Age‐dependent expression of cancer‐related genes in a long‐lived seabird

Abstract Studies of model animals like mice and rats have led to great advances in our understanding of the process of tumorigenesis, but this line of study has less to offer for understanding the mechanisms of cancer resistance. Increasing the diversity of nonmodel species from the perspective of molecular mechanisms of natural cancer resistance can lead to new insights into the evolution of protective mechanisms against neoplastic processes and to a wider understanding of natural cancer defense mechanisms. Such knowledge could then eventually be harnessed for the development of human cancer therapies. We suggest here that seabirds are promising, albeit currently completely ignored candidates for studying cancer defense mechanisms, as they have a longer maximum life span than expected from their body size and rates of energy metabolism and may have thus evolved mechanisms to limit neoplasia progression, especially at older ages. We here apply a novel, intraspecific approach of comparing old and young seabirds for improving our understanding of aging and neoplastic processes in natural settings. We used the long‐lived common gulls (Larus canus) for studying the age‐related pattern of expression of cancer‐related genes, based on transcriptome analysis and databases of orthologues of human cancer genes. The analysis of differently expressed cancer‐related genes between young and old gulls indicated that similarly to humans, age is potentially affecting cancer risk in this species. Out of eleven differentially expressed cancer‐related genes between the groups, three were likely artifactually linked to cancer. The remaining eight were downregulated in old gulls compared to young ones. The downregulation of five of them could be interpreted as a mechanism suppressing neoplasia risk and three as increasing the risk. Based on these results, we suggest that old gulls differ from young ones both from the aspect of cancer susceptibility and tumor suppression at the genetic level.

While studying model animals like mice and rats, which are highly susceptible to cancer, has led to great advances in our understanding of the process of tumorigenesis, this line of study has less to offer for understanding the mechanisms of cancer resistance (Seluanov et al., 2018). As a general rule, larger and longer-lived animal species have more effective cancer suppression mechanisms (Caulin & Maley, 2011). In recent years, several discoveries providing novel insights on the natural mechanisms of cancer resistance have been made in nonstandard mammalian species, including naked mole rats (Heterocephalus glaber, Liang, Mele, Wu, Buffenstein, & Hornsby, 2010), blind mole rats (Spalax ehrenbergi, Gorbunova et al., 2012), long-lived bats (Seim et al., 2013), whales (Keane et al., 2015), and elephants (Sulak et al., 2016). In each of these species, evolution has taken a different path, leading to novel mechanisms of cancer defense (Seluanov et al., 2018). Accordingly, studying a diversity of nonmodel species from the perspective of molecular mechanisms of natural cancer resistance can lead to new insights into the evolution of protective mechanisms against neoplastic cells and to a wider understanding of natural cancer defense mechanisms, which could eventually be harnessed for the development of human cancer therapies (Lemaître et al., 2019;Seluanov et al., 2018;Sepp, Ujvari, Ewald, Thomas, & Giraudeau, 2019).
So far, most of the studies of wild animal cancer defenses have focused on interspecific comparisons (but see Giraudeau et al., 2019), often looking at the duplication of specific known tumor suppressor genes. For example, it has been shown that elephants (Loxodonta africana) have more copies of tumor suppressor gene TP53 than other species (Sulak et al., 2016). Similarly, copy number gains involving genes associated with cancer have been shown in bowhead whales (Balaena mysticetus, Keane et al., 2015). Data for genomewide expression of aging-associated or cancer-related genes are also available for a few wild model species (i.e., wolves Canis lupus, Charruau et al., 2016, greater mouse-eared bats Myotis myotis, Huang, Jebb, & Teeling, 2016;Huang et al., 2019). A comparison of greater mouse-eared bats miRNA (microRNAs, regulators of gene expression) expression with that of humans, pigs, and cows revealed four upregulated cancer-related genes, three of which likely function as tumor suppressors and one as a tumorigenesis promoter (Huang et al., 2016). In addition, a study comparing short-lived and longer-lived strains of the fish Nothobranchius furzeri revealed differential expressions of miRNAs related to known tumor suppressor genes (Baumgart et al., 2012).
Here, we apply an intraspecific approach for studying tumor suppression mechanisms in wild animals. Since age has been shown to be one of the most important risk factors for cancer in humans and captive animals (Rozhok & DeGregori, 2016), comparing cancer suppression mechanisms between young and old individuals from a wild long-lived species could give a valuable insight into the mechanisms of age-related cancer suppression preferred by natural selection. For example, a recent longitudinal study in greater mouse-eared bats indicated that, several miRNAs acting as tumor suppressors were upregulated, while miRNAs promoting cell cycle or carcinogenesis were downregulated with age . From this perspective, seabirds are a promising, albeit currently completely ignored candidate for studying cancer defense mechanisms, as they have a longer maximum life span than expected from their body size and rates of energy metabolism (Holmes & Ottinger, 2003).
In this study, we used a known-age breeding colony of common gull (Larus canus, maximal recorded life span 34 years) to study the links between age and cancer defenses at the genetic level. To gain insight into the key molecular mechanisms underlying cancer defenses, we characterized the transcriptome of birds and compared the results with databases of known human cancer-related genes to assess whether cancer-related genes are differently expressed between old and young gulls, and between males and females.
Additionally, we analyzed whether the cancer-related genes differently expressed in our analyses overlap with aging-associated genes in other species, to understand whether the age-related patterns of cancer resistance in gulls are more likely phylogenetically conserved or unique to long-lived seabirds.

| Field methods
Samples were collected on the May 24, 2018, from a free-living, known-age breeding colony of common gulls located on Kakrarahu islet in Matsalu National Park on the west coast of Estonia (58°46' N, 23°26' E). This colony has been constantly monitored for over 40 years. All birds hatched on this islet are banded as chicks, so the exact age of the birds who return to the hatching colony to breed is known. In addition to a metal band, birds are also marked with a plastic band for ease of monitoring, and details of their breeding (i.e., partner choice, start of breeding, number of eggs) are recorded each year. Breeding birds are highly faithful to their colony, less than 3% of them change colonies between years, moving mostly to neighboring colonies (Rattiste, 2004). Male and female take turns in feeding and incubating throughout the day. For replacing damaged plastic bands, a subset of birds is caught every year from their nests using spring traps. To avoid nest abandoning, all birds are caught after the tenth day of incubation. From the birds needing to be caught for plastic band replacement, we chose 20 for collecting blood samples.
All of these birds were caught on the same day, between 9 and 12 a.m., and released immediately after band change and blood sampling. Up to 50 μl of blood was collected from the brachial vein using blood lancets. Blood was collected in 200-μl microvette tubes with EDTA as an anticoagulant, and about 10 μL of whole blood was then immediately transferred to RNAlater buffer. Samples were placed in a cooled and light-protected box and transported to storage at −80ºC until analyzed. Ten birds were selected for transcriptome analysis, based on the extracted RNA quality, distribution of birds between age classes, and their gender (Table 1). Maximum life span of common gulls is 34 years, and about half of the birds survive over the age of 8-9 years (Rattiste & Lilleleht, 1986). The experimen-

| Sequencing
Total RNA was extracted from RNALater preserved whole blood using RNeasy Mini Kit (Qiagen) according to manufacturer's instructions. The extraction also incorporated the optional DNAse digestion step. The initial quality and quantity of total RNA was determined using TapeStation (Agilent). On average 10 ± 3.4(SD) μg of total RNA was extracted with an average RIN 8.2 ± 2 (SD). The mRNA was extracted and cDNA generated using IlluminaTruSeq RNA
To identify possible genes that vary with age, we used the list of mammal age-related genes from the GenAge database (Magalhães & Toussaint, 2004). Custom database joins were performed using either SQLite version 3.24.0 (Hipp, Kennedy, & Mistachkin, 2018)

| RE SULTS
The Trinity de novo assembly resulted in 273,527 transcripts with a median length of 538 bp. The mean length of the assembled con-  Table S1) and 431 by age (220 matched genes, see Table S4). After removing the transcripts that had low abundance, no matches were found in the COSMIC database for sex-related differentially expressed genes. Two hundred of genes differentially expressed between age groups were downregulated and 20 upregulated in old birds compared to the young birds. Thirty of the transcripts that displayed differential expression by age were also present in the COSMIC cancer gene database (Table S2). After removing transcripts that had low abundance (logCPM < 1) and grouping similar transcripts, this list was reduced to eleven cancerrelated genes (Table S3). Out of the eleven differentially expressed cancer-related genes between young and old common gulls, three were likely artifactually linked to cancer after checking with latest OrthoDB version 10.1. The remaining eight were downregulated in old gulls compared to young ones. These eight transcripts were linked to the following cancer-related genes: TRIM33, USP6, PRDM16, SETD1B, MLLT3, KEAP1, CHD2, and DCAF12L2 ( Figure 1 and Table 2). The downregulation of the first five could be interpreted as a mechanism suppressing neoplasia risk and the downregulation of the last three as increasing the risk (see Table 2 for a description of the gene functions). We identified seven age-related genes from the GenAge list that could be matched to one or several differentially regulated gull transcripts (see Table S4). These were as follows: JAK2, A2M, TFDP1, EGF, RICTOR, SIRT7, VCP. We found no overlap between the list of cancer-associated and age-related transcripts. Accordingly, none of the cancer-related genes that were expressed differently between old and young gulls are known senescence genes in other studied species.
F I G U R E 1 Cancer-related differentially expressed genes between old and young common gulls. logCPM denotes the average log2-countsper-million transcripts (i.e., transcript abundance); logFC is log2 fold change between the groups. For example, value 4 means that the expression has increased 16-fold. Y-axis gene names are shortened names from the OrthoDB name; values in parentheses refer to the associated human gene symbols. See also Table S3 for detailed statistics

Ke
Tr TA B L E 2 The links of the transcripts that were differently expressed between young and old gulls with cancer-related genes, and the possible function of these genes

Hypothetical effect on neoplasia risk References
Serine/threonine protein kinase (STK) Tripartite motif-containing 33, also known as transcriptional intermediary factor 1 gamma (TRIM33, TIF1-γ) Prevents apoptosis, has a variety of cellular functions, including cell growth, differentiation, immune response, and carcinogenesis.
Lower expression in old birds, which possibly lowers cancer risk Wang et al. (2015), Lee (2018) Ubiquitin-specific proteases (USP) USP6, also known as TRE17 Upregulation of USP6 leads to bone neoplasms, and this gene is therefore considered as an oncogene. While USP6 is a hominoid-specific gene that was formed as result of a recent evolutionary fusion of the ancestor genes TBC1D3 to USP32, both of the ancestor genes have also been linked with oncogenic processes.

| D ISCUSS I ON
The well-known Peto's paradox highlights the gaps in our current knowledge explaining why animals with larger bodies and longer life spans do not have higher incidence of cancers (Rozhok & DeGregori, 2016). While the paradox remains unresolved, most investigations focused on the evolution of intracellular mechanisms that reduce the risk of cell transformation (Rozhok & DeGregori, 2016). The current study applies a novel, intraspecific approach in a nonmodel wild animal and improves our understanding of the mechanisms of agerelated cancer risk and cancer suppression.
While there has been a push toward applying transcriptome methods in ornithological studies over the recent years (Jax, Wink, & Kraus, 2018), little is known about the age-specific gene expression in birds. Age-specific gene expression in follicles of Peking ducks (Anas platyrhynchos) indicated no clear pattern regarding expression levels in relation to age (Ren et al., 2019), while in great tits (Parus major), age classes were not associated with differential gene expression levels in blood and liver (Watson, Videvall, Andersson, & Isaksson, 2017).
Conversely, in the present study, we found that more than 90% of the differentially expressed genes (including also transcripts that were not related to cancer) were downregulated in older birds. The average life expectancy for ducks is three years and for great tits 2-3 years, with maximal life span of twenty years for ducks and 15 years for great tits, accordingly, both of these species have much shorter life expectancies than common gulls. We can therefore hypothesize that the lower level on gene expression in the blood of old common gulls is related to a longer life span, slower pace of life and metabolism level necessary for reaching older age (Auer, Dick, Metcalfe, & Reznick, 2018). It should nonetheless be noted that the 220 genes differentially expressed between young and old common gulls comprise only about 0.57% of all the transcripts compared in our analysis.
Our analysis allowed the identification of 11 genes that could be linked to cancer and which were differently expressed between young and old common gulls. Our comparison of age groups was rather conservative, since the applied false discovery filtered out only the most clearly differentially expressed transcripts.
Accordingly, we are confident in the biological relevance of the found differences. Out of these eleven genes, one was expressed in higher levels in older gulls, while for ten others, the expression was downregulated in the "old" group. This could indicate a reduced expression of these genes with increasing age, and/or that this lower expression is a prerequisite condition for reaching old age in common gulls. Given the cross-sectional nature of our study, we cannot outrule the possibility that some birds in the "young" group will also reach old age (16+ years in our study). However, as only 17% of common gulls who start breeding at the study site reach the age of 16 (Rattiste & Lilleleht, 1986), we can suggest that the "old" group consists of birds that are physiologically or genetically better adapted to reach old age. These genes play critical roles in many cellular processes, including cell proliferation, survival, DNA repair, and genomic integrity. DCAF12L2 has been shown to be mutated in several human cancers.
Lower expression in old birds, which possibly increases cancer risk Lee & Zhou, 2007, Liu et al. (2012, Gylfe et al. (2013) Note: For their role in cancer, please see Appendix S1.

TA B L E 2 (Continued)
Based on the known functions of these genes in humans, we divided the differently expressed genes into two groups: genes that decrease cancer risk, and genes that increase cancer risk (see the Discussion below). Additionally, three of the transcripts that were identified in our analyses as being similar to cancer-related genes were more convincingly linked to other aging-associated genes and the links with cancer genes in these cases are likely an artifact of our analysis. This is confirmed by the fact that the analysis where these links appeared was performed on OrthoDB version 10.0 while using the most recent OrthoDB version 10.1 did not exhibit these links. Therefore, instead of being linked to cancer-related human orthologues suggested by the analysis, these three transcripts are more strongly associated with the following age-related human orthologues. (a) ferritin gene (linked to CDK6), which could be linked back to age-associated pathologies in humans (Touitou et al., 1985); (b) amyloid beta precursor A4 gene (linked to EXT1), which is widespread in the majority of vertebrate species that do not cease reproduction in senescence, and where selection pressure is maintained into old age (Moir & Tanzi, 2019), and (c) adenylate cyclase 3 (linked to PTEN), which is regulating fat accumulation and insulin levels in mammals (Liang et al., 2016). More details about the functions of these and also all the other transcripts are included in the Appendix S1.
The description of the functions of the eight remaining genes is presented in Table 2. The downregulation of 5 of these genes in older gulls could be a mechanism of cancer suppression. The first transcript could represent TRIM33 gene (tripartite motif-containing 33, also TIF1-γ), a transcriptional cofactor that prevents apoptosis (Wang et al., 2015), and similar role could be ascribed to the second transcript linked to PRDM16 gene (Zhu et al., 2016). Downregulation of these genes in older gulls could stimulate apoptosis and thus limit cancer progression. The third transcript was linked to USP6, which, when upregulated, is considered an oncogene causing bone neoplasms (Oliveira et al., 2004), and downregulation of this gene in older gulls suggests an anticancer mechanism. The fourth transcript was linked to SETD1B, which supports mitotic processes (Tajima et al., 2019) and is overexpressed in several cancer types (e.g., Yang & Ernst, 2017;Chen et al., 2019). A lower expression of SETD1B in older gulls thus suggests a method for suppressing uncontrolled cell growth and thus neoplasia. The fifth transcript was linked to MLLT3 (protein AF-9), which has been associated with leukemia in several vertebrate species (Ney Garcia et al., 2015). While its upregulation enables cell proliferation (Calvanese et al., 2019), downregulation reduces it (Zhang, Luo, Wang, & Yang, 2012). Hence, the downregulation of MMLT3 in common gull blood cells with age might reflect increased cancer resistance. On the other hand, the 3 other genes that were downregulated in older birds could potentially lead to increased cancer risk. The first transcript was linked to KEAP1 (Kelch-like ECH-associated protein 1), which, through its link to oxidative stress regulating genes, such as glutamate-cysteine ligase and glutathione reductase (Yang et al., 2013), is considered as a tumor suppressor gene. Interestingly, our previous studies in common gulls have indicated that the glutathione system is indeed linked to the longevity of these birds (Urvik et al., 2016). Lower expression of a tumor suppressor in older gulls could increase cancer risk. The second transcript was linked to the gene DCAF12L2, which was shown to be mutated in several human cancers (Gylfe et al., 2013;Liu et al., 2012). However, the role of this gene in human cancers is still under investigation. The third transcript was associated with CHD2 and CHD4 genes, which play a critical role in tumor suppression (Nagarajan et al., 2009), but also in cellular proliferation, senescence, and apoptosis (Mills, 2017). Again, a lower expression of tumor suppressor genes in older gulls could suggest higher cancer risk.
To summarize these results, out of eleven differentially expressed cancer-related genes between young and old common gulls, three were most likely artifactually linked to cancer, since these transcripts are associated with aging processes through more straightforward physiological mechanisms. Out of the eight remaining genes, the downregulation of five in old birds could be interpreted as a tumor suppressor mechanism, and three as potentially tumor promoting (Figure 2). Since cancer (and cancer suppression mechanisms) is an evolutionarily very old issue (Nesse, 2017) (Ladds, 2009;Reece, 1992). It is still not known how strong selection pressure cancer could be for gulls.
What we do know, based on studies in common gulls, is that on the last year of life, their reproductive success will drop (Rattiste, 2004), suggesting that the death of gulls is not only random predation, but that physiological senescence plays a role. Future effort should be targeted at trying to determine the cause of mortality in common gulls, for example, by collecting carcasses and performing necropsy.
Even species with very low cancer prevalence have to invest in cancer prevention, possibly even more than species with high cancer prevalence, making them intriguing models for understanding the evolutionary mechanisms of cancer defense .
Common gulls might have genetic mechanisms for suppressing cancer that allow these birds to bypass some of the processes linked to senescence, allowing them to reach old age without increased risk of cancer. This is further supported by the finding that none of the identified cancer-related genes expressed differently between age groups has been associated directly with aging in humans or model species. We did identify seven differently expressed transcripts that could be linked with known human aging-associated genes as curated in the GenAge database, but there was no overlap with cancer-related genes revealed by our analysis. These results suggest that cancer suppression might represent an evolutionary adaptation that enables these seabirds to achieve longevity, despite their high metabolic rate and small body size. For example, when looking at telomere shortening, which is considered a mechanisms mediating trade-offs between senescence and cancer susceptibility (Nesse, 2017), previous studies indicated that unlike many other species, older gulls do not have shorter telomeres when compared to younger gulls (Rattiste et al., 2015). Early in life, in the fastest growth phase, their telomeres might actually elongate instead of shortening (unpublished data). Our results therefore call for further investigations focused on (a) the comparison of gene expression between common gulls that exhibit signs of cancer or not (although cancer diagnostics in wild animals need to take a step forward), and (b) the comparison of our results with data from other long-lived seabirds, in order to evaluate if the found patterns in cancer-related genes are universal among bird species with similar life histories.
In conclusion, we have shown that old gulls differ from young ones both from the aspect of cancer susceptibility and tumor suppression at the genetic level. This is the first study to look at intraspecific variations in cancer defenses in relation to aging at the genetic level in a wild bird species. It is intriguing to speculate that these seabirds have, through evolutionary pressure for a longer life span, found physiological or genetic pathways to bypass the inevitable process of senescence (see also Rattiste et al., 2015;Urvik et al., 2016). Hopefully, future years will bring a fast accumulation of data on the genetic mechanisms of cancer defenses in nonmodel species, thereby improving our understanding of this phylogenetically very old, but at the same time very contemporary issue.

ACK N OWLED G EM ENTS
We thank Janek Urvik for help with field work. This study was supported by the Estonian Research Council (IUT34-8, PUT653, PSG458), French-Estonian cooperation program PARROT, and Marie Sklodowska-Curie grant agreement no. 701747 to TS.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
The raw sequence reads and the transcriptome assembly are made available in the European Nucleotide Archive under the study accession no. PRJEB35479. Rest of the additional data can be found from the Appendix S1.