expressed in at least one tumor type (Figure 2A). If a more stringent threshold is used (15% of all samples for a given tumor type), we found 678 putative CT genes (176 as new CT genes) (Figure 2A). The same procedure was
Oncotarget
92968
www.impactjournals.com/oncotarget
(155 new) and 418 (132 new) putative CT genes using the 10% and 15% thresholds, respectively (Figure 2A). Supplementary Table 2 lists all genes with their corresponding expression in all analyzed tumor types. A comparative analysis between all tumor types, shown in Figure 2B, reveals that among the putative CT genes one can find several that are tumor-specific. In several cases, a particular CT gene showed a significant expression in only one tumor (17% of all CT genes). For example, 32 CT genes were expressed exclusively in leukemia, 11 in melanoma, and 14 in ovarian cancer (Figure 2B). This finding demonstrated the potential of our strategy for proposing particular biomarkers and targets for different types of tumors.
Much has been discussed on the role of CT genes in driving tumorigenesis [12]. Based on that, we decided to explore our set of putative CT genes regarding their status as cancer genes in the TCGA dataset used here. A method recently developed by us, the S-score [21], integrates information such as mutation screening, methylation status, copy-number variation and expression profiling and was used to infer whether our CT candidates showed a pattern of either an oncogene or tumor suppressor. The S-score was calculated for all 1,103 genes preferentially expressed in testis for all 15 tumor types studied here. When a stringent cut-off was used, as in De Souza et al. [21] (S-score >3, indicating an oncogene, or <-3, showing a tumor suppressor), we found 313 cancer genes (128
Figure 1: Identification of putative CT genes in 15 tumor types. (A) Schematic view of the pipeline used to identify CT genes with the identification of 1,103 genes predominant or exclusively expressed in testis and 745 putative CT genes. (B) Venn diagram showing the intersection of the 1103 genes predominant or exclusively expressed in testis regarding their source. (C) Gene Ontology enrichment analysis using the set of 1,103 genes predominant or exclusively expressed in testis.
Oncotarget
92969
www.impactjournals.com/oncotarget
differently in distinct tumor types. Supplementary Table 2 lists the S-score values for all testis-biased genes in all 15 tumor types used here.
To evaluate whether the set of putative CT genes is enriched or depleted of cancer genes in any tumor type, a Monte Carlo simulation was performed using random sets of 1,103 genes. As shown in Figure 3, there is an overall depletion of oncogenes in this set of testis-biased genes (pattern found in 12 out of 15 tumor types). Only five
and SKCM). No oncogene enrichment was found. The same pattern, general depletion of cancer genes in the set of CT genes, was also observed when we used the set of 745 putative CT genes in the simulations (Supplementary Figure 1). This finding strongly suggests that, while there are several cancer genes within the CT genes dataset, there is overall a depletion of cancer genes in that set except for SKCM and LUAD, both presenting enrichment for tumor suppressors.
Figure 2: Comparative analysis of putative CT genes. (A) Number of testis-biased genes and putative CT genes according to different stringent criteria for selection. (B) Matrix reporting the number of shared CT genes for all possible paired tumor type. The list at the right is the number of exclusive CT genes per tumor.
Oncotarget
92970
www.impactjournals.com/oncotarget
To serve either as a biomarker or a therapeutic target, a given CT gene should be expressed at the protein level. Mass spectrometry-based strategies are becoming powerful resources to query the proteome, and the golden standards technologies are now able to identify around 10,000 proteins per sample. We capitalize on the availability of mass spectrometry data from different sources for a variety of tumor samples to determine among the putative CT genes those expressed at the protein level.
Using a cohort of 209 samples from different tumor types (04 melanoma, 95 colorectal tumors, 40 breast tumors, 36 prostate tumors and 34 ovary tumors), we identified 136 putative CT genes at the protein level (Supplementary Table 3). Although this is a significant number, it should be emphasized that a high rate of false-negatives is expected in this analysis due to the non-exhaustive nature of mass spectrometry-based approaches. By plotting the sum of the area under curve measurements of all identified proteins in each sample group (Supplementary Figure 2), it is possible to observe that many of these CT gene products belong to the top
A comparison of the Top 15 most abundant CT genes in the proteomic dataset shows that some genes are more globally present in most or all tumors represented here, such as PBK, SPATA22, IL4I1, HIST1H1A, among other. Some genes seem to show a more specific profile depending on the tumor type, such as C17ORF104, only highly-abundant in the melanoma samples, FUT5 in colon, PAGE1 in ovarian tumor and CSNK1A1L in prostate. CT signature for cancer prognosis
Clinical data regarding overall survival from TCGA patients was used to evaluate the impact of expression of a putative CT gene in the outcome of the corresponding patients. A computer program evaluated all putative CT genes (745) for all 15 tumor types regarding any survival difference between samples expressing (FPKM or RSEM >1) or not expressing a given CT gene. Supplementary Table 4 provides the raw data for this survival analysis for all tumor types. Genes reporting a q-value < 0.05 (as defined by a log-rank test) for the difference in survival between the groups of patients expressing or
Figure 3: Enrichment analysis of CT genes for oncogenes and tumor suppressors. Box plots represent the distribution of the number of cancer genes in the 10,000 Monte Carlo simulations. Red star indicates the true number of cancer genes in the set of CT genes for each respective tumor type. Upper, middle and lower panels correspond to the real and simulated sets for oncogenes and suppressors together, oncogenes and suppressors, respectively.
Oncotarget
92971
www.impactjournals.com/oncotarget
whose expression affected the outcome of patients in at least one tumor type. Expression of CT genes was more associated with poor prognosis (179 genes) than to better prognosis (57 genes). Few genes behave differently in distinct tumor types. We confirmed the finding from Yao et al. [20] who found a high frequency of CT genes affecting prognosis in kidney renal clear cell carcinoma (KIRC). Most of the CT genes in KIRC were associated with poor prognosis (129 out of 141 genes). A more even distribution of CT genes associated to good and poor prognosis were found for LGG (30 and 32 CT genes, respectively) and SKCM (5 and 5 CT genes, respectively). To further select potential candidates, we split samples based on the expression of a given CT gene in three categories: no expression, expression below the median and expression above the median. If the expression of a corresponding CT gene was truly associated with patient outcome, we would expect that patients expressing more of the corresponding CT gene would have a stronger survival effect. A manual inspection of the Kaplan-Meyer plots for all 236 CT genes (179 and 57 associated to poor and good prognosis, respectively) was performed looking for the above pattern, and 113 genes were found more strongly associated with patient outcome (illustrative plots for KIRC and SKCM are shown in Figure 4).
Data from Senbabaoglu et al. [22], who also used expression data from TCGA, was then used to assess the association between expression of a given CT gene and the number of infiltrating CD8+ cells. Senbabaoglu et al. [22] developed a method that uses expression data from TCGA to estimate the number of imune cells infiltration in a given tumor. Our analysis was only possible for four tumor types (BLCA, HNSC, LUAD and SKCM) due to either the lack of a significant number of CT genes associated with good prognosis or the lack of data of infiltrating CD8+ cells for some tumors. For two tumor types, LGG and KIRC, the high number of CT genes associated to both, good and poor prognosis, rendered the comparison impossible since both groups of samples were almost identical (since they comprised the totality of the samples). Figure 5A shows that samples with high expression of CT genes associated with good prognosis have a significantly higher number of CD8+ cells in BLCA (p < 9.9e-5), HNSC (p<9.5e-4) and SKCM (1.1e-6).
When we split the samples according to the expression of a given CT gene, for most of the genes associated with good prognosis there is a significant association between the CT gene expression and infiltration of CD8+ cells for SKCM (Figure 5B), HNSC (Figure 5C) and BLCA (Figure 5D). A scatchard plot (showing the association between expression and CD8+ infiltration) for the same genes is shown in Supplementary Figure 3. Stronger associations were obtained for ZNF683 in BLCA, GPR31 in HNSC and C5ORF58, GTSF1L, HSF5 and HEATR9 in SKCM.
We have capitalized on the availability of NGS data to perform one the largest screening for CT genes so far. One first critical issue, as discussed by [5] and [14], is the classification of CT genes based on its expression pattern in normal tissues. These genes exhibit different expression profiles and can be categorized into testis-restricted, testis/ brain-restricted, or testis-selective group. This pattern imposes a challenge to identify the CT genes that are most suitable for the development of cancer therapies. Hofmann et al. [14], for example, performed an expression survey of 153 known CT genes and showed that only 39 had an exclusive expression in testis while the remaining had at least some expression in other tissues. Our screening of genes with a testis-biased expression involved the use of two proportional scores with different levels of stringency (0.9 and 0.99). While we used the less stringent dataset in most of the remaining analysis, the more stringent dataset is available to the community (Supplementary Table 1).
Several reports in the literature have indicated that CTs are mainly expressed in lung, ovarian, bladder, breast tumors and especially melanoma [3]. Most of the studies so far have used approaches based on the interrogation of few genes although Hofmann [14] used a series of high-throughput gene expression analyses to validate putative CT genes and Wang et al. [5] have used TCGA data from 19 tumor types. TCGA has provided a unique opportunity to screen for putative CT genes in a large panel of samples from many different tumor types and associated clinical features. Yao [20] have used TCGA data to explore the pan-cancer expression landscape of CT genes restricting their analysis to a subset of the CT database. Here, again, we used two thresholds with different levels of stringency (expression in at least 10% or 15% of samples in a given tumor) to identify putative CT genes. This allowed us to identify 745 putative CT genes (using the 10% threshold). By comparing this set of CT genes with the catalogs from the CT database and from Wang et al. [5] we found 201 new CT genes in our dataset. Several CT genes showed expression in only one tumor, which demonstrate the potential of our analysis for proposing biomarkers and targets for particular types of tumor.
The CT gene catalog generated by this study allowed us to evaluate some questions regarding CT genes. For example, much has been discussed on the role of CT genes in driving tumorigenesis [12]. One of the arguments that support a more important role of CT genes in tumorigenesis is related to their functions, many of which are related to tumorigeneses, like signal transduction and gene regulation. Another indirect support comes from an apparent similarity between germ cell and tumor developments [12, 23, 24]. In that line, some authors [25, 26] proposed that the expression of CT genes, usually restricted to germline cells, would trigger a gametogenic program in other somatic cells that would
Oncotarget
92972
www.impactjournals.com/oncotarget
undifferentiated state of stem cells as reported by Cheng
et al. [27] and Lifantsenva et al. [28]. Nevertheless, few CT45A1 was shown to work as an oncogene and drive tumorigenesis in breast cancer [29]. Deletion of SSX2 in
Figure 4: Kaplan-Meyer plots for representative CT genes. (A) Most representative plots for SKCM including four genes associated to good prognosis. (B) Most representative plots for KIRC including four genes associated to poor prognosis.
Oncotarget
92973
www.impactjournals.com/oncotarget
in cancer cell survival, frequently acting in the p53
pathway [31, 32]. Involvement with cancer progression, evaluated for any potential enrichment of oncogenes and tumor suppressor, using the S-score method developed
Figure 5: Correlation between expression of CT genes and the number of infiltrating CD8+ cells. (A) Samples with higher expression for CT genes associated to good prognosis have a higher rate of infiltrating CD8+ cells for BLCA, HNSC and SKCM. (B) Rate of infiltrating CD8+ cells for samples with no, low or high expression for the respective CT genes associated to good prognosis in SKCM.
(C) Rate of infiltrating CD8+ cells for samples with no, low or high expression for the respective CT genes associated to good prognosis
in HNSC. (D) Rate of infiltrating CD8+ cells for samples with no, low or high expression for the respective CT gene associated to good prognosis in BLCA.
Oncotarget
92974
www.impactjournals.com/oncotarget
finding strongly suggests that, while there are several cancer genes within the CT genes dataset, there is overall a depletion of cancer genes in that set except for melanoma and lung adenocarcinoma, both presenting enrichment for tumor suppressors.
Although few CT genes have been associated with either poor or good prognosis in a variety of tumor types [34-37], no large-scale analysis has been performed with a complete set of putative CT genes. Djureinovic et al. [38] generated a wide list of putative CT genes in non-small-cell lung cancer and found no gene associated to prognosis. Yao et al. [20] have recently shown that although some CT genes from the CT Database are associated with patient outcome, not many are independent prognostic markers. We explore this issue in a systematic way through an exhaustive analysis on the association between the expression of CT genes and cancer prognosis. Dozens of CT genes were associated with either good or poor prognosis. The robustness of our method is shown by the identification of genes clearly associated with disease progression among the set of 113 CT genes more strongly related to patient outcome by our analysis. Included in this set of genes are TEX101 [39], HORMAD2 [40], OTP [41] and TEX19 [42]. It is interesting to notice that CT genes are not significantly enriched for genes associated to prognosis (data not shown).
It is tempting to speculate that the CT genes associated to a better prognosis are eliciting an immune response against the tumor, which could be the reason for a better outcome in such patients. ROPN1 has been demonstrated to induce autoantibodies in patients with prostate cancer [43] and multiple myeloma [44]. Spontaneous tumor immune response was also detected for SPAG6 in sera from patients with gastric cancer, melanoma and prostate cancer [45]. CTAGE1 antibodies were also found in sera of colorectal cancer patients [46].
The finding that expression of several CT genes is associated with good prognosis led us to hypothesize that this effect could be the result of infiltrating CD8+ cells driven by the CT gene expression. Several methods have been developed that evaluate the intra-tumor immune landscape based on gene expression analysis of the bulk tumor [22, 47, 48]. Data from Senbabaoglu et al. [22] is quite suitable for our analysis since they used their method in most of the tumor types evaluated by us in this report. We found several CT genes, especially in SKCM, whose expression is significantly associated to both good prognosis and CD8+ cell infiltration. We suggest that these CT genes be considered for further studies that would evaluate their immunotherapeutic potential.
RNA-Seq data source
Four datasets were used for the identification of CT genes, including three from normal tissues, obtained from the Expression Atlas Portal [18]: the Human Body Map (GEO accession: GSE30611), GTEx [18] and Human Protein Atlas (HPA) [19]. The 16 samples from the Human Body Map were processed using our pipeline [49]. Expression values for GTEx and HPA datasets were obtained directly from the projects web page as FPKM and TPM values, respectively. Data from 15 tumor types from the TCGA consortium was used for the identification of putative CT genes. Expression, methylation and GISTIC CNV data were obtained from the cBIO portal by using the CGDS-R package, which provides processed data for each tumor type. Furthermore, somatic mutation data from COSMIC [50] and a local compilation of all somatic mutations found in the literature [49, 50] were used. For each sample in each tumor type, an expression threshold equal to 1 RSEM was applied to separate samples based on the expression of the putative CTs. Only genes with expression in more than 10% or 15% of samples were considered CTs for a particular tumor.
Enrichment ontology analysis
The R package “clusterProfiler” version 3.3 [51] was used to perform the ontology enrichment analysis based on Gene Ontology (GO) with a hyper-geometric test and correction method of Benjamini-Hochberg (BH), with cutoff parameters of p-value < 0.05 and q-value < 0.05. To remove redundancy of enriched GO terms the function “simplify” with default parameters was used.
Proteomic analysis of public cancer datasets The following MS raw file datasets were downloaded from ProteomeXchange: ovarian cancer dataset PXD003668 [52]; breast cancer dataset PXD002619 [53]; melanoma dataset PXD001724 [54]; colon cancer dataset PXD002041-50 [55]; prostate cancer dataset (PXD003430, PXD003452, PXD003515, PXD004132, PXD003615, PXD003636, PXD004159) [56]. All datasets were submitted to MaxQuant software version 1.5.2.8 [57] for protein identification. Parameters were set as follows: protein N-acetylation and methionine oxidation as variable modifications; carbamidomethylation of cysteine as fixed modification; first search error window of 20 ppm and main search error of 6 ppm at MS level. Furthermore, trypsin without proline restriction enzyme option was used, with two allowed mis-cleavages. Minimal unique peptides were set to 1, and FDR allowed was 0.01 (1%) for peptide and protein identification. The
Oncotarget
92975
www.impactjournals.com/oncotarget
assign FDR rates. A contaminants filter was performed, removing all occurrences presents on columns “Reverse” or “Potential contaminant” from the output of MaxQuant. S-score simulation
Identification of cancer genes was performed using the S-score metric [21] in both the set of testis-enriched genes (1103) and the set of 745 putative CTs. The Monte Carlo simulation was performed against each tumor type (with extreme S-score), where 10.000 simulated sets were compared to the real sets. In this step, three different tests were carried out: enrichment for oncogenes (genes with S-score ≥ 3), enrichment for tumor suppressor (genes with S-score ≤ -3) and enrichment for cancer genes (including both oncogenes and tumor suppressors).
Survival signatures and patients prognosis
To test the association of CT genes with patient outcome in a given tumor type all putative CTs expressed in at least 30 samples were used. All putative CTs were tested individually using a log-rank test and genes were selected based on a threshold (q-value ≤ 0.05), as defined by the “qvalue” R package [58], and classified as associated with “Good” or “Poor” prognosis. Next, samples expressing a given CT associated with prognosis were separated in two subsets based on a median expression of the corresponding CT gene. Kaplan-Meyer curves were plotted using the ggplot2 (from the R package). CD8+ profiling for TCGA samples was obtained from Senbabaoglu et al. [22].
Abbreviations
CT: Cancer/testis; CG: cancer-germline; NGS: next-generation sequencing; TCGA: The Cancer Genome Atlas; HBM: Human Body Map; GTEx: The Genotype-Tissue Expression; KIRC: kidney renal clear cell carcinoma; SKCM: skin cutaneous melanoma; GO: Gene Ontology.
ACKNOWLEDGMENTS AND FUNDING
This work was supported by a CAPES grant (23038.004629/2014-19) to SJS, by the Ludwig Institute for Cancer Research to SJS, and by the Institute of Bioinformatics and Biotechnology to SJS. Data analyses were performed on supercomputers from the Digital Metropolis Institute at the Federal University of Rio Grande do Norte.
The authors declare that they have no conflicts of interest.
REFERENCES
1. Krishnadas DK, Shusterman S, Bai F, Diller L, Sullivan JE, Cheerva AC, George RE, Lucas KG. A phase I trial combining decitabine/dendritic cell vaccine targeting MAGE-A1, MAGE-A3 and NY-ESO-1 for children with relapsed or therapy-refractory neuroblastoma and sarcoma. Cancer Immunol Immunother. 2015; 64:1251–60.
2. Scanlan MJ, Simpson AJG, Old LJ. The cancer/testis genes: Review, standardization, and commentary. Cancer Immun. 2004; 4:1–15.
3. Simpson AJG, Caballero OL, Jungbluth A, Chen YT, Old LJ. Cancer/testis antigens, gametogenesis and cancer. Nat Rev Cancer. 2005; 5:615–25.
4. Koslowski M, Bell C, Seitz G, Lehr H, Roemer K, Mu H. Frequent Nonrandom Activation of Germ-Line Genes in Human Cancer. Cancer Res. 2004; 64:5988–93.
5. Wang C, Gu Y, Zhang K, Xie K, Zhu M, Dai N, Jiang Y, Guo X, Liu M, Dai J, Wu L, Jin G, Ma H, et al. Systematic identification of genes with a cancer-testis expression pattern in 19 cancer types. Nat Commun. 2016; 7:10499. 6. Akcakanat A, Kanda T, Koyama Y, Watanabe M, Kimura
E, Yoshida Y, Komukai S, Nakagawa S, Odani S, Fujii H, Hatakeyama K. NY-ESO-1 expression and its serum immunoreactivity in esophageal cancer. Cancer Chemother Pharmacol. 2004; 54:95–100.
7. Gnjatic S, Atanackovic D, Jäger E, Matsuo M, Selvakumar A, Altorki NK, Maki RG, Dupont B, Ritter G, Chen YT, Knuth A, Old LJ. Survey of naturally occurring CD4+ T cell responses against NY-ESO-1 in cancer patients: correlation with antibody responses. Proc Natl Acad Sci U S A. 2003; 100:8862–7.
8. Ayyoub M, Zippelius A, Pittet MJ, Rimoldi D, Valmori D, Cerottini JC, Romero P, Lejeune F, Liénard D, Speiser DE. Activation of human melanoma reactive CD8+ T cells by vaccination with an immunogenic peptide analog derived from Melan-A/melanoma antigen recognized by T cells-1. Clin Cancer Res. 2003; 9:669–77.
9. Ayyoub M, Rimoldi D, Guillaume P, Romero P, Cerottini JC, Valmori D, Speiser D. Tumor-reactive, SSX-2-specific CD8+ T cells are selectively expanded during immune responses to antigen-expressing tumors in melanoma patients. Cancer Res. 2003; 63:5601–6.
10. Ayyoub M, Brehm M, Metthez G, Talbot S, Dutoit V, Taub RN, Keohan ML, Gure AO, Chen YT, Williamson B, Jungbluth AA, Old LJ, Hesdorffer CS, et al. SSX antigens
Oncotarget
92976
www.impactjournals.com/oncotarget
11. Tsuji T, Altorki NK, Ritter G, Old LJ, Gnjatic S. Characterization of Preexisting MAGE-A3-Specific CD4+ T Cells in Cancer Patients and Healthy Individuals and Their Activation by Protein Vaccination. J Immunol. 2009; 183:4800–8.
12. Gjerstorff MF, Andersen MH, Ditzel HJ. Oncogenic cancer/testis antigens: prime candidates for immunotherapy. Oncotarget. 2015; 6:15772–87. https://doi.org/10.18632/oncotarget.4694. 13. Scanlan MJ, Gure AO, Jungbluth AA, Old LJ, Chen YT.
Cancer/testis antigens: an expanding family of targets for cancer immunotherapy. Immunol Rev. 2002; 188:22–32. 14. Hofmann O, Caballero OL, Stevenson BJ, Chen YT, Cohen
T, Chua R, Maher CA, Panji S, Schaefer U, Kruger A, Lehvaslaiho M, Carninci P, Hayashizaki Y, et al. Genome-wide analysis of cancer/testis gene expression. Proc Natl Acad Sci U S A. 2008; 105:20422–7.
15. Almeida LG, Sakabe NJ, de Oliveira AR, Silva MCC, Mundstein AS, Cohen T, Chen YT, Chua R, Gurung S, Gnjatic S, Jungbluth AA, Caballero OL, Bairoch A, et al. CTdatabase: A knowledge-base of high-throughput and curated data on cancer-testis antigens. Nucleic Acids Res. 2009; 37:2007–10.
16. da Cunha JPC, Galante PAF, de Souza JE, de Souza RF, Carvalho PM, Ohara DT, Moura RP, Oba-Shinja SM, Marie SKN, Silva WA, Perez RO, Stransky B, Pieprzyk M, et al. Bioinformatics construction of the human cell surfaceome. Proc Natl Acad Sci. 2009; 106:16752–7.
17. Lee SY, Obata Y, Yoshida M, Stockert E, Williamson B, Jungbluth AA, Chen YT, Old LJ, Scanlan MJ. Immunomic analysis of human sarcoma. Proc Natl Acad Sci U S A. 2003; 100:2651–6.
18. Petryszak R, Keays M, Tang YA, Fonseca NA, Barrera E, Burdett T, Füllgrabe A, Fuentes AMP, Jupp S, Koskinen S, Mannion O, Huerta L, Megy K, et al. Expression Atlas update - An integrated database of gene and protein expression in humans, animals and plants. Nucleic Acids Res. 2016; 44:D746–52.
19. Uhlen M, Fagerberg L, Hallstrom BM, Lindskog C, Oksvold P, Mardinoglu A, Sivertsson A, Kampf C, Sjostedt E, Asplund A, Olsson I, Edlund K, Lundberg E, et al. Tissue-based map of the human proteome. Science. 2015; 347:1260419–1260419.
20. Yao J, Caballero OL, Yung WKA, Weinstein JN, Riggins GJ, Strausberg RL, Zhao Q. Tumor Subtype-Specific Cancer-Testis Antigens as Potential Biomarkers and Immunotherapeutic Targets for Cancers. Cancer Immunol Res. 2014; 2:371–9. 21. De Souza JES, Fonseca AF, Valieris R, Carraro DM, Wang
JYJ, Kolodner RD, De Souza SJ. S-score: A scoring system for the identification and prioritization of predicted cancer genes. PLoS One. 2014; 9:1–9.
22. Şenbabaoğlu Y, Gejman RS, Winer AG, Liu M, Van Allen EM, de Velasco G, Miao D, Ostrovnaya I, Drill E, Luna A, Weinhold
prognostic and immunotherapeutically relevant messenger RNA signatures. Genome Biol. 2016; 17:231.
23. Caballero OL, Chen YT. Cancer/testis (CT) antigens: Potential targets for immunotherapy. Cancer Sci. 2009; 100:2014–21.
24. Rousseaux S, Debernardi A, Jacquiau B, Vitte AL, Vesin A, Nagy-Mignotte H, Moro-Sibilot D, Brichon PY, Lantuejoul S, Hainaut P, Laffaire J, de Reynies A, Beer DG, et al. Ectopic Activation of Germline and Placental Genes Identifies Aggressive Metastasis-Prone Lung Cancers. Sci Transl Med. 2013; 5:186ra66–186ra66.
25. Old LJ. Cancer is a somatic cell pregnancy. Cancer Immun. 2007; 7:19.
26. Akers SN, Odunsi K, Karpf AR. Regulation of cancer germline antigen gene expression: implications for cancer immunotherapy. Futur Oncol. 2010; 6:717–32.
27. Cheng YH, Wong EW, Cheng CY. Cancer/testis (CT) antigens, carcinogenesis and spermatogenesis. Spermatogenesis. 2011; 1:209–20.
28. Lifantseva N, Koltsova A, Krylova T, Yakovleva T, Poljanskaya G, Gordeeva O. Expression Patterns of Cancer-Testis Antigens in Human Embryonic Stem Cells and Their Cell Derivatives Indicate Lineage Tracks. Stem Cells Int. 2011; 2011:1–13.
29. Shang B, Gao A, Pan Y, Zhang G, Tu J, Zhou Y, Yang P, Cao Z, Wei Q, Ding Y, Zhang J, Zhao Y, Zhou Q. CT45A1 acts as a new proto-oncogene to trigger tumorigenesis and cancer metastasis. Cell Death Dis. 2014; 5:e1285.
30. Greve KB, Lindgreen JN, Terp MG, Pedersen CB, Schmidt S, Mollenhauer J, Kristensen SB, Andersen RS, Relster MM, Ditzel HJ, Gjerstorff MF. Ectopic expression of cancer/testis antigen SSX2 induces DNA damage and promotes genomic instability. Mol Oncol. 2015; 9:437–49. 31. Yang B, O’Herrin SM, Wu J, Reagan-Shaw S, Ma Y, Bhat
KMR, Gravekamp C, Setaluri V, Peters N, Hoffmann FM, Peng H, Ivanov AV, Simpson AJG, et al. MAGE-A, mMage-b, and MAGE-C proteins form complexes with KAP1 and suppress p53-dependent apoptosis in MAGE-positive cell lines. Cancer Res. 2007; 67:9954–62.
32. Marcar L, MacLaine NJ, Hupp TR, Meek DW. Mage-A cancer/testis antigens inhibit p53 function by blocking its interaction with chromatin. Cancer Res. 2010; 70:10362–70. 33. Caballero OL, Shousha S, Zhao Q, Simpson AJG, Coombes
RC, Neville AM. Expression of Cancer/Testis genes in ductal carcinoma in situ and benign lesions of the breast. Oncoscience. 2013; 1:14–20. https://doi.org/10.18632/ oncoscience.4.
34. Sang M, Gu L, Yin D, Liu F, Lian Y, Zhang X, Liu S, Huang W, Wu Y, Shan B. MAGE-A family expression is correlated with poor survival of patients with lung adenocarcinoma: a retrospective clinical study based on tissue microarray. J Clin Pathol. 2016.
Oncotarget
92977
www.impactjournals.com/oncotarget
Survival of Lung Adenocarcinoma Patients Predicted from Expression of PD-L1, Galectin-9, and XAGE1 (GAGED2a) on Tumor Cells and Tumor-Infiltrating T Cells. Cancer Immunol Res. 2016; 4:1049–60.
36. Liu S, Sang M, Xu Y, Gu L, Liu F, Shan B. Expression of MAGE-A1, -A9, -A11 in laryngeal squamous cell carcinoma and their prognostic significance: a retrospective clinical study. Acta Otolaryngol. 2016; 136:506–13. 37. Ohue Y, Wada H, Oka M, Nakayama E. Antibody response
to cancer/testis (CT) antigens: A prognostic marker in cancer patients. Oncoimmunology. 2014; 3:e970032. 38. Djureinovic D, Hallström BM, Horie M, Mattsson JSM,
La Fleur L, Fagerberg L, Brunnström H, Lindskog C, Madjar K, Rahnenführer J, Ekman S, Ståhle E, Koyi H, et al. Profiling cancer testis antigens in non-small-cell lung cancer. JCI insight. 2016; 1:e86837.
39. Yoshitake H, Yokoi H, Ishikawa H, Maruyama M, Endo S, Nojima M, Yoshida K, Yoshikawa H, Suzuki F, Takamori K, Fujiwara H, Araki Y. Overexpression of TEX101, a potential novel cancer marker, in head and neck squamous cell carcinoma. Cancer Biomark. 2013; 12:141–8.
40. Zhang K, Tang S, Cao S, Hu L, Pan Y, Ma H, Guo X, Wu S, Shen H, Hu Z. Association of polymorphisms at HORMAD2 and prognosis in advanced non-small-cell lung cancer patients. Cancer Epidemiol. 2014; 38:414–8. 41. Swarts DRA, Henfling MER, Van Neste L, Van Suylen
RJ, Dingemans AMC, Dinjens WNM, Haesevoets A, Rudelius M, Thunnissen E, Volante M, Van Criekinge W, Van Engeland M, Ramaekers FCS, et al. CD44 and OTP are strong prognostic markers for pulmonary carcinoids. Clin Cancer Res. 2013; 19:2197–207.
42. Zhong J, Chen Y, Liao X, Li J, Wang H, Wu C, Zou X, Yang G, Shi J, Luo L, Liu L, Deng J, Tang A. Testis expressed 19 is a novel cancer-testis antigen expressed in bladder cancer. Tumor Biol. 2016; 37:7757–65.
43. Adeola HA, Smith M, Kaestner L, Blackburn JM, Zerbini LF. Novel potential serological prostate cancer biomarkers using CT100+ cancer antigen microarray platform in a multi-cultural South African cohort. Oncotarget. 2016; 7:13945–64. https://doi.org/10.18632/oncotarget.7359. 44. Chiriva-Internati M, Mirandola L, Yu Y, Jenkins MR, Gornati
R, Bernardini G, Gioia M, Chiaramonte R, Cannon MJ, Kast WM, Cobos E. Cancer testis antigen, ropporin, is a potential target for multiple myeloma immunotherapy. J Immunother. 2011; 34:490–9.
45. Siliņa K, Zayakin P, Kalniņa Z, Ivanova L, Meistere I, Endzeliņš E, Ābols A, Stengrēvics A, Leja M, Ducena K, Kozirovskis V, Linē A. Sperm-associated Antigens as Targets for Cancer Immunotherapy. J Immunother. 2011; 34:28–44.
46. Gerhardt A, Usener D, Keese M, Sturm J, Schadendorf D, Eichmüller S. Tissue expression and sero-reactivity of
47. Newman AM, Liu CL, Green MR, Gentles AJ, Feng W, Xu Y, Hoang CD, Diehn M, Alizadeh AA. Robust enumeration of cell subsets from tissue expression profiles. Nat Methods. 2015; 12:1–10.
48. Gong L, Parikh S, Rosenthal PJ, Greenhouse B. Biochemical and immunological mechanisms by which sickle cell trait protects against malaria. Malar J. 2013; 12:317.
49. Fonseca AL, da Silva VL, da Fonsêca MM, Meira ITJ, da Silva TE, Kroll JE, Ribeiro-dos-Santos AM, Freitas CR, Furtado R, de Souza JE, Stransky B, de Souza SJ. Bioinformatics Analysis of the Human Surfaceome Reveals New Targets for a Variety of Tumor Types. Int J Genomics. 2016; 2016:1–7.
50. Forbes SA, Beare D, Boutselakis H, Bamford S, Bindal N, Tate J, Cole CG, Ward S, Dawson E, Ponting L, Stefancsik R, Harsha B, Kok CY, et al. COSMIC: Somatic cancer genetics at high-resolution. Nucleic Acids Res. 2017; 45:D777–83.
51. Yu G, Wang LG, Han Y, He QY. clusterProfiler: an R Package for Comparing Biological Themes Among Gene Clusters. Omi A J Integr Biol. 2012; 16:284–7.
52. Coscia F, Watters KM, Curtis M, Eckert MA, Chiang CY, Tyanova S, Montag A, Lastra RR, Lengyel E, Mann M. Integrative proteomic profiling of ovarian cancer cell lines reveals precursor cell associated proteins and functional status. Nat Commun. 2016; 7:12645.
53. Tyanova S, Albrechtsen R, Kronqvist P, Cox J, Mann M, Geiger T. Proteomic maps of breast cancer subtypes. Nat Commun. 2016; 7:10259.
54. Welinder C, Pawłowski K, Sugihara Y, Yakovleva M, Jönsson G, Ingvar C, Lundgren L, Baldetorp B, Olsson H, Rezeli M, Jansson B, Laurell T, Fehniger T, et al. A protein deep sequencing evaluation of metastatic melanoma tissues. PLoS One. 2015; 10:1–16.
55. Zhang B, Wang J, Wang X, Zhu J, Liu Q, Shi Z, Chambers MC, Zimmerman LJ, Shaddox KF, Kim S, Davies SR, Wang S, Wang P, et al. Proteogenomic characterization of human colon and rectal cancer. Nature. 2014; 513:382–7. 56. Iglesias-Gato D, Wikström P, Tyanova S, Lavallee C,
Thysell E, Carlsson J, Hägglöf C, Cox J, Andrén O, Stattin P, Egevad L, Widmark A, Bjartell A, et al. The Proteome of Primary Prostate Cancer. Eur Urol. 2016; 69:942–52. 57. Cox J, Mann M. MaxQuant enables high peptide
identification rates, individualized p.p.b.-range mass accuracies and proteome-wide protein quantification. Nat Biotechnol. 2008; 26:1367–72.
58. Storey JD, Bass AJ, Dabney A, Robinson D. qvalue: Q-value estimation for false discovery rate control. R package version 2.6.0. Bioconductor; 2015. Available from http://github.com/jdstorey/qvalue.
6. DISCUSSÃO
No artigo principal desta Tese, realizamos a identificação de CT genes, baseado em seu padrão de expressão específico ou predominante em tecido de testículo (HOFMANN et al. 2008). Em nossa abordagem utilizamos três bases de dados de RNA-Seq de tecidos normais, para selecionar os genes com proporção de expressão de (0.9 e 0.99) em testículo, em relação com os demais tecidos analisados. Foram identificados 745 CT genes putativos, destes, 210 previamente não classificados como CTs, quando comparado contra outras bases de dados. Em nosso pipeline utilizamos diferentes filtros de seleção, tornando mais restringente a seleção de CT genes putativos.
Um ponto levantado neste trabalho foi sobre a importância dos CT no processo de tumorigenesis, argumentado por alguns autores (GJERSTORFF et al. 2015; CABALLERO and CHEN, 2009; OLD, 2007; AKERS et al. 2010). Nossa lista de CT genes foi avaliada quanto ao potencial enriquecimento de oncogenes e supressor de tumor, usando o método S-score desenvolvido por nosso grupo (DE SOUZA et al. 2014).
Descobrimos que, em geral, há uma depleção de genes associados a câncer no conjunto de CT genes analisados. Sugerindo que, embora existam vários genes associados a câncer na lista dos CT genes, em conjunto eles não mostram enriquecimento, exceto melanoma e adenocarcinoma de pulmão, ambos apresentando enriquecimento para supressores de tumor.
Embora poucos CT genes tenham sido associados a um prognóstico ruim ou bom em uma variedade de tipos de tumores (SANG et al. 2017; OHUE et al. 2016; LIU et al. 2016; OHUE et al. 2014), nenhuma análise em larga escala foi realizada com um conjunto completo de CT genes. Djureinovic et al. (DJUREINOVIC et al. 2016) geraram uma ampla lista de genes putativos de CT no câncer de pulmão de células não pequenas e não encontraram nenhum gene associado ao prognóstico. Exploramos essa questão de maneira sistemática através de uma análise exaustiva da associação entre a expressão dos CT genes e o prognóstico do câncer. Dezenas de CT genes foram associados a um prognóstico bom ou ruim. A robustez do nosso método é demonstrada pela identificação de genes claramente associados à progressão da doença entre o conjunto de 113 genes de CT mais fortemente relacionados ao resultado do paciente por nossa análise. Incluídos neste conjunto de genes estão TEX101 (YOSHITAKE et al. 2012), HORMAD2 (ZHANG et al. 2014), OTP (SWARTS et al. 2013) e TEX19 (ZHONG et al. 2016). É interessante ressaltar que os CT genes não são significativamente enriquecidos para os genes associados ao prognóstico.
A descoberta de que a expressão de vários CT genes está associada a um bom prognóstico nos levou a supor que esse efeito poderia ser o resultado da infiltração de células CD8+ impulsionadas pela expressão do CT gene. Foi demonstrado que o ROPN1 induz anticorpos em pacientes com câncer de próstata (ADEOLA et al. 2016) e mieloma múltiplo (CHIRIVA-INTERNATI et al. 2011). Também foi detectada resposta imune espontânea do tumor para SPAG6 em soros de pacientes com câncer gástrico, melanoma e câncer de próstata (SILINA et al. 2011) e anticorpos CTAGE1 também foram encontrados em soros de pacientes com câncer colorretal (GERHARDT et al. 2004). Encontramos vários CT genes, especialmente em SKCM, cuja expressão está significativamente associada ao bom prognóstico e à infiltração de células CD8+. Sugerimos que esses CT genes sejam considerados para estudos futuros que avaliam seu potencial imunoterapêutico.
7. CONCLUSÃO
A integração de diferentes bancos de dados públicos utilizados neste trabalho é um ponto importante a se destacar, aqui nós utilizamos 5 bases de dados distintos, contendo cada, dezenas de diferentes tipos de amostras, nos fazendo enxergar a bioinformática em um contexto de Big Data.
Devido a constante utilização da base de dados de mutação do TCGA, por alunos da bioinformática e membros do BioME, elaborei um pipeline para atualização e processamento deste banco, ver APÊNDICE B.
No artigo principal desta tese conseguimos realizar a identificação em escala genômica de genes de CT, dentre estes novos ainda não descritos na literatura, com um grande potencial para serem utilizados como biomarcadores de câncer. Além disso, encontramos uma associação entre os genes de CT com prognósticos de pacientes. Os resultados mais empolgantes em nossos analises foi encontrar que a expressão de genes de CT poderia estar induzindo uma resposta imune de células CD8+ que poderiam estar dando um bom prognóstico aos pacientes.
8. REFERÊNCIAS
ABBAS, Abul K.; LICHTMAN, Andrew H. H.; PILLAI, Shiv. Imunologia Celular e Molecular. [s.l.]: Elsevier Brasil, 2015.
ADEOLA, Henry A. et al. Novel potential serological prostate cancer biomarkers using CT100 cancer antigen microarray platform in a multi-cultural South African cohort. Oncotarget, v. 7, n. 12, 2016. Disponível em: <http://dx.doi.org/10.18632/oncotarget.7359>.
AKCAKANAT, Argun et al. NY-ESO-1 expression and its serum immunoreactivity in esophageal cancer. Cancer chemotherapy and pharmacology, v. 54, n. 1, p. 95–100, 2004.
AKERS, Stacey N.; ODUNSI, Kunle; KARPF, Adam R. Regulation of cancer germline antigen gene expression: implications for cancer immunotherapy. Future Oncology, v. 6, n. 5, p. 717–732, 2010. Disponível em: <http://dx.doi.org/10.2217/fon.10.36>.
AYYOUB, Maha et al. SSX antigens as tumor vaccine targets in human sarcoma. Cancer immunity, v. 3, p. 13, 2003.
AYYOUB, Maha et al. Tumor-reactive, SSX-2-specific CD8+ T cells are selectively expanded during immune responses to antigen-expressing tumors in melanoma patients. Cancer research, v. 63, n. 17, p. 5601–5606, 2003.
AYYOUB, Maha et al. Activation of human melanoma reactive CD8+ T cells by vaccination with an immunogenic peptide analog derived from Melan-A/melanoma antigen recognized by T cells-1.
Clinical cancer research: an official journal of the American Association for Cancer Research, v.
9, n. 2, p. 669–677, 2003.
BANCHEREAU, Jacques; PALUCKA, Karolina. Immunotherapy: Cancer vaccines on the move.
Nature reviews. Clinical oncology, v. 15, n. 1, p. 9–10, 2018.
CABALLERO, Otavia L.; CHEN, Yao-Tseng. Cancer/testis (CT) antigens: Potential targets for immunotherapy. Cancer Science, v. 100, n. 11, p. 2014–2021, 2009. Disponível em: <http://dx.doi.org/10.1111/j.1349-7006.2009.01303.x>.
CANCER GENOME ATLAS RESEARCH NETWORK. Comprehensive genomic characterization defines human glioblastoma genes and core pathways. Nature, v. 455, n. 7216, p. 1061–1068, 2008. CARVALHO, Marilia Sá et al. Análise de sobrevivência: teoria e aplicações em saúde. 2011. Disponível em: <http://dx.doi.org/10.7476/9788575413029>.
CHIRIVA-INTERNATI, Maurizio et al. Cancer testis antigen, ropporin, is a potential target for multiple myeloma immunotherapy. Journal of immunotherapy, v. 34, n. 6, p. 490–499, 2011.
DARK, Graham G. Oncology at a Glance. 1 ed. Editora Wiley 2013.
DE SOUZA, Jorge E. S. et al. S-score: a scoring system for the identification and prioritization of predicted cancer genes. PloS one, v. 9, n. 4, p. e94147, 2014.
DJUREINOVIC, Dijana et al. Profiling cancer testis antigens in non-small-cell lung cancer. JCI insight, v. 1, n. 10, p. e86837, 2016.
EID, John et al. Real-time DNA sequencing from single polymerase molecules. Science, v. 323, n. 5910, p. 133–138, 2009.
GERHARDT, Axel et al. Tissue expression and sero-reactivity of tumor-specific antigens in colorectal cancer. Cancer letters, v. 208, n. 2, p. 197–206, 2004.