Expression of evolutionarily novel genes in tumors

The evolutionarily novel genes originated through different molecular mechanisms are expressed in tumors. Sometimes the expression of evolutionarily novel genes in tumors is highly specific. Moreover positive selection of many human tumor-related genes in primate lineage suggests their involvement in the origin of new functions beneficial to organisms. It is suggested to consider the expression of evolutionarily young or novel genes in tumors as a new biological phenomenon, a phenomenon of TSEEN (tumor specifically expressed, evolutionarily novel) genes.


Background
Evolutionarily novel genes are those novel genes which originate in the germ cells of multicellular organisms and thus can participate in evolution. Genes that originate in somatic cells (e.g. in tumor cells) and cannot be passed to the progeny organisms are not considered as evolutionarily novel.
Novel genes can originate from pre-existing genes or de novo. The theory of the origin of novel genes is well developed and the mechanisms of the origin of evolutionarily novel genes are well understood and described [8,45,58,70,76,77,110,131,132,189,194,217]. But there is a question in which cells of the evolving multicellular organisms genes determining the evolutionary innovations and morphological novelties are expressed.
There is a general correlation between the increase in the gene number in the genomes of evolving organisms, from one side, and the increase in the number of cell types, the origin of other innovations and the overall complexity, on the other [34,91,215]. The question is how such adequate correlation was realized at the multicellular level. An adequate increase in cell number that accompanied the process of the origin of novel genes is hard to imagine. More likely, some autonomous cellular proliferative processes were recruited to provide the space for the expression of new genes.
In my previous publications [88][89][90] and in my recently published book "Evolution by Tumor Neofunctionalization" [91] I suggested that heritable tumorsbenign tumors or tumors at the early stages of progressionmay provide extra cell masses for expression of evolutionary novel genes and for emergence of evolutionary innovations and morphological novelties. The non-trivial prediction of this hypothesis is that we may find the expression of evolutionarily novel genes in tumors.
Experiments in this direction performed in my lab since early 2000s have indeed demonstrated the specific or predominant expression of many evolutionarily young or novel genes in tumors. These data will be discussed in the first part of this review.
I also found in the literature descriptions of many genes with similar dual specificitytumor specifically expressed, evolutionary novel. Such genes with dual specificity were not purposefully searched for by the authors and the connection of tumors and evolution was not emphasized. Rather, the data on evolutionary novelty and specificity of expression of certain genes were the result of descriptive experiments and often can be found among other described features of the studied genes. Similar information may be found in the results of genome-wide studies. Tumor specificity of expression of genes originated by gene duplication, from retrotransposons and endogenous retroviruses, by exon shuffling or de novo will be discussed in the second part of this review.

The purposeful experimental search for evolutionarily novel genes with tumor-specific expression
To study experimentally the prediction concerning the expression of evolutionarily young or novel genes in tumors we used two complementary approaches. One was to study the evolutionary novelty of genes/ sequences with proven tumor specificity of expression. The other was to study tumor specificity of expression of genes/sequences with proven evolutionary novelty. Both approaches found out genes/sequences with dual specificity, i.e. tumor-specifically or tumorpredominantly expressed and evolutionarily young or novel.

The evolutionary novelty of tumor-specifically expressed sequences
To find the sequences which are expressed in tumors but not in normal tissues the global comparison of cDNA sequences from all available tumor-derived libraries with cDNA sequences from all available normal tissue-derived libraries was performed. The normal EST set was subtracted in silico from the tumorous EST set [11].
The results showed that, in accordance with my prediction, tumors indeed express hundreds of sequences that are not expressed in normal tissues. About half of discovered tumor-specific sequences lack long reading frames (i.e., may be referred to non-coding RNAs) and defined function [11,51]. Among non-coding RNAs, the long non-coding RNA [94] and candidate microRNA (see ELFN1-AS1, a novel primate gene expressed predominantly in tumors) have been described.
The analysis of the relative evolutionary novelty of sequences retrieved from the paper [51] was performed. The protein-coding sequences were studied by Protein-Historian tool [28]. The nucleotide BLAST algorithm and the original Python script [3] were used to analyze the novelty of noncoding sequences. The orthologs of tumor-specifically expressed sequences described by Baranova and co-authors were searched in 26 completely sequenced eukaryotic and prokaryotic genomes. The curves of phylogenetic distribution of orthologs of these sequences have been generated. The data suggest that both sets of tumor-specifically expressed sequences are relatively evolutionary novel. The non-coding tumorspecifically expressed sequences are younger than protein-coding tumor-specifically expressed sequences. During last 39 million years of evolution, these sequences represented the youngest gene class in human ancestors' genomes [115,116].
In vitro experiments intended to confirm that the sequences found in silico are indeed specifically expressed in tumors were also carried out. cDNA panels from normal and tumor tissues were used for PCR with specific primers. In total, 56 sequences described in [11] have been studied in this way. Among them, nine were confirmed to be highly tumor-specific [94,95,138]. The sequences that have been confirmed to be tumor-specific are expressed in a vast variety of tumors. For example, the sequence Hs. 202247 is expressed in 46 tumor samples out of 56 examined and in none of 27 normal tissues. One of the protein products of the sequences that proved to be tumorspecific appeared to be a promising immunogen for antitumor vaccine development [138,170]. However, most of experimentally confirmed tumor-specific sequences appear to be non-coding RNAs.
The nine experimentally confirmed tumor-specific sequences were studied for their evolutionary novelty using molecular-biological techniques, comparative genomics analysis, the search for orthologous sequences and sequence conservation analysis [92,163,164]. Eight of the nine tumor-specifically expressed sequences are either evolutionarily new (primates or humans) or relatively young (mammals) ( Table 1) and evolve neutrally [92,93,[162][163][164]. I suggest to call such sequences Tumor-Specifically Expressed, Evolutionarily New Sequences, or TSEEN sequences.
The sequence Hs.285026 (HHLA1) contains ORF, although the corresponding protein is not shown experimentally. This sequence is similar to human de novo protein-coding genes [86]. As far as corresponding protein has not been shown, this sequence may represent the earlier stage of the novel gene origin comparing to those described by D.G. Knowles and A. McLysaght. This and other sequences described in our studies (besides protein-coding sequences with established functions) may represent proto-genes (gene precursors which have not yet acquired functions and evolve neutrally [29]) at different stages of their evolution towards novel genes with protein or RNA related functions. The sequence Hs.633957 represents this transition.

ELFN1-AS1, a novel primate gene expressed predominantly in tumors
The human transcribed locus resides in the 7th chromosome and corresponds to the UniGene EST cluster Hs.633957. It was found by our group to be expressed in a tumor-specific manner by in silico analysis [11]. Later these data were supported experimentally: specific transcripts of the locus were detected in tumors of various histological origins, but not in most of the healthy tissues [94,149,150].
Experimental and in silico evidence that locus is a stand-alone gene which has its own promoter and capability for alternative splicing was obtained. However, only one splicing isoform is predominant. The gene was assigned a gene symbol ELFN1-AS1, ELFN1 antisense RNA 1 (non-protein coding), gene name approved by  The use of the different tool results in six "primates", one "humans" and one "mammalia" in this column (Makashov, personal communication) Human Gene Nomenclature Committee. Our data point to the miRNA function of ELFN1-AS1 with DPYS mRNA being its primary target [151,152]. This gene originated de novo from an intronic region of a conservative gene ELFN1 (NCBI Ref. Seq. NM_001128636.2) in primate lineage. Homologous sequences of this gene were identified by us in all primates, but the DNA sequence from the representative of suborder Strepsirrhini Otolemur garnettii has more than 50 % differences from its human counterpart and forms an outgroup on the phylogenetic tree. Thus ELFN1-AS1 could become transcriptionally active after divergence of Strepsirrhini and Haplorhini primates. It is noteworthy that all the Haplorhini primates have a region with 5 or more E-boxes downstream of the DS site. This suggests that ELFN1-AS1 gene since its origin could be c-Mycresponsive.
Taken together, the data indicate that human transcribed locus contains a gene for some non-coding RNA, likely a microRNA. This gene combines features of predominant expression in tumors and evolutionary novelty [151,152].

PBOV1, de novo originated human gene with tumor-specific expression
In the study of PBOV1 gene the other approach was used, i.e. the evolutionary novelty of the gene was studied first.
PBOV1 (UROC28, UC28) is a human protein-coding gene with a 2501 bp single-exon mRNA and 135aa ORF. The gene has been originally characterized by An and co-workers [4]. This gene was mentioned among 12 human genes without orthologs in the mouse and dog genomes in the paper of Clamp and co-authors [38]. We studied the evolutionary novelty of this gene more carefully and found that the coding sequence of PBOV1 is poorly conserved in the mammalian evolution and originated de novo in primate evolution through a series of frame-shift and stop codon mutations. Consequently, 80 % of protein sequence is unique to humans. The Ka/Ks ratio both in pairwise alignments and in multiple alignment of all primate sequences syntenic to human coding sequence didn't show any significant differences from 1.0, indicating that the amino acid sequence evolved neutrally. PBOV1 protein lacks any annotated or predicted domains and over 60 % of its sequence is predicted to be disordered. These findings strongly suggest that human PBOV1 is a protein of a very recent de novo evolutionary origin [165].
After establishing the evolutionary novelty of PBOV1 gene, the specificity of its expression in tumors and normal tissues was studied. PBOV1 has been previously reported to be overexpressed in prostate, breast, and bladder cancers [4]. We studied the expression of PBOV1 using PCR on panels of cDNA from various normal and tumor tissues. The gene had a highly tumorspecific expression profile. It was expressed in 20 out of 34 tumors of various origins but was not expressed in any of the normal adult or fetal human tissues that we tested (Figs. 1 and 2). The interesting feature of this result is that tumor specificity of PBOV1 expression was predicted by us from its evolutionary novelty [96,165].
Unlike cancer/testis antigens genes PBOV1 is expressed from a GC-poor TATA-containing promoter which is not influenced by DNA methylation and is not active in testis. PBOV1 activation in tumors may depend  (12,14,17), seminoma (7), meningioma (63), non-Hodgkin lymphomas (67,82,92,102,113). From open access paper [165]. Copyright of authors on sex hormone receptors, C/EBP transcription factors and Hedgehog signaling pathway. Although the PBOV1 protein has recently originated de novo and thus has no identifiable structural or functional signatures, a missense SNP (single nucleotide polymorphism) in it has been previously associated with an increased risk of breast cancer. Using publicly available data we found that higher level of PBOV1 expression in breast cancer and glioma samples were significantly associated with a positive disease outcome. PBOV1 is also highly expressed in primary but not recurrent high-grade gliomas, suggesting that immunoediting against PBOV1-expressing cancer cells might occur over the course of disease. We propose that PBOV1 is a novel tumor suppressor gene which might act by provoking the cytotoxic immune response against cancer cells that express it. We speculate that this property might be a source of phenotypic feedback that facilitated PBOV1 gene fixation in human evolution [165].

The evolutionary novelty of human cancer/testis antigen genes
Cancer/testis antigen genes (CTA or CT genes) code for a subgroup of tumor antigens expressed predominantly in testis and different tumors. CT antigens may be also expressed in placenta, in female germ cells, and in the brain [33,64,175,209,210] (see discussion of CT genes expression in the brain in [91]).
At the time of the study, CTDatabase (http:// www.cta.lncc.br) included 265 CT genes and 149 CT gene families.
The hypothesis of the expression of evolutionarily novel genes in tumors explains this otherwise strange cancer-testis association paradox: as far as the origin of evolutionarily novel genes is connected with their expression in germ cells, cancer/testis genes are novel genes which are expressed in tumors.
So I suggested that cancer/testis antigen genes should be evolutionarily new or young genes. In order to prove this prediction, the presence of genes orthologous to human cancer-testis genes in human lineage was studied [44]. This analysis was performed separately for genes located on the X chromosome and autosomal cancer/testis genes, as far as extensive traffic of novel genes has been described for mammalian X chromosome [16,46,103].
Altogether 36.7 % of all human CT genes originated in Catarrhini, Homininae and humans. It was also found that 30 % of all human CT genes originated in Eutheria. These CT genes acquired functions in Eutheria. This indicates the importance of processes in which tumors and CT antigens were involved during the evolution of Eutheria. CT genes originated in Eutheria are located mainly on autosomes. CT genes originated in Catarrhini, Homininae and humans are located predominantly on X chromosome. This difference is probably related to important events in evolution of mammalian X chromosome since the origin of Eutheria [99], especially to the acquisition of a special role in the origin of novel genes [77].
Thus the majority of CT-X genes are either novel or young for humans, and majority of all human CT genes (>70 %) originated during or after the origin of Eutheria. These results suggest that the whole class of human CT genes is relatively evolutionarily new [44].
Our data are in good correspondence with evidence obtained by other groups on particular families of CT genes. I found the evidence in the literature that at least 7 families (of 149 families know by that time) of CT genes (MAGE-1, PRAME, SPANX-A/D, GAGE, XAGE, CT45 and CT47) and many CT genes located on the X chromosome (CT-X genes) were either new or young (reviewed in [91]. Later it was found that one more CT gene family, CTAGE (cutaneous T-cell-lymphoma-associated antigen) shows a rapid and primate specific expansion, especially in humans, which starts with an ancestral retroposition in the Haplorhini ancestor followed by DNA-based duplications [214]. But our study [44] was the first systematic study of the evolutionary novelty of the whole class of CT genes which showed that it is relatively evolutionarily novel. Thus our prediction of the evolutionary novelty of the whole class of CT genes turned out to be correct.
The relative evolutionary novelty of the whole class of CT genes confirms the prediction about expression of evolutionarily young and novel genes in tumors. The expression of cancer/testis genes in tumors thus appears as a natural phenomenon, not an aberrant process as interpreted by most of authors (e.g. [1,27,32,36,175,214]). More discussion of evolutionary novelty of CT genes may be found in my recent book [91].
The list of single genes and gene classes studied by our group at the Biomedical Center is presented in Table 1.
The data obtained by our group, both on individual genes and on large groups of genes, suggest that tumor specifically expressed, evolutionarily novel (TSEEN) genes could represent a new biological phenomenon, a phenomenon of TSEEN genes [91]. That is why I looked in the literature for the evidence about similar kind of genes, i.e. evolutionarily novel, tumor specifically expressed.

Analysis of the literature data related to TSEEN genes
It turned out that many examples of genes with dual specificity -evolutionarily novel, tumor specifically expressedcould be found in the literature but serious attention was never paid to this association. Below I will discuss the tumor specificity of expression of genes originated by different mechanisms -by gene duplication, from retrotransposons and endogenous retroviruses, by exon shuffling or de novo. As far as positive Darwinian selection is a feature of many evolutionarily novel genes, human tumor-related genes positively selected in primate lineage will be also discussed.

Expression of pseudogenes in tumors
Gene duplication is a major way of genome evolution. The original hypothesis [131] suggested that pre-existing genes are under control of natural selection, and their evolution is constrained within their existing function. The extra copy of existing gene gets out of control of the natural selection, so that accumulation of mutations in this extra copy may lead to the origin of a novel gene with related or even new function. Gene duplication is considered as providing the "row material" for the origin of new genes. This concept also suggests that the majority of duplicates becomes inactive pseudogenes due to degenerative mutations, and only rarely beneficial mutations would lead to the emergence of a new gene with a novel function [131]. But the term "pseudogene" was first introduced by C. Jacq and co-authors in 1977 [72].
The DNA-mediated mechanisms of gene duplication include unequal crossing over, tandem, segmental, chromosomal or genome duplications. The resulting gene duplicates may be organized in tandem, interspersed or polyploid manner. Segmental duplications are large interspersed segments of DNA with high sequence identity (>90 %), usually separated by >1Mb of unique sequences [120].
RNA-based gene duplication, or retroposition, creates duplicate genes by reverse transcription of RNAs from parental genes. RNAs from all categories generate retrosequences that may be exapted as novel genes or regulatory elements [21]. Retrogenes are most abundant in mammals where long interspersed nuclear elements (LINEs) that provide the enzyme reverse transcriptase for retroposition are widespread. The majority of retrogenes is produced by genes with high levels of germline expression. They often originate from the X chromosome [16,76]. A new retrogene is intronless, contains a poly(A) tract, and may be flanked by short duplicate sequences [15,104].
DNA-mediated gene duplication is more frequent event in genome evolution, while RNA-based gene duplication is more capable to generate genes with novel functions. The retroposition is less likely to provide expressed daughter retrocopies than segmental DNA duplication because retrocopies do not contain regulatory elements. So, new promoters and enhancers should somehow be recruited for the origin of new genes, and several mechanisms of such recruitment are described [76,77]. Retrogenes usually locate on chromosomes different from that of parental genes. Mammalian X chromosome demonstrates extensive retrogene traffic [46]. For reasons of different location and new promoter recruitment, the transcribed retrogenes are more capable to evolve new expression patterns and novel functional roles than gene duplicates arising by DNA segmental duplication [76,77]. Retrogenes, like duplicates originated through DNA-mediated mechanisms, might provide the raw material for the origin of evolutionarily novel genes and functionally important evolutionary innovations [76,119,197]. At least one functional retrogene per million years originated in primate lineage that led to humans [119].
In accordance with two major ways of gene duplication -DNA-based and RNA-based mechanismstwo types of pseudogenes are categorized as duplicated and processed pseudogenes, accordingly [105,148]. One more group of pseudogenes includes so called "unitary" pseudogenes that arise through spontaneous mutations of single coding genes [216]. Other pseudogene biotypes may include polymorphic pseudogenes (loci known to be coding in some individuals), IG pseudogenes (immunoglobulin segments with disabling mutations) and TR pseudogenes (T-cell receptor gene segments with disabling mutations) [147].
Hundreds to thousands of pseudogenes have been identified in different species. In humans, 11,216 pseudogenes have been recently annotated, including~8,000 processed pseudogenes [61,147]. The extrapolation estimates suggest that the number of pseudogenes in human genome may be~14,000 [147]. This is smaller than earlier estimates [190,217]. The processed pseudogenes are the most abundant type of pseudogenes in human genome which is connected with the burst of retroposition activity in ancestral primates [135,217]. Pseudogenes have long been considered as non-functional or "junk" DNA. But during the last decade, the attitude has changed substantially. The evidence is accumulating that many pseudogenes are transcribed and functional in development and diseases (reviewed in [105,148,154,173]. Laura Poliseno determines the following types of pseudogene functions: related to the parental gene and parental gene independent functions; mediated by the pseudogene DNA, by pseudogene RNA transcribed in sense, by pseudogene RNA transcribed in antisense, or by pseudogene-encoded proteins [154]. Pseudogenes transcribed as noncoding RNAs may regulate their parental genes as antisense RNAs, short interfering RNAs (siRNAs) or as microRNA decoys [173]. Pseudogenes participate in the regulation of variety of biological processes including cancer [105,148,154]. One of the earliest indications of the functional role of pseudogenes was demonstration that in mouse oocytes pseudogenederived small interfering RNAs regulate gene expression [188,204]. Besides fully functionally active pseudogenes, partially active pseudogenes in the process of either losing or gaining function are described [147].
The authors who study pseudogenes come to conclusion that pseudogenes serve as a source of novel functions for the evolving organisms [10,22,105]. A special term -"potogeneswas generated to designate pseudogenes as DNA sequences with a potentiality for becoming new genes [10,22]. This is in accordance with the major postulate of original hypothesis of evolution by gene duplication [131], and we may consider pseudogenes with novel or evolving functions as evolutionarily novel or evolving genes.
Transcription of pseudogenes is an important indication of their functionality. The evidence of pseudogenes transcription was accumulating during the last years [10,219]. The ENCODE and GENCODE projects provided information about transcription of 876 pseudogenes including 531 processed and 345 duplicated pseudogenes [147]. The other group of authors studied RNA-Seq transcriptome data from 248 cancer and 45 benign samples of 13 different tissue types and described the expression of 2,082 distinct pseudogenes [78]. What is important for our consideration of expression of evolutionarily novel genes in tumors, they observed 218 pseudogenes expressed only in cancer samples, of which 178 were observed in multiple cancers [78].
One of the first demonstrations that pseudogenes are activated in tumors was description of the new tumor antigen (NA88-A) generating an HLA class I-restricted CTL response against melanoma coded for by a processed pseudogene [126]. At the same time, the expression of parental gene HPX42B did not lead to similar CTL response. The transcription of NA88-A pseudogene was limited with significant expression found only in some metastatic melanomas [126].
Among other earlier works was detection of ψPTEN expression in central nervous system high-grade astrocytic tumors [211]. The ψPTEN expression was complementary to PTEN mutation because the majority of glioblastomas showed either PTEN mutation or ψPTEN expression. In the later study [153] the functional relationship between the mRNAs produced by the PTEN tumor suppressor gene and its pseudogene PTENP1 (the other name of ψPTEN) was demonstrated. PTENP1 was able to regulate cellular levels of PTEN and exerted a growth suppressive role acting as a microRNA-decoy [153].
In a comprehensive paper devoted to human processed pseudogenes Zhaolei Zhang and co-authors [217] described several pseudogene families with implication to tumors (see Table 5 in the above mentioned paper).
Other examples of pseudogenes expressed in tumors but not in normal tissues are presented in Table 2.
As we can see from the data presented in this part of the paper, the expression of pseudogenes in tumors is widespread. Thus the evolution of pseudogene towards functional novel gene may involve its expression in tumors as a part of the whole process (see [91] for more discussion of the role of gene expression in the origin of novel genes).

Endoretroviral sequences and other retrotransposons are expressed in tumors
Transposable elements are classified in two groups, Class I and Class II. Class I mobile elements use RNA intermediate and reverse transcriptase activity for transposition, while Class II elements use a DNA intermediate and a 'cut and paste' mechanism. Class I elements include long terminal repeat (LTR) retrotransposons, also called 'endogenous retroviruses' (ERVs), and nonlong terminal repeat (non-LTR) retrotransposons (LINEs and SINEs) [155]. Human transposable elements comprise about 40 % of the human genome: HERVs, 4.64 %; MaLR, 3.65 %; LINEs, 20.42 %; and SINEs, 13.14 % [100]. That is why mobile elements were called the "drivers of genome evolution" [83]. The role of transposons in gene origin was recently reviewed in [91].
Endogenous retroviruses (ERVs) have been shown to have originated as the result of repeated germ cell retroviral infection of their ancestral hosts [13,19,63,118,205]. The genes of ERVs were evolutionarily new for their ancestral hosts. Together with other retrotransposons, ERVs participated in the origin of genes with the novel functions to their hosts (reviewed in [91]). There are 203,000 copies of human ERVs (HERVs) in the human genome [100]. Different authors define different numbers of HERV families, from 26 [53] to about 50 [114,121] or even 350 families [136].
Recent reviews confirm the upregulation of HERVs in tumors [80,113,127,158,161], which is connected with general trend of HERVs demethylation in tumors [127,158], and similar data continue to accumulate [26,181,208]. ERVs of mice also demonstrate hypomethylation and transcriptional upregulation in mice tumors [66,112,158].
Endogenous retroviruses may serve as targets for antitumor immunity. For example, HERV-K-MEL, a HERV-K pseudogene expressed in most melanomas and in many other types of tumors, encodes the antigenic peptide that is targeted by CTLs in melanoma patients [30,169]. HERV-E was found to be selectively expressed in clear kidney cell cancer but not in normal tissues. This tumor-specific expression is connected with inactivation of the von Hippel-Lindau tumor suppressor and hypomethylation. Antigens encoded by HERV-E are immunogenic and stimulate cytotoxic T-cells that kill cancer cells. HERV proteins that act as tumorassociated antigens have also been detected in other types of tumors [37].
Especially interesting for my consideration is HERV-K family because it contains the most recently active members that entered the ancestral human genome after the divergence of humans and chimps and may be considered as evolutionarily novel for humans [12,13,185]. Many HERV-K proviruses are unique to humans [12]. HERV-K continued to replicate in human lineage until at least 250,000 years ago [114,117], and might still expand [113]. HERV-K is also most widely expressed in different tumors (see above). In HERV-K and in other younger families such as HERV-H and HERV-W the most pronounced DNA demethylation was reported [49,158]. Not only mRNA, but also HERV-K antibodies are already elevated in the blood at the early stage of breast cancer [202,203].
RNA transcripts from various HERV LTRs have been described in various types of human tumors and cell lines. For example, elevated HERV-K 5′LTR mRNA was detected in prostate cancer tissues (reviewed in [207]).
Other primate-specific retrotransposons such as SVA, LINE-1P, AluY, and MaLR families are also known for the loss of DNA methylation in tumors. The younger retroelements are highly methylated in healthy tissues, while in many tumors these young elements suffer the most dramatic loss of methylation [49,130,186]. L1 and Alu sequences are silenced in normal human cells and activated in tumors [14,155,171]. Full length L1 RNA in cancer cell lines and expression of ORF1p in tumors have been shown (reviewed in [130]). The majority of the retrotransposition events seem to be harmless "passenger" mutations [191].
There are in silico data supporting the increased transcription of retrotransposons in transformed human cells [41]. Although originally it was thought that HERVs are transcriptionally silent in most normal tissues, in silico [57,84,166,178] and PCR and microarray [6,50,140,174,179] data suggest that HERV-derived RNAs are more widely expressed in normal tissues than originally anticipated. HERV-K is transcribed during normal human embryogenesis [56]. Syncytin, the envelope gene of human defective endogenous retrovirus HERV-W, is expressed in multinucleated placental syncytiotrophoblasts and may mediate placental cytotrophoblast fusion [18,123,198].

Genes originated by exon shuffling are expressed in tumors and may lead to oncogenic transformation
The principle of gene origin by exon shuffling is the following: new genes are created by recombining previously existing exons that leads to the origin of mosaic genes and proteins [54,75,110,[141][142][143]. The exon shuffling is important mode of the origin of new genes: at least 19% of the exons in data base were involved in exon shuffling [109]. The correlation between exon-intron organization of the gene and the domain organization of the corresponding protein is most evident in the case of young vertebrate genes, e.g. genes coding for proteases of blood coagulation, fibrinolytic and complement cascades, etc. That is why the first evidence for exon shuffling came from studies on proteases of blood coagulation and fibrinolysis [143].
Modular domain rearrangements can lead to cancer. The fusion of the self-oligomerizing SAM domain from the gene TEL to the catalytic domain of the nonreceptor tyrosine kinase Abl in some human leukemias results in constitutively clustered chimeric protein, persistent activation of tyrosine kinase and oncogenic transformation. Tyrosine kinases other than Abl are also activated in fusion proteins by oligomerization of SAM domain of TEL [106]. Activation of Abl tyrosine kinase seen in patients with chronic myelogenous leukemia is caused by

Pseudogene description References
HERV-K-MEL Expressed in melanomas, sarcomas, lymphomas, bladder and breast cancers, but not in normal tissues [30,169] CYP4Z2P A pseudogene of mammaryrestricted cytochrome CYP4Z1, shows 4.8 times higher transcription level in breast cancer tissue than in normal mammary gland with almost no transcription in other tissues [157] ΨCx43 Connexin43 (CX43) pseudogene was shown to be expressed in mammary cancer cell lines and to inhibit growth. ΨCx43 acts as a posttranscriptional regulator of CX43 in breast cancer cells [17,79] rac1 The rac1 processed pseudogene overexpression was detected in the human brain tumors as compared to normal brain tissues [69] Oct4-pg5 Oct4-pg1 NANOGP8 Oct4 and Nanog, embryonic stem cell-specific genes coding for transcription factors, have multiple retropseudogenes, expressed in many cancer tissues but not in normal tissues. [71,73,137,184,213] POU5F1P1 POU5F1B OCT4-pg3 OCT4-pg4 A processed pseudogene POU5F1P1 (POU5F1 is another name of Oct4) is overexpressed in prostatic carcinoma, and OCT4-pg1, OCT4-pg3 and OCT4-pg4 were found to be expressed in glioma and breast carcinoma. POU5F1B (also known as Oct4-pg1 and POU5F1P1) is amplified and promotes aggressive phenotype to gastric cancer. [62,81,206,218] CRIPTO3 A presumed pseudogene expressed in many cancer samples [183] BRAF Expression of BRAF pseudogene was described in thyroid tumors. [220]

CSNK2A1P
Protein kinase CKα intronless gene CSNK2A1P, a presumed CK2α pseudogene, plays the oncogenic role in lung cancer. It is amplified and over-expressed in non-small cell lung cancer and leukemia cell lines and in lung cancer tissues. [67]

MYLKP1
A transcribed pseudogene MYLKP1 is strongly expressed in cancer cells and its parental gene MYLK is highly expressed in non-neoplastic cells. [60] MO1P3 SUMO1P3 pseudogene-expressed lncRNA is up-regulated in gastric cancer tissues and may be a potential biomarker in the diagnosis of gastric cancer [122] HMGA1P6 HMGA1P7 The overexpression of HMGA1 pseudogenes HMGA1P6 and HMGA1P7 is induced in human pituitary tumors and anaplastic thyroid carcinomas. [47,48] [25,98,187] ψPPM1K Transcribed pseudogene ψPPM1K exerts tumor-suppressor activity in hepatocellular carcinoma by generation of siRNA [31] translocation of the tip of chromosome 9 encoding Abl to chromosome 22 encoding BCR and formation of fusion protein. Oligomerization of coiled-coil domains from BCR leads to constitutive activation of Abl [106]. The Tre2(USP6) oncogene is a hominoid-specific gene. It originated by the fusion of two genes, USP32 (NY-REN-60) and TBC1D3. USP32 is an ancient gene and highly conserved. TBC1D3 is young and originated by recent segmental duplication in primates. Tre2 is young for humans as far as it originated 21-33 million years ago after TBC1D3 segmental duplication in primates [144].
Atypical splicing in combination with retrotransposition may also lead to exon shuffling. Moreover atypical splicing of existing genes may be the most prevalent mechanism of novel protein creation. Atypical splicing includes alternative splicing within the single-gene transcripts and intergenic splicing of transcripts from tandemly located genes. Transcription-induced chimeras may evolve into gene fusions, and alternative splicing may evolve to gene fission (reviewed in [8]). For instance, the chimeric PIPSL gene was formed by L1-mediated retrotransposition of a readthrough, intergenically spliced transcript in hominoids [9]. This phenomenon was called transcription-mediated gene fusion. Many examples of intergenic splicing have been described in the human genome. The authors suggest that it is a novel mechanism of gene origin, where transcription-induced chimerism followed by retroposition may result in new gene [2]. At least 4 %-5 % of the tandem gene pairs in the human genome can be transcribed into a single RNA coding for chimeric protein [139].
Alternative splicing often participates in exonization process. When the new exon is alternatively spliced and expressed at low levels, splice variants with and without new exon are represented, and the pre-existing function is not destroyed. This opens the way to the origin of new gene with a new function and/or new functional module due to novel exon [54,128,177,199]. The comparison of human, mouse and rat genomes indicates that alternative splicing is associated with an increased frequency of exon creation and/or loss [124].
Transposed element exonization may be a source of new constitutively spliced exons. Alu-containing exons are alternatively spliced. Comparative analysis of transposed element insertion within human and mouse genomes reveals Alu's unique role in shaping the human transcriptome [172,176].
The alternative splicing is widespread in cancer. The splice changes in cancer are global. Up to half of all alternative splicing events may be changed in tumors. Some splice isoforms are upregulated in all studied cancers, the others are characteristic to certain types of tumors. Affected proteins include transcription factors, cell signal transducers, transmembrane proteins, secreted extracellular proteins, proteins involved in metabolism, angiogenesis, apoptosis, cell motility and invasion, oncoproteins and tumor suppressor proteins. Genes with alternative transcripts associated with various cancers include CD44, p53, p73, PTEN, APC, BCL-X, VEGF4, mdm2, BRCA1, TACC1, TERT, KLF6, SURVIVIN, ASIP, NF1, Caspase 8, CDH17, Ron, BARD1, AR, FGFR2, RUNX1, HOXA9, WT1, BIM, TF, HERV-K env (np9), HNRPK and many others. Many of these genes have multiple splicing patterns, e.g. mdm2 gene locus produces over 72 mdm2 variants. Alternative splicing in cancer-related genes may have impact on all major aspects of tumor cell biology. All hallmarks of cancer have alternatively spliced regulators. There are also many cancerassociated splice variants with unknown functions [7,35,42,52,59,85,101,102,133,156,160,182,195,196].
Atypical splicing events do not alter the number of genes in DNA, but produce altered proteins which influence all aspects of tumor biology. In evolutionary perspective, atypical splicing combined with retrotransposition may lead to the origin of novel genes. The promising direction of research would be to study what proportion of spicing events involved in cancer have already generated (through retroposition) novel genes in the germ plasm.
Genes originated de novo are specifically expressed in tumors "Senseless" DNA sequences may acquire new functions in the organism and become new genes. New functions may be connected not only with protein-coding genes, but also with various functional non-coding RNAs. This mechanism of novel genes origin is called de novo origin.
New promoter elements such as GC-islands, TATAboxes, LINE1 promoters or retroviral LTRs may arise as a result of mutational process, gene rearrangements, retrotransposition or viral infection. Such events can lead to expression of "senseless" DNA sequences that subsequently may accumulate mutations that alter their protein-coding capacity. The senseless DNA sequences acquire new functions. Noncoding RNAs may eventually acquire ORFs and become protein-coding mRNAs. These could be mechanisms of de novo gene origin. Exonization by alternative splicing may be the mechanism of de novo exon origin (see discussion above in Genes originated by exon shuffling are expressed in tumors and may lead to oncogenic transformation).
Three novel human protein-coding genes have been shown to originate from noncoding DNA since the divergence with chimp. These genes have no proteincoding homologs in any other genome. Few humanspecific mutations altered protein-coding capacity by destroying "disablers" in the ancestral sequences. The existence of protein-coding genes is supported by expression and proteomic data [86]. One of those genes -CLLU1has been shown earlier to be specifically expressed in chronic lymphocytic leukemia (CLL) [23]. The CLL expression specificity of CLLU1 was later confirmed in several studies [24,74,134,159]. It was also shown that CLLU1 is expressed in other tumors (tumors of lung, stomach, prostate and spleen), but in no normal tissue [ [97], in press]. We may conclude that CLLU1 belongs to TSEEN genes.
PBOV1, a gene of the recent de novo origin specific to humans, has highly tumor-specific expression profile [165] (see discussion above in PBOV1, de novo originated human gene with tumor-specific expression).
PBOV1 expression levels positively correlate with relapse-free survival in breast cancer patients and with overall longitude of survival in glioma patients [165]. On the contrary, CLLU1 is highly expressed in poorprognostic patients [23,24,74,134,159].

Positive selection of human tumor-related genes in primate lineage
Positive Darwinian selection participates in the evolution of the novel genes. Comparison of the rate of amino acid replacement substitution with the rate of synonymous substitution, population genetic analyses of polymorphisms and the findings of convergent evolution support the adaptive evolution of the novel genes. There are many examples of rapidly evolving novel genes and gene families supported by positive selection. In humans, strong positive selection and accelerated evolution was documented for lactase gene and for many other genes with different molecular functions, e.g. transcription factors, genes involved in nuclear transport, DNA metabolism/cell cycle, protein metabolism, pigmentation pathways, dystrophin protein complex, heat shock proteins; various types of genes related to sensory perception, immune response, reproduction, morphology, hostpathogen interactions, and neuronal functions. Examples of positively selected gene families are also numerous, including those in African great apes and hominids. Several gene families have expanded or contracted rapidly in primates, including brain-related families in humans. Many of such families show evidence for positive selection. The proportion of positively selected genes is significantly higher in younger genes in humans, i.e. positive selection may play a role in faster evolution of younger genes. Many examples of rapid evolution and positive selection of new genes described in the literature points out that this phenomenon is widespread. It supports involvement of novel genes and gene families in adaptation and speciation and in evolution and enhancement of new functions (reviewed in [91]).
For our consideration, it is important that positive selection in primate lineage was described for many human tumor-related genes [39,40,43,129,145,180].
SPANX, GAGE, PRAME and CTAGE families of cancer/testis antigen genes, with unknown functions yet, undergo positive selection in primate evolution [43,55,87,108,214]. Comparison of human/chimp orthologues of CT-X genes has shown that they diverge faster and undergo stronger positive selection than those on the autosomes [180].
Adaptive evolution of the tumor suppressor BRCA1 in humans and chimps was demonstrated [68]. Most of the internal BRCA1 sequence is variable between primates and evolved under positive selection [145].
Angiogenin (ANG) is the tumor-growth promoter due to its ability to stimulate the formation of new blood vessels. Its expression is elevated in variety of tumors. The study among several primate species showed that ANG gene has a significantly higher rate of nucleotide substitution at nonsynonymous site than at synonymous sites, an indication of positive selection [212].
Comparison of 7645 chimp gene sequences with their human and mouse orthologs showed accelerated evolution in functions related to oncogenesis [39]. A search for positively selected genes in the genomes of humans and chimps showed the evidence for positive selection in many genes involved in tumor suppression, apoptosis and cell cycle control [129].
More examples of positively selected tumor-related genes are reviewed in [40].
Positive selection of many human tumor-related genes in the evolution of primates confirms the prediction of evolution by tumor neofunctionalization hypothesis concerning expression of evolutionarily new genes in tumors and selection for their new organismal functions. If an evolutionarily new gene is expressed in tumors, or a sequence that is expressed in tumors acquires a function beneficial to the organism and becomes an evolutionarily new gene, selection of organisms for the enhancement of the new function should take place, as predicted by the hypothesis. This is exactly what was found in papers discussed above: the positive selection of genes and proteins in different primate groups, not the somatic evolution of tumor cells. More discussion of positive selection in relation to the possible evolutionary role of tumors may be found in [91].
The paradox of the positive selection of many tumorassociated genes is difficult to explain otherwise than by the postulation that tumors play a positive evolutionary role. The other attempt to explain positive selection of tumor-related genes is based on the concept of genomic conflict and antagonistic coevolution [40,129].
Some evolutionarily novel genes are cellular oncogenes. The Tre2(USP6) oncogene is a hominoid-specific gene [144] (see discussion above in part 2.3). Evolutionarily novel genes CT45A1, TBC1D3 and NCYM may act like oncogenes (reviewed in [215]). Y. Zhang and M. Long suggest that these genes may also assume other biological functions, and attract the selection, pleiotropy and compensation hypothesis of M. Pavlicev and G.P. Wagner [146] to explain the paradox related to their oncogene role.

Conclusion
The phenomenon of tumor specifically expressed, evolutionarily novel genes (TSEEN genes) This review discusses the data obtained in my lab and the data described in the literature. My group looked for genes with dual specificity, i.e. evolutionarily novel and tumor specifically expressed. We studied single genes, the complex class of CT genes with many gene families, and two newly described gene classes obtained by global subtraction of normal cDNA sequences from tumor cDNA sequences. Using different approaches, we have been able to describe many genes with tumor specific or tumor predominant expression which are also evolutionarily novel or young.
We have also described tumor-specifically expressed, evolutionarily new sequences which look like protogenes, i.e. gene precursors which have not yet acquired functions and evolve neutrally. Expression of protogenes, novel and young genes in tumors may represent different stages of the origin of a new genes and novel organismal functions (which are not related to tumor progression) in multicellular organisms.
The analysis of published information about evolutionarily novel genes and/or sequences originated through different molecular mechanisms (by gene duplication, from endogenous viruses and retrotransposons, by exon shuffling or de novo) reveals that evolutionarily novel genes/sequences tend to be expressed predominantly in tumors, independent of the mechanism of origin. Sometimes the expression of evolutionarily novel genes in tumors is highly specific. Moreover, positive selection of many human tumor-related genes in primate lineage suggests their involvement in the origin of new functions beneficial to organisms. I suggested considering the expression of evolutionarily young or novel genes in tumors as a new biological phenomenon, a phenomenon of TSEEN (tumor specifically expressed, evolutionarily novel) genes [91]. This phenomenon is similar to phenomenon of carcinoembryonic antigens in that it represents a phenomenon of dual specificity, i.e. evolutionary and tumor specificities.
Some TSEEN genes are oncogenes, the others acquired functions beneficial to organism, but many TSEEN genes have no known functions. The lack of know functions is usually associated with the youngest TSEEN genes. We may infer that they are in the process of acquisition of function in the organism as suggested by positive selection of many of them in primate lineage.
TSEEN genes may thus represent a new interesting link between different but connected processes of gene origin, genome evolution, tumorigenesis and progressive evolution.

Funding
This study was supported by the grant №833 of Ministry of Education and Science of the Russian Federation, and grant of Russian State Project "5-100-2020" in Peter the Great SPb Polytechnic University to A.P. Kozlov.

Competing interests
The author declares that he has no competing interests.