Discussion

Gene losses in inflammation–related gene families and positive selection in single–copy genes associated with immune and cell repair functions in mammalian models have been evident since the very first bat genome assemblies were published (G. Zhang et al., 2013). Although subsequent studies have confirmed those initial results (Ahn et al., 2016; Seim et al., 2013), confidence in assessing both gene losses and gene family expansions has strengthened only recently, with the publication of highly contiguous assemblies for a few bat species (Jebb et al., 2020; Scheben et al., 2020). Examining a comprehensive sample of bat lineages while checking against high quality genome assemblies and multi organ RNA Seq, our analyses reveal system wide gene losses with the potential to modify the sensitivity, targets, and magnitude of immune responses across all bats. These inferred losses are particularly concentrated along inflammasome activation pathways, which are triggered by the innate immune recognition of pathogenic signals through both pathogen-associated molecular patterns (PAMPs) and damage associated molecular patterns (DAMPs). In contrast with more pathogen–driven PAMPs, DAMPs result from host cellular distress signals such as mitochondrial stress and reactive oxygen species (ROS) (Zheng, Liwinski, & Elinav, 2020), which bats produce during active flight (Costantini, Lindecke, Petersons, & Voigt, 2019). Bat cells, in turn, display exceptional mechanisms of repair (Pickering, Lehr, Kohler, Han, & Miller, 2014) and resist damage (Harper, Salmon, Leiser, Galecki, & Miller, 2007), connecting molecular signaling and cell processes to extreme longevity (Salmon et al., 2009; Wilkinson & Adams, 2019).
Based on our genomic surveys, immune-related losses can be divided into three categories: the epithelial defense receptors (defensins), the Natural Killer gene complex (NKC) and the interferon-induced pathway (IFI; HIN; PYHIN) (Figure 2). This particular combination of losses in crucial components of immune activation seems surprising, as it would imply that bats mount a low intensity immune response to primarily intracellular pathogens: viruses. But integrating these genomic findings with published functional data suggests complex, systemic adaptation, in line with both previous analyses of bat immune system responses (A. Banerjee et al., 2020; Basler, 2020; P. Zhou, 2020) and the growing body of evidence for cellular mechanisms underlying longevity (Z. Huang, Whelan, Dechmann, & Teeling, 2020; Z. Huang et al., 2019, Kacpryzk et al., 2017). We review these losses in a stratigraphic order, from the outer cellular matrix to the inner cellular pathways, starting with the defensins.
While defensins are the primary barrier of the immune system, with broad antimicrobial activity that covers bacteria, fungi and viruses (Semple & Dorin, 2012; Xu & Lu, 2020), bat defensin losses consist mainly of orthologs of genes localized to epithelial cells. Defensins can function as modulators of the host’s cell surface receptors, and α and β defensins genes have pleiotropic effects on the regulation of carcinogenesis and inflammation (Xu & Lu, 2020). By acting as chemokines to alter the adaptive immune response, defensins also serve as a bridge between innate and adaptive immunity (Grigat, Soruri, Forssmann, Riggert, & Zwirner, 2007). In bats, both α and β defensin genes have undergone a rapid evolutionary change through either loss or positive selection (Table 1, Figure 2a, Supplementary Table 4). In humans, defensins can elicit proinflammatory cytokine production (Niyonsaba et al., 2010; Wiens, Wilson, Lucero, & Smith, 2014), but overexpression of certain defensins can actually enhance viral infection (Rapista et al., 2011). We hypothesize that specific defensin losses in bats (Figure 2a) complement several other mechanisms (Ahn et al., 2019; A. Banerjee et al., 2017; Xie et al., 2018) contributing to a dampened inflammatory response, reduced host–driven damage from viral infections, and enhanced longevity (Baker & Schountz, 2018; Brook & Dobson, 2015; Gorbunova et al., 2020). For example, modifying defensin repertoires on epithelial cells would result in fewer instances of both immune cell recruitment and initiation of inflammatory pathways known to damage healthy tissue (e.g., focal necrosis in lungs, spleen and lymph nodes during the inflammatory response during SARS-Cov2 infection (Merad & Martin, 2020)). In humans, loss of β-defensins prevents the inhibition of neutrophil apoptosis and thus averts the production of proinflammatory cytokines and chemokines (Nagaoka, Niyonsaba, Tsutsumi-Ishii, Tamura, & Hirata, 2008), avoiding the amplification of the immune response, and may have a similar effect in bats. Losses of some epithelial surface defensins would thus reduce inflammation without compromising responses to intracellular pathogens.
Another result with inferred implications for reducing proinflammatory reactions involves losses of Natural Killer (NK) receptors that play an important role in the recognition of MHC-I molecules and regulation of cytotoxic activity against virus–infected cells. While killer-cell immunoglobulin like receptors (KIR) and killer cell lectin-like receptors (KLR) receptor losses has been previously reported forPteropus alecto and Myotis davidii (Papenfuss et al., 2012; G. Zhang et al., 2013), our analyses confirm these losses across Chiroptera (Supplementary Table 4). Although the Killer Cell Lectin Like Receptor K1 (KLRK1 or NKG2D) gene is present in bats, its ligands, gene subfamilies RAET1 and H60 responsible for binding and activating NKG2D receptors, recruiting natural killer cells, and stimulating them to secrete Interferon gamma (IFN-γ) (Zhi et al., 2010), were absent in all bat species (Figure 2b).
We hypothesize that these losses lead to low recruitment of proinflammatory NK cells and reduce B-cell signaling (Arapović et al., 2009; Stolberg et al., 2014; Takada et al., 2008; Wortham et al., 2012), as they do in mice and humans. Loss of this particular mechanism of activation of the MHC-I pathway prevents proliferation of immune cells, which can be cytotoxic, proinflammatory, and targets of viral infections (Djelloul, Popa, Pelletier, Raguénez, & Boucraut, 2016; Wortham et al., 2012). For example, NKG2D–deficient mice infected with influenza viruses exhibit less airway damage and reduced inflammation without compromising viral clearance; NKG2D stimulation is a central pathway to viral-mediated NK cell hyperresponsiveness (Wortham et al. 2012), and has been shown to be involved in age-dependent COVID–19 severity (Akbar & Gilroy, 2020). During viral exposure, rarer activation of NKG2D function would therefore lead to less inflammatory exacerbation. Reducing instances of NKG2D activation might also reduce B cell signaling, as it occurs in NKG2D–deficient mice (Lenartić et al., 2017; Zafirova et al., 2009), and complements losses of immunoglobulin heavy chain variable regions IGHV1, IGVH3 and IGHV14 genes that modify the B cell receptor signaling pathway, and thus B lymphocyte differentiation (M. Banerjee, Mehr, Belelovsky, Spencer, & Dunn-Walters, 2002; McHeyzer-Williams, Okitsu, Wang, & McHeyzer-Williams, 2012; Reddy et al., 2010). Based on the roles of both NKG2D and B cell activation in promoting inflammation in viral infection, and since some viral proteins have been shown to specifically target the NKG2D receptor via the RAET1 and H60 loci (Arapović et al., 2009), we propose that these losses resulted from selection during viral infections early in the evolutionary history of bats. While the functional implications for bats need to be tested, in humans, lack of specificity of the T and B cells in children results in a broader immune response to novel viruses (Pierce et al., 2020), and it may confer analogous advantages in bats.
Complementing losses in defensins and NK signaling, the third large group of gene losses involves the IFN-γ pathway (Figure 2c). While representatives of the PYRIN and HIN domain (PYHIN) gene family, immune sensors of cytosolic DNA activating the inflammasome and IFN-γ, are present in all mammals, they have not been found in any of the bat genomes analyzed thus far examined (Ahn et al., 2016; G. Zhang et al., 2013; Jebb et al., 2020). Previous genomic analyses linked losses in this inflammasome pathway to the unique demands of bat flight and in response to increased ROS production (G. Zhang et al., 2013), but there are immune implications as well. In other mammals, the presence of dsDNA, DAMPs and PAMPs, or, especially, bacteria and DNA viruses, induces the (PYHIN) AIM2 inflammasome, while the IFI16 inflammasome (Interferon-inducible protein 16, also missing in bats) recognizes viruses replicating in the nucleus (Zheng et al., 2020). Hence, these bat gene losses could undermine innate defense against viruses. We hypothesize that bats have evolved mechanisms to overcome this potential disadvantage in rapid recognition and response against viruses through expansion of MHC-I class genes (Supplementary Table 7). These genes are involved in the recognition and binding of intra cellular peptides, and previous studies have described a unique 5–amino acid insertion at the exon 2 peptide binding region (PBR) on bats which may allow the host to recognize longer peptides (Ng et al., 2016; Papenfuss et al., 2012). Besides implications for immunity, IFN-γ pathway gene losses also point to changes in autophagy. In mice, loss of the IFN-γ inducible immunity related GTPase gene (IRGM1 and IRGM2) results in an IFN-γ induced autophagic death program in lymphocytes (Feng et al., 2008). Along with the loss of other IFN-γ related genes (IGTO, IIGP, TGTP2), these losses may help achieve apoptosis of infected cells without runaway inflammation.
While some mechanisms of activation of IFN-λ are lost in bats, IFN-γ itself is under positive selection within branches (Table 1 Supplementary Table 7). IFN-γ is a crucial part for the first line of defense against viruses, helps shape adaptive immune memory (Schroder, Hertzog, Ravasi, & Hume, 2004), and its deficiency increases inflammation (Loo et al., 2017). Thus, evolutionary adaptation may have shaped bats’ unique ability to induce a rapid antiviral response without triggering runaway inflammation. This fine-tuned response may be achieved by expressing high levels of IFN-γ early on, which recruits broad-spectrum immune cells to the site of injury, while negatively regulating the IFN-γ pathway receptors that trigger inflammation (Ahn et al., 2019; Ferber et al., 1996).
In contrast to a pattern of proinflammatory signal losses, most other variation in gene families in Chiroptera corresponded to cell processes and metabolic functions. Shifts from the ancestral bat insectivorous diet to including nectar and fruit and the resulting mutualistic relationships between bats and plants appear to have led to elevated rates of diversification and the evolution of new morphological traits (Dumont et al., 2012; Jones, Bininda-Emonds, & Gittleman, 2005), but gene family evolution has remained underexplored. Regarding significant expansions (Supplementary Table 7), we identified few —only nine— sets of duplications independently replicated across all pteropodids and phyllostomids with convergent, plant–based diets (Figure 1). In addition to a trace amine associated receptor (TAAR) of unknown chemosensory function (Liberles & Buck, 2006) and a putative homolog of the yeast protein transport protein YIP1, two genes stand out as candidates for diet–linked adaptive gene family evolution: those encoding homologs of inositol monophosphatase 1 (IMPA1) andintegrin alpha-D/beta-2 (ITAD). Glycolysis, the metabolic pathway that breaks down glucose to ultimately phosphorylate more ADP into ATP than the reverse, begins with the phosphorylation of glucose into D-glucose 6-phosphate (Berg, Tymoczko, & Stryer, 2002). This metabolite, however, cannot diffuse through the membrane and is thus highly osmotic; its accumulation would cause cells to swell. Through the synthesis of myo -inositol from D-glucose 6-phosphate, IMPA1 provides one avenue to protect cells, particularly in the brain (Parthasarathy, Parthasarathy, & Vadnal, 1997), from the osmotic stress of this glucose metabolite (Rafikov et al., 2019). We found independent IMPA1 duplications in the pteropodid ancestor, A. jamaicensis ,A. caudifer , P. discolor , and the common ancestor of phyllostomids and Mormoops . Except for the aerial insectivoreMormoops , all the lineages with IMPA1 duplications include nectar and fruit in their diet (Figure 1), are expected to at least occasionally experience high blood glucose levels (Amitai et al., 2010; Ayala & Schondube, 2011; Kelm, Simon, Kuhlow, Voigh & Ristow, 2011; Welch, Herrera & Suarez, 2008; Meng, Zhu, Huang, Irwin, & Zhang, 2016), and therefore require options for processing metabolites from glycolysis. Although beta integrins, including ITAD, are regulators of leukocyte function and therefore not annotated as directly involved in metabolism, leukocyte adhesion has been found to modulate glucose homeostasis via lipid metabolism (Meakin et al., 2015). Specifically, mice deficient in a paralogous beta-2 integrin become spontaneously obese in old age despite a normal diet (Z. Dong, Gutierrez-Ramos, Coxon, Mayadas, & Wagner, 1997), and when fed a fat rich diet show obesity, inflammation, high neutrophil activity and insulin resistance in skeletal muscle (Meakin et al., 2015). Likewise, mice deficient in this same integrin are unable to respond to fasting by increasing fat uptake and reduce insulin levels slowly compared to normal mice (Babic et al., 2004). We found single ITAD duplications in lineages that include sugar rich foods in their diet: ancestral pteropodids and phyllostomids, as well as Leptonycteris yerbabuenae , two each inMacroglossus , Anoura , and Tonatia , and three inArtibeus jamaicensis . While the function of these lineage–specific bat paralogs remain unknown, their phylogenetic distribution warrants future exploration and functional analysis.
In summary, our results, grounded on the most comprehensive survey of bat genomes to date, suggest bats have evolved complex mechanisms of inflammasome regulation. These may have evolved to prevent uncontrolled inflammatory response against DAMPs byproducts of the high metabolic rate required for powered flight (Banerjee et al., 2017; Banerjee et al., 2020; Subudhi, Rapin & Misra, 2019; Xie et al., 2018), to better respond against intra-cellular pathogens such as viruses, or some combination of both. Regardless of the ecological origin of selection, compared to mammals such as humans or mice, bat genomes reveal systemwide immune evolution that prevents or dampens aggressive inflammatory responses. In contrast with these gene losses, we found significant expansions in gene families involved with glucose degradation, coinciding with the transition from a diet based mainly on insects to a high-glucose content diet that includes fruit and nectar.
By undertaking large-scale comparative genomic analyses encompassing many ecologically divergent lineages, the present study demonstrates the impact of genomics in non-model organisms. Such analyses allow elucidating the broad evolutionary mechanisms in a given clade, with potential for functional implications. Yet, heterogeneity in assembly quality continues to limit the scope of inference. Hence, the need to generate high quality genomes for future studies endures.