key: cord-1007997-cqmr71au authors: Hecht, Luke B.B.; Thompson, Peter C.; Rosenthal, Benjamin M. title: Assessing the evolutionary persistence of ecological relationships: a review and preview date: 2020-07-01 journal: Infect Genet Evol DOI: 10.1016/j.meegid.2020.104441 sha: 063ed2746184093dd145cff4571d47dfaf457345 doc_id: 1007997 cord_uid: cqmr71au Species interactions, such as pollination, parasitism and predation, form the basis of functioning ecosystems. The origins and resilience of such interactions therefore merit attention. However, fossils only occasionally document ancient interactions, and phylogenetic methods are blind to recent interactions. Is there some other way to track shared species experiences? “Comparative demography” examines when pairs of species jointly thrived or declined. By forging links between ecology, epidemiology, and evolutionary biology, this method sheds light on biological adaptation, species resilience, and ecosystem health. Here, we describe how this method works, discuss examples, and suggest future directions in hopes of inspiring interest, imitators, and critics. The stability of specific relationships may be of practical interest. Non-human primates have been frequent sources of zoonotic infection, presumably owing to the "low barrier to entry" for microbes adapted to physiologically similar environs (Wolfe et al., 2007) . Livestock have been another important source, presumably owing to their population densities and physical proximity to people. Why might bats serve as frequent sources of zoonotic infections, such as that of SARS and the COVID-19 coronavirus? Has this long been the case, or do such instances ensue from only recent patterns in habitat incursion? Farming, especially of crop monocultures, invites specialist crop pests, weeds, and pathogens; can we learn something about the typical sources of such uninvited guests? Can we better harness nature's bounty using approaches modeled on more resilient natural systems? The "parasite release" hypothesis holds that invasive species thrive when rid of the infections that held them in check in their places of origin (Keane and Crawley, 2002; Shea and Chesson, 2002; Torchin et al., 2003 Torchin et al., , 2001 Torchin and Lafferty, 2009) . Might this mean that long-term ecosystem dynamics are bound by longstanding antagonisms from competitors and pathogens? Or do such outcomes instead ensue from a revolving cast of pathogens? When we introduce natural enemies in hopes of reducing the harms wrought by invasive species, for how long can we hope such biocontrol to endure? Assessing the longevity of biological associations has traditionally required direct physical evidence (i.e. fossilized parasites (Morris, 1981; Peñalver et al., 2017; Poinar, 2003; Poinar and Boucot, 2006) ) or indirect evidence derived from comparative phylogeny (Galbreath et al., 2009; Hafner et al., 1995 Hafner et al., , 1994 Hafner and Nadler, 1988; Hoberg and Brooks, 2015) . Before reviewing our new method, we will briefly describe the promise and limitations of the preceding two approaches. Direct physical evidence of a longstanding association has come in the form of fossils that attest to ancient pairings. Such evidence has illuminated, for example, for how long particular insect groups have served as pollinators for particular plant groups (Poinar, 2003) . Parasitic mites, preserved in amber, have been documented on feathered dinosaurs (Peñalver et al., 2017) . DNA originating from Borrelia spirochetes was isolated from ticks in museum collections, preserved a century before the recognition of Lyme Disease in North America (Hubbard et al., 1998; Marshall et al., 1994; Persing et al., 1990) . Such findings excite the imagination by showing us how the ancient world resembled the present. Moreover, subsequent evolutionary radiations affirm that, once established, new biological relationships (i.e. insect pollination of land plants; the establishment of photosynthetic symbioses in coral and in lichens) changed the subsequent course of life on the planet (Pellmyr, 1992) . Grazers require grasslands; limpets require seas. It is hard to imagine anything more persuasive than direct physical evidence to establish the minimum age of any form of biological interaction. However, most of life's partnerships and pathogens have not been fossilized. What can be said about their origins and persistence? chewing lice of pocket gophers (Hafner et al., 1995; Hafner and Nadler, 1988 ) requires explanation; the simplest explanation would seem to be a causal one (the lice became subdivided owing to allopatric speciation of their solitary hosts). Stated more generally, the stability or transience of biological interactions has been profitably explored through the joint consideration of phylogenies, despite warranted debate as to the fairest interpretation of concordant (or discordant) patterns of evolutionary diversification. Strict co-speciation may have been too-often accepted when a looser process of "ecological fitting" (accommodating frequent switching among often related host groups) would suffice (Brooks and Hoberg, 2007; Brooks, 2015, 2008) . Recently, we introduced a third approach to examining evidence for shared histories, premised on the simple idea that given pairs of species appear to have undergone parallel episodes of population growth and decline (Hecht et al., 2018) (Figure 1 ). On a short timeframe, antagonisms between host and parasite or predator and prey might be expected to engender inversely correlated population sizes; indeed, the classical Lotka-Voltera equations modeling predator/prey dynamics assume that growth in a predator population thereby suppress prey populations, cyclically limiting the growth of each (Wilson and Bossert, 1971) . Over longer intervals, we suggest that biological dependencies would engender concerted population growth or contraction, whether driven by climate, resource availability, or other factors. That is, climates favorable to expanding the savannah would favor growth of its various inhabitants and their pathogens, commensals, and parasites. Such intuition was invoked to explain evidently concordant expansions and contractions in the populations of leaf-eating monkeys and pandas, two unrelated consumers of similar plant resources (Zhou et al., 2014) . The spread and retreat of glaciers, the genesis of a floodplain, or the desertification of a grassland would similarly impose new opportunities or limitations to a broad suite of organisms, irrespective of the nature or strength of their biological interdependencies. Bearing in mind such caveats, and cognizant that no correlation suffices to establish causation, comparing demographic histories constitutes a valuable and, until recently, entirely unexplored avenue to consider the temporal durability of biological dependencies. Concerted growth and contraction in species pairs (or groups) should be viewed within a broader consideration of their historical and/or extant interconnections, and can serve to generate hypotheses and better delimit the plausible historical depth of their interaction. By contrast, extant dependent species pairs (or groups) lacking a parallel pattern of population growth and decline through time might be suspected as only recently establishing ecological or epidemiological dependency. Our prior work suggests exactly that for potatoes and the oomycete agent of potato blight (Phytophthera infestans): it appears they did not commence growing in tandem until potatoes were domesticated. Larger and older populations should generally harbor more neutral genetic diversity than smaller and younger populations. This idea serves as the foundation for modern population genetics writ large, and the neutral theory of molecular evolution, in particular (Kimura, 1983) . The study of island biogeography well illustrates the idea, establishing empirically that small founding populations (at the extreme, a single clonal individual or a single breeding pair) bequeath to descendent populations a small fraction of variation then circulating in source populations (i.e. those on the mainland (Johnson et al., 2000; MacArthur and Wilson, 2001) . Diversity among the descendants of such populations will remain low for extended intervals (because new mutations are slow to accrue) unless infused by new variants imported by additional immigrants. Building on this intuition, a system of inferential tools has been constructed to express the expected rate at which new neutral alleles are birthed, their probabilities of survival, and their probabilities of ultimate fixation. Viewed in reverse, these models describe the coalescence of extant alleles into their shared common ancestors. These probabilities are well-characterized for theoretical populations of constant size, and data may be evaluated against expectations from other demographic scenarios (Rosenberg and Nordborg, 2002; Erikkson et al 2010) (Figure 2 ). As in the case of island populations founded by small numbers of individuals, any population experiencing a rapid population decline will limit ensuing variability for estimable intervals (Galtier et al., 2000) . Long-lived population bottlenecks most profoundly reduce diversity long after a population's census size has rebounded. Each allelic lineage surviving such a bottleneck may foster its own clusters of descendant lineages. By contrast, rapid population expansion produces "star phylogenies" wherein many extant variants differ only minimally from one another, and without deeper subdivisions among them (Kuhner et al., 1998; Slatkin and Hudson, 1991) . Such patterns are most readily apparent in non-recombinant loci (such as most mitochondrial haplotypes); recombination ultimately obscures the history of any single descent history (by generating myriad, ephemeral genotypes each incorporating innovations arising from all reproductively successful progenitors). Coherent blocks of linked variants of common descent cannot long persist in sexual populations (excepting where recombination is locally or globally suppressed), meaning that their occurrence and distribution provides additional means to draw inferences about a population's past size. Models of recombination provide well-founded expectations for how genetic variation should be distributed throughout the genome of individuals sampled from sexual populations of constant size (Hudson and Kaplan, 1988) . Departures from these expectations provide a basis to recognizing population inconstancy (Donnelly and Tavare, 1995; Schraiber and Akey, 2015) . Several tools have been developed to mine extant genetic and genomic information to reconstruct a population's demographic past (Heled and Drummond, 2008; Li and Durbin, 2011; Pybus et al., 2000; Schiffels and Durbin, 2014; Terhorst et al., 2017) . Each infers a history of demographic stability, growth, or contraction from information regarding the frequency distribution and/or coalescence history of extant alleles. By incorporating estimates of pergeneration mutation rate, generation time, and recombination rates, these tools enable estimates of approximately when ancestral populations experienced stasis or change. These methods for Since the development of theory for how varying population size would be reflected in a coalescent tree (Donnelly and Tavare 1995) , numerous programs have been written to assess the likelihood of complex demographic histories based on the inferred genealogy of a population sample. Comparing two or more demographies focusses on the "shape" of those histories -the order and relative timing of population growth and decline. The absolute size of either population, at any given time, seems less important to know. Several families of method satisfy this requirement (Table 1) . Each method performs at its own optimal rang e of timescales, primarily depending on its investment in sample size or genome coverage. Fundamental progress on model-agnostic demographic reconstruction was made by Pybus, Rambaut and Harvey (2000) with the introduction of the "skyline plot." The skyline plot has seen several iterations, leading to the "extended Bayesian skyline plot" (EBSP; Heled and Drummond 2008) , widely used for inferring relatively recent demographic histories from multilocus population samples. The most recent past is illuminated b y sampling a few loci from numerous individuals. The development of the pairwise sequentially Markovian coalescent (PSMC) model by Li and Durbin (2011) marked the beginning of a new approach to demographic history reconstruction. It exploits the fact that the genome of a single diploid individual is a mosaic of genes from numerous ancestors, assembled through recombination ( Figure 2 ). The approach infers past population size by evaluating the extent and genomic distribution of heterozygous positions, comparing whole genome data to a coalescent model incorporating recombination (Figure 3 ). Whole-genome methods are most effective in probing a population's demographic history between about 500 and 500,000 generations in the past. As the cost of whole genome resequencing has fallen, more recent methods have sought to combine the advantages of PSMC and the EBSP. For example, MSMC and SMC++ utilize whole genomes of multiple individuals to resolve population sizes over a wider timescale than either PSMC or the EBSP. For example, SMC++ utilizes the allele frequency spectrum (AFS) of a population sample, while MSMC uses haplotype phase data. Such additional information reduces uncertainty about the recent past, which can be difficult to discern based on heterozygosity alone because substantial differentiation among alleles requires time. One limitation of most existing wholegenome methods (the SMC family) is that confidence in their demographic reconstructions is only expressed in terms of bootstrap variability. The EBSP more formally allows estimation of posterior probabilities. Terhorst et al. 2017 In principle, any method producing a credible estimate of demographic history could be used as a basis to compare past population dynamics in two or more species. Box 1 summarizes some of those now available, and describes their individual strengths, requirements, and precision. Among these is a method that mines information from the range of coalescent histories represented by maternal and paternal alleles at the many loci in the genome of a single diploid individual (Figures 3-4) . The Pairwise Sequentially Markovian Coalescent (PSMC) model (Li and Durbin, 2011) , has been shown to accurately reconstruct demographic histories from simulated data. When applied to human genomes, the method demonstrated an impressive ability to discern "out of Africa" and other salient features of human demography supported by a strong foundation in physical anthropology, linguistics, and population genetics. Consequently, this method has found broad application to gather insight in a range of biological taxa (Liu et al., 2016; Miller et al., 2012; Nadachowska-Brzyska et al., 2016; Zhou et al., 2018) . The sufficiency of but a single diploid genome makes this method attractive for studying non-model organisms, providing a convenient starting point for seeking to understand demographic history for any diploid taxon of interest. The methods we developed to compare demographic histories inferred from the PSMC (Hecht et al 2018 and summarized below) could in principle be applied, with little modification, to demographic histories reconstructed using any of several approaches (summarized in Box 1). The elaborations and improvements we envision ought likewise to be applied to other members of this family of methods. To the extent that the survival and reproduction of individuals of one species depend on the bodies or ecosystem services of another species, their numbers should rise and fall in tandem. A classic example of this is found in models of predator-prey dynamics, where a predator's reproductive rate depends on the abundance of its prey (Kendall et al., 1998) . Shared success should occasion species that thrive under shared conditions; conversely, adverse events (such as glaciation) threaten many species simultaneously. Temporal correlations have been noted between specific environmental variables and the abundance of unrelated species depending on shared habitats and food resources (Zhou et al., 2014) . Comparative demography examines ecological dependency when extrapolated over hundreds of generations, using patterns of similarity or difference to discern for how long the ecological dependency has held. Our initial approach to comparative demographic history did not concern itself with absolute estimates of population size and the magnitude of population change because we reasoned that even species locked in a tight mutualistic relationship might differ in their reproductive rates, and that other factors would also influence their respective population sizes (Wells et al., 2017) . For example, a plant might be limited by nutrient availability even if its pollinators are abundant (Mattila and Kuitunen, 2000) , and a parasite's population size may be limited by the availability of both primary and secondary hosts, or by the density of its host population in addition to its absolute size. We therefore limited our focus to directional change in estimated population size, identifying intervals of growth, stasis, or decline. By this (intentionally simple) formulation, we endeavored to understand when pairs of species grew or contracted in concert. To achieve this, their demographic histories must first be synchronized. Using methods briefly summarized below, we devised a method of curve-fitting constrained by reasonable estimates of the relative annual mutation rate of each member of the species pair ( Figure 5 ). Each interval was coded for each species as one of growth (1), stasis (0), or decline (-1). Subtracting the value for one species from the other in each interval identifies intervals of agreement or conflict in growth history. For example, if both species experienced growth, the difference in this index (1-1) is zero. The same would be true for intervals where each species declined, or each experienced stasis. Slight disagreement would be registered if, say, one population was stable but the other grew or declined (absolute value of the difference =1). Conflict (one population growing while the other is contracting) yields an index difference with an absolute value of 2. When compared across all time intervals in the demographic curves, species pairs that have the smallest average slope differences deserve further consideration. (It would be notable as well if two species had average slope differences close to 2, as this would suggest that they respond in opposite ways to variation in prevailing conditions). In theory, unrelated species might have average slope differences close to 1; in our experience, species sharing no obvious ecological association have averaged differences of 0.5. We cannot say whether this apparent resemblance derives from actual commonalities among all species so far considered, or whether this reflects some bias in the reconstruction process. That is, shared environmental responses may explain non-random episodes of population growth and decline in many species. Any systematic bias in estimating population size (say, consistent exaggeration of the most distant past) would inflate the extent of correlation between pairs of demographic estimates. That said, our prior work demonstrated significantly better fits when comparing true host-parasite species pairs to arbitrary species pairs. Additional evaluation of diverse biological data, and further exploration data simulated under defined (known) demographic histories, are needed to further clarify this question. We took up several case studies when introducing comparative PSMC (C-PSMC). These included the agent of severe human malaria (Plasmodium falciparum) in relation to its mosquito vectors and the great apes (gorillas, chimpanzees, and human beings). Here, we saw a striking correspondence between the ancient historical growth patterns of the parasite with its Anopheline vectors and with all great ape lineages in the ancient past, followed by a striking population explosion unique to the rise of human beings, alone ( Figure 6 ). That is to say, early human demographic history resembles patterns shared by non-human primates, but more recent primate declines contrast markedly with our exponential growth. It stands to reason that human-feeding mosquitoes that thrive in human-disturbed habitats would flourish in the Anthropocene, and that parasites reproducing in such mosquitoes and in us would likewise flourish: patterns of genomic variation in each appear entirely consistent with exactly this expectation. We also examined the demographic history of the parasite Trichinella spiralis in relation to swine in Europe and Asia, and noted with great interest a regional distinction in the growth histories of parasite that mirrors a corresponding (and previously published (Groenen et al., 2012) ) difference in the population histories of Asian and European swine. The regional specificity of evidently shared population histories in the parasite and this host provides J o u r n a l P r e -p r o o f additional confidence that historical information can be extracted with these tools, and supports phylogenetic inferences indicating that Asian and European populations of the parasite have been distinct since long before the domestication of swine in each region (Thompson et al. in prep) . We determined that the long-term growth history of a photosynthetic alga mirrored that of the anemone species with which it continues to partner in a symbiotic relationship. Given that warming waters cause coral bleaching events (Berkelmans and van Oppen, 2006) we had suspected that no such pattern of long-term concerted population growth might result. The data indicated otherwise, supporting a much stronger correlation of population growth between the species involved in an extant symbiotic relationship than with an unrelated coral from a different geographic region. Thus, short-term fluctuations in symbiotic associations may be bounded, reverting to long-established affinities. Lastly, we applied this analysis to the potatoes and the oomycete agent of potato blight. Here, comparing demographic histories yielded a provocative insight: only since the age of potato domestication has both the plant and its pathogen exploded in abundance. Prior to that time, we discern no convincing parallelism in their respective growth histories. A fair interpretation of these data suggests that the origins of this plant-pathogen coincided with the time of potato domestication (not earlier). It is tempting to speculate that domestication itself created conditions conducive to the establishment and spread of potato blight. Such a causal interpretation cannot be proven by mere temporal correlation; but this interpretation gains plausibility by examining phylogenies of the pathogen and plant species, which indicate rampant facultative opportunism in the ancestors of Phytophthora (Cooke et al., 2000; Gómez-Alpizar et al., 2007; Haas et al., 2009; Martin et al., 2014) . Demographic histories cannot be read directly from genetic data, but instead require fitting a demographic model. Most methods for doing this are based on the principle of the coalescent, which assumes that effective population size is proportional to the average number of genetic differences between individuals in the population (Rosenberg and Nordborg, 2002) . However, this number depends not only on the population size, but also the rate at which new mutations are introduced. More genetically variable samples are understood as having derived from larger, older ancestral populations; but scaling this requires knowing (or estimating) the mutation rate. In PSMC, the independent histories sampled by each of a diploid individual's two parents provides the basis for estimating the relative size and age of the population (Li and Durbin, 2011) . However, to scale this history into absolute terms of effective population size and years for comparison with other populations' histories, an annual mutation rate must be specified. The annual mutation rate can be derived empirically from the data when using methods that incorporate asynchronous sampling (including ancient DNA), such as the Bayesian Skyline Plot (Drummond et al., 2005; Heled and Drummond, 2008) ), or assumed based on previously J o u r n a l P r e -p r o o f published estimates. If a per-generation mutation rate has been previously published, the annual mutation rate can be estimated by dividing the generational mutation rate by the population's generation time. The generation time can, in turn, be estimated from life tables as the average age of mothers, or the age at first reproduction can be assumed as a lower bound (Staerk et al., 2019) . For rare or under-studied species, none of this information may be available. In such cases, a further option is to determine plausible ranges of mutation rate and/or generation time and determine which parameters best align a distinctive feature of the demographic history (N e peaks/troughs) to a known historical event, such as the Last Glacial Maximum or the date an invasive species was introduced. The demographic history of an ecologically related species may also be used. Such a fitting procedure can be facilitated by the C-PSMC program. Once their demographic history plots have been appropriately scaled, historical effective population sizes can be compared. However, it is important to remember that effective population size estimates (N e ) are not mere estimates of census size (N). Generally, effective population size is a fraction of census population size, reflecting variance in the proportion of individuals who actually reproduce and contribute to the long-term genetic diversity of the population (Frankham, 1995) . However, the ratio N e /N may vary over time if evolutionary or environmental changes lead to different rates of reproductive success (Turner et al., 2006) . To our knowledge, no study of long-term demographic history has yet attempted to account for these natural vagaries in N e /N, though short-term studies have documented them (e.g. (Shrimpton and Heath, 2003) ). Domesticating animals and plants has significantly altered how long they live and what proportion of them reproduce. Indeed, exponential growth of the human population has resulted in approximately 24-fold growth in only the last thousand years, and 7fold growth (an extra 6 billion people) in only the last 100 years. Because effective population size is computed as a harmonic mean that is sensitive to lower values, extant human variation resembles what would be expected for a population numbering only tens of thousands; human beings will likely never accumulate genetic variability commensurate with our current abundance (as it would take more time than the Earth will likely remain habitable) (Kliman et al, 2008) . Such distinctions are worth remembering when inferring ecological information from genetic evidence. Population structure is another cause of variation in N e /N. In a metapopulation, subdivision restricts an individual to a subset of possible ancestors; this reduces N e /N. Conversely, recent admixture can inflate the N e /N ratio, by pooling the inheritance of long-separated demes (Baalsrud et al., 2014) . In both cases, the apparent change in N e /N is relative to the researcher's understanding of the true population structure. In demographic history reconstruction, these problems can largely be overcome by analyzing population structure before grouping samples into putative populations for coalescent analyses (Heller et al., 2013) . However, population structure may also have varied over time, introducing artifacts to the reconstructed historical effective population sizes. For example, after a glacial episode, populations previously isolated J o u r n a l P r e -p r o o f in refugia may then appear to grow in effective population size due only to their subsequent admixture, even if the total population has not grown (Mazet et al., 2016) . A range of ecological processes could explain correlated demographic histories. Specialized predation, mutualism, parasitism or strong commensalism would all be expected to lead to demographic dependence in one or both species. On the other hand, a pair of species might simply be limited by a common environmental constraint (Hardie and Hutchings, 2010) , as may often hold true for populations in extreme environments (e.g. very cold or dry) or during extreme events such as glacial periods or El Niño years (Trillmich and Limberger, 1985) . However, shared dependency on a third factor is unlikely to account for tens of thousands of years of demographic correlation between a randomly selected pair of species in the temperate zone, judging by the variety of demographic histories which have been published. The dependency of an obligate parasite on its host simplifies the process of determining the cause of any correlation between their demographic histories. When a parasite is entirely dependent upon a single host species, its population is all but certain to be regulated by the availability of hosts (Arneberg et al., 1998; Buck and Lutterschmidt, 2017) . Of course, without strong a priori expectations about the identity of the host, attribution can still be complicated if there are multiple plausible host species which exhibit similar demographic histories. Species with complex demographic histories (i.e. those involving multiple discrete bottleneck and expansion events) will be most amenable to this comparative approach, since models involving more discrete demographic events afford more opportunities to identify concordant or discordant periods of growth or decline, making it easier to determine if and when a pair of populations experienced specifically intertwined fates. The methods we have employed require accurate assessment of heterozygosity across an organism's genome; certain genomic features and limitations of data collection may complicate this goal. For example, haploid organisms are devoid of heterozygosity, limiting the application of PSMC. Detecting heterozygosity may also be confounded by misalignment of repetitive regions (which can confuse duplicated regions as alleles of the same locus). Finally, most available data of this sort has been generated in short reads (at most 300 bp with current next generation sequencing technology). While 50 base pairs generally suffice to locate a read in a genome, the presence of repetitive elements may cause small misalignments of reads to the reference sequence, resulting in systematic local overestimation of heterozygosity. Haploid genomes are not uncommon among parasites. Falciparum malaria, for example, has been repeatedly sequenced in its haploid state. Nevertheless, a reasonable assessment of heterozygosity can be made from these organisms by imagining the diploid state that would be J o u r n a l P r e -p r o o f Journal Pre-proof formed by joining two such haploid sporozoites in the ookinete (akin to any fertilized oocyte or ovule). There are several means to synthesize such hypothetical diploid genomes (pseudodiploids) from two haploid genome sequencing projects. The consensus sequences for two haploid individuals may be generated from a typical genome sequencing project and then aligned over the entire length of chromosomes (or contigs and scaffolds depending on the status of the genome assembly) using any number of alignment tools (Armstrong et al., 2019) . However, these alignment tools are not perfect and often leave considerable gaps between chromosomal sections, rendering true heterozygosity calls across entire chromosomes difficult. As an alternative, short-reads from each of two sequencing projects may be simultaneously mapped to a common reference sequence. Assuming equal numbers of reads from each haploid sequencing project, the resulting assembly should have generally comparable numbers of reads from both "parental" haploid sequencing projects at every locus; where disagreement occurs between the two sequenced samples, a consensus basecaller will indicate a heterozygous position. This method, which we have successfully employed, removes the need to curate errorprone chromosome alignments. The combined reads should not introduce large amounts of new variation in the genome assembly and should reliably reflect the mating of the two haploid lineages sampled. One caveat is that the haploid genomes should derive from the same interbreeding population (Mazet et al., 2015) . Creating pseudodiploids from two isolates from separate breeding populations would violate the assumption that the sampled genome derives from a population close to mutation/drift equilibrium, which is at the core of demographic reconstruction programs. Repetitive regions of the genome notoriously impede confident alignment of short reads (Treangen and Salzberg, 2011) . These repetitive regions may derive from short tandem repeats (microsatellites), interspersed nuclear elements, transposable elements, segmental duplications, and a variety of other less common repeat types (Biscotti et al., 2015) . A single duplication can quickly cause short-read alignment problems as the two copies diverge from each other within the same genome, making it difficult for algorithmic alignment. Misaligned reads can systematically inflate local estimates of heterozygosity. Consensus base calls rely on proper mapping of multiple reads to a specific homologous portion of the genome. If two duplicates have diverged by 10%, misalignment would introduce 3 heterozygotes for a 300 bp read. These incorrectly assigned heterozygotes may confuse demographic reconstruction programs by inflating the estimated occurrence of anciently-diverged alleles (characterized by elevated heterozygosity). PSMC treats short repetitive regions by scoring them as a single heterozygous recombination block. However, if the repetitive region extends for thousands of basepairs, multiple heterozygous blocks will provide erroneous evidence for portions of the genome harboring ancient variation. Overestimating the occurrence of such "ancient" sections would bias estimates of ancient population size. We scrutinized the occurrence and impact of such misalignment errors first by algorithmic screening for known eukaryotic repetitive elements and eliminating them from subsequent analyses by "masking" them. Repeat masking can be performed before or after aligning shortreads to the reference sequence, but should be accomplished prior to submitting the assembly to demographic reconstruction. Repetitive regions can also be suspected when a disproportionate J o u r n a l P r e -p r o o f number of reads maps to portions of the genome: too many reads will map to a "locus" if it has been condensed from repetitive paralogs. Elevated read depth therefore signals the possibility of local misassembly. Removing such suspect regions may provide less biased estimates of historical population size. (Some variation in read depth does occur even when repeats are not the cause; in practice, areas with greater than 2-fold difference in mean read depth might deserve to be pruned from an assembly prior to reconstructing demographic histories). A more conservative approach would remove all sections of the genome with greater than mean coverage, though this may sacrifice some areas of legitimate data harboring real biological signal. If genome sequences are long enough (longer than 1 million base pairs after unusual readdepth removal), the demographic signal should remain despite aggressive pruning. Just such a need arose when reconstructing the demographic history of the potato which, like many related plants, underwent a tripling of the genome (Jiao et al., 2012) . Anticipating that this would result in multiple misalignments of reads, we produced demographic reconstructions of from the genome before and after correcting for regions suspected, on the basis of excess read coverage, as misaligned. In this case, we eliminated all sections of the genome assembly that had greater than mean read depth. By eliminating suspected duplicated regions, we drastically reduced estimates of ancient population size while not materially affecting the shape of the demographic curve. Our conclusions concerning the history of the potato and potato blight proved robust to this potential confounder. Populations are limited by those factors that most constrain per-capita growth rate (Berryman, 2004) . A combination of such limiting factors determines the maximum sustainable size to which a population can grow under prevailing environmental conditions, defining its realized niche. Identifying the contemporary limiting factors of population size and growth rate through demographic sensitivity analysis is one of the main contributions of population ecology to conservation, as it is assumed that the factors most limiting of a population's growth are those most deserving of stewardship (Conroy and Brook, 2003; Gerber and Heppell, 2004) . However, this defines only current constraints. It is not usually possible to account for compensatory changes in other parameters, such as mortality being shifted onto a different life stage or the antagonistic coevolution of a parasite (Halpern et al., 2005) . However, feedback processes like these may have left their mark on the demographic response to analogous events in the past. In time, the factors that limit instantaneous population size and growth rate drive demographic history, as change in one of them may cause growth or decline in the population. Comparative demographic history might be able to provide evidence to complement demographic sensitivity analysis, trading precision for a broader perspective. For example, Zhou et al. (2014) found that the population sizes of the giant panda and snub-nosed monkey have tracked the abundance of leaves and seeds in their native habitat for at least a million years. Bottom-up control by plant productivity would seem the most likely explanation. While an analysis of their modern population would also indicate that their population is chiefly limited by habitat/food availability, comparative demography shows that this relationship has persisted through significant environmental change, with the same limitation being in place even when their food J o u r n a l P r e -p r o o f was much more abundant, suggesting that there is much scope for improvement. In contrast, another species' population size might be found to have tracked the abundance of their primary food only during periods of low productivity (e.g. glaciations), but instead tracked the inverse abundance of a predator during periods of higher productivity. When the generation time and mutation rate of a species can be estimated with confidence (and assuming these have not appreciably changed through time), it is possible to estimate when episodes of population growth or contraction took place. This implies that the reverse is also true: if the timing of a population's growth or decline can be assumed a priori (driven, say, by the retreat of glaciers, the birth of a volcanic island, or the flourishing of a biological resource), then these methods provide a new means to estimate generation time and/or mutation rate. Assuming anthropogenic causes for the recent growth in Anopheline mosquitos provided just such an example, yielding a mutation rate estimate considerably consistent with prior empirical estimates (Hecht et. al. 2018) . Given enough information, the comparative demography approach can also be used to estimate these parameters. Generation time and N e /N are also related to demographic sensitivity and conservation outcomes because they reflect recruitment rate of juveniles, as well as the genetic resources of a population and the rate at which it can recover from disturbance or grow in response to a successful conservation intervention (Belmar-Lucero et al., 2012; Frankham, 1995; Palstra and Fraser, 2012; Romiguier et al., 2014) . For the reasons outlined above, we believe much can be gained by comparing the inferred demographic histories among biological species. The purpose of this review is to provoke critics, inspire adherents, and otherwise incite the scholarly community into testing and improving upon this approach to better address pressing and as-yet unimagined questions. We view C-PSMC as a step towards extracting ever more rigorous information on the history of biological dependencies and, more broadly, the concerted response of biological communities to changing conditions for life on the planet. We can imagine more formal approaches to modeling the hypothesized population dependencies among various groups of organisms. Methodological research might focus on developing models seeking to integrate demographic reconstruction with the detection of dependent population dynamics (say, in the form of a free parameter for correlation between effective population sizes). For example, we could imagine a Bayesian framework to evaluate whether, and to what extent, estimates of a given species' history benefits by considering information derived from exogenous sources (such as the climate record or the demographic histories of other organisms). As increasing statistical sophistication is being integrated into efficient bioinformatic analyses of heretofore unimagined volumes of data, we hope comparative demography will be embraced as a concept worth exploring, refining, and improving. Below are some questions we invite the community at large to consider as worthwhile endeavors. We advocate development of methods to compare among multiple reconstructed demographic history plots to identify good matches among those derived from various interacting species. Using such a tool, one could test explicit a priori hypotheses regarding the longevity of species interactions. Such a tool could use any number of metrics to measure changes over time, whether it be slope, inflection points, or more complex statistical representations of shape and curvature. Ultimately this could be expanded to a likelihood algorithm that compares new demographic reconstructions to a database of all of those that have previously been published, returning a series of best "hits" in the database, like the BLAST algorithm used for DNA sequences (Altschul et al., 1990) . Furthermore, a tool that could break down portions of demographic curves into different time segments could help delineate specific events that changed species relationships from independent to dependent, or vice versa. This could provide many interesting insights into species interactions and, by examining many curves, could provide supporting evidence for periods of great biological upheaval or unusual stability. Iterating over many curves from independent studies might resolve general patterns revealing how groups of species have jointly responded to common environmental perturbations. The establishment of a tool that matches curves that span over 100,000 years in a database could provide tremendous insights into adaptability characteristics of extant species. Why stop at comparing a mere pair of demographic curves? The growth and contraction of entire landscapes (through i.e. glaciation, deforestation, desertification) has influenced the abundance and distribution of entire communities (Allen, 2003; Asner et al., 2004; Binney et al., 2017) . The genomes of entire suites of organisms may retain a record of their common fate, as appears to be true for leaf-eating monkeys and pandas with overlapping habitats. Would such a principal be evident on a far broader scale? Tropical rainforests, now critically threatened, are the home to the Earth's terrestrial biodiversity (Bradshaw et al., 2009) . One reason for this is that the tropics have nurtured communities for longer (other explanations include more abundant sun and rainfall, greater habitat fragmentation, and greater resource competition) (Pianka, 1966; Rohde, 1992; Willig et al., 2003) . Whereas temperate regions have undergone episodic climate extremes (for example, glacial advance and retreat), the tropics may be viewed as more enduring nurseries. Comparative demography holds promise for establishing whether species endemic to tropical rainforests have indeed experienced more stable populations, over longer intervals, than comparable species native to subtropical, temperate, or boreal regions, and whether species interactions (i.e. pollination, predation, parasitism) have endured longer. Species least capable of withstanding or escaping environmental extremes should be especially vulnerable to habitat loss and climate change (Foden et al., 2013) . Their genomes may harbor J o u r n a l P r e -p r o o f evidence of this. Have such species undergone notably episodic growth and decline? Conversely, more resilient and more vagile species might show greater independence in their demographic histories. Do actual data bear out this expectation? Comparative demography seems a natural approach to addressing such a question, especially if modified to jointly consider multiple demographic histories. Biological communities are structured by nutrient and energy flows from photosynthetic producers, herbivores, predators, and their parasites (Hairston and Hairston, 1993; Kuris et al., 2008; Odum, 1968) . How does an organism's ecological role influence its long-term demographic responsiveness to changing circumstances? Have the primary producers in a given landscape responded, in concert, to changing climatological conditions? Are such effects mirrored, dampened, or exaggerated at higher trophic levels? And are such patterns characteristic of particular landscapes, or can more general patterns be discerned through a comparative demographic lens? Changing transmission opportunities induce marked changes in pathogen abundance and distribution. Human pathogens, for example, reflect changes in human abundance, behavior, hygiene, population density, migration patterns, trade (Araújo et al., 2009; Zanotto et al., 1996) . Indeed, viruses have provided notably informative examples of the evolutionary dimensions of population growth and spread (Bush et al., 1999; Ferguson et al., 2003; Rambaut et al., 2004; Simmonds et al., 2019; Wertheim et al., 2019) . We see much more opportunity in using a comparative demographic approach to assess outstanding questions in historical epidemiology and, for example, the history of host shifts (respecting the specific requirements of each method for demographic reconstruction). A comparative framework can contribute to understanding whether, for how long, and where a given pathogen's ancestors have depended on a given host. Similarly, we believe such tools could be used to better understand the impact of specific interventions (i.e. vaccines) or of general environmental conditions (i.e. hygiene) in minimizing the prevalence and impact of specified pathogens. It is reasonable to suppose that predatory, pathogenic, or parasitic organisms may, in the short term, reduce the abundance of the organisms upon which they depend. As consumers generally benefit from growth in their resources, predators and parasites should flourish with the success of their prey and hosts over the longer term. Conversely, adverse conditions may undermine their collective welfare. Thus, population dynamics may be negatively correlated short-term but positively correlated in the long term. Indeed, such scale-dependent contrasts might provide signature evidence for enduring parasitism. (Persistent predation or resource competition might engender similar patterns). As discussed above, estimates of effective population size derive from harmonic means, driven more by long-term trends than by short-term fluctuations. Therefore, each time scale provides valuable information on the nature of antagonistic biological dependencies. Our species is blessed with the ability, and cursed with the responsibility, to understand our impact on our surroundings. Every human economic activity directly or indirectly requires the use of natural resources, the harvesting of nature's bounty, the cultivation of domesticated crops and livestock for human consumption, and the alteration of the conditions of life for wildlife, pests, weeds, pathogens, and the like. Distinct events in the human story have unleashed intensified waves of anthropogenic impact, including the advent of agriculture, the advent of global seafaring, the industrial revolution, the green revolution, and the advance of hygiene, public health, and medicine. We are surrounded by the evidence of our immense impact on the demography of countless other organisms, be it billions of chickens or zero dodo birds. In the history of life on the planet, human beings are now said to be causing its sixth global extinction crisis. Meanwhile, an estimated 11% of the world's land is under crop cultivation, approximately 80% of which feeds livestock. The biomass of our livestock outweighs our own biomass, and together outweigh all other vertebrates (excepting fish) by over twenty times; the globe's chickens collective ly weigh more than three times the globe's wild birds (Bar-On et al., 2018) . Overfishing, overhunting, habitat destruction, resource competition, climate forcing, and pathogen pollution have driven the most widely-recognized cases threatening species with extinction (such as whales, cheetahs, elephants, coral declines, the global die off of amphibians) but we remain only dimly aware of the extent and pace at which we drive the fate of life on the planet, writ large. We do not need comparative demography to understand these basic truths. But comparative demography does offer additional means to delineate the pace and origins of our impact. Perhaps, one day, it can be used to identify cases where we succeeded in reversing course. Figure 1: Comparative demography seeks to infer whether, and when, distinct species have undergone shared intervals of population growth and contraction. Here temporal changes in the estimated effective population size of each of two hypothetical species has been illustrated. Note that the abundance of each species appears to have grown and contracted in concert over one temporal interval (to the left of the vertical dashed line; =) but at another interval, one species grew as the other contracted (to the right; ≠). Changes in population size influence when alleles coalesce. A. Here, the expected temporal distribution of branching in a population of constant size (above, orange) is contrasted with that expected for a population that has undergone a bottleneck (below, blue) (after Erikkson et al, 2010) . In the constant-sized population, fewer coalescences are ascribed to the interval represented by the grey shading. B. The age of alleles diminishes exponentially in populations of constant size, but bottlenecks accelerate allelic loss, temporally concentrating coalescence events. C. Therefore, intervals of notably frequent coalescence events may demarcate periods of past population decline. Similar reasoning underlies procedures used to identify periods of population growth. Other factors (selection, population subdivision) may similarly influence such temporal patterns. A single diploid genome can be harnessed to estimate the age distribution of maternal and paternal alleles. Blocks of the genome are characterized by their heterozygote density, which is assumed to reflect for how long they have been accumulating differences. Loci distinguished by few heterozygotes are assumed to have diverged more recently than loci distinguished by many heterozygotes. C. Population decline favors allelic loss, whereas population growth favors allelic persistence. Adjusted for statistical expectations (which assume that most alleles in any given population are young), past population size estimates (red line) are inversely related to the age distribution of alleles in the genome (black histogram). Each segment of a temporally aligned pair of demographic history curves (A) is scored based on whether the population is growing, declining, or unchanged over that period (B). The residual differences in slope direction identifies periods of consistent or inconsistent population dynamics and summarize the overall fit between the pair of demographic history curves (C). The origins and durability of such interactions merit study but evidence is often lacking. Comparative Demography elucidates long-term species interactions by comparing histories of population growth and contraction. The method has broad application and untapped potential. J o u r n a l P r e -p r o o f Response of past and present Mediterranean ecosystems to environmental change Basic local alignment search tool Phylogeography and evolutionary history of dengue virus type 3 Whole-Genome Alignment and Comparative Annotation Host densities as determinants of abundance in parasite communities Grazing Systems, Ecosystem Responses, and Global Change Effects of population characteristics and structure on estimates of effective population size in a house sparrow metapopulation The biomass distribution on Earth Concurrent habitat and life history influences on effective/census population size ratios in stream-dwelling trout The role of zooxanthellae in the thermal tolerance of corals: a 'nugget of hope' for coral reefs in an era of climate change Limiting factors and population regulation Vegetation of Eurasia from the last glacial maximum to present: Key biogeographic patterns Repetitive DNA in eukaryotic genomes Tropical turmoil: a biodiversity tragedy in progress How will global climate change affect parasite-host assemblages? Parasite abundance decreases with host density: evidence of the encounter-dilution effect for a parasite with a complex life cycle Predicting the evolution of human influenza A Demographic sensitivity and persistence of the threatened white-and orange-bellied frogs of Western Australia A Molecular Phylogeny of Phytophthora and Related Oomycetes Coalescents and genealogical structure under neutrality Bayesian coalescent inference of past population dynamics from molecular sequences The total branch length of sample genealogies in populations of variable size Identifying the world's most climate change vulnerable species: a systematic trait-based assessment of all birds, amphibians and corals Effective population size/adult population size ratios in wildlife: a review When cold is better: climate-driven elevation shifts yield complex patterns of diversification and demography in an alpine specialist (american pika, Ochotona princeps) Detecting bottlenecks and selective sweeps from DNA sequence polymorphism The use of demographic sensitivity analysis in marine species conservation planning An Andean origin of Phytophthora infestans inferred from mitochondrial and nuclear gene genealogies Analyses of pig genomes provide insight into porcine demography and evolution Phylogenetic trees support the coevolution of parasites and their hosts Molecular phylogenies and host-parasite cospeciation: gophers and lice as a model system Disparate rates of molecular evolution in cospeciating hosts and parasites Cause-effect relationships in energy flow, trophic structure, and interspecific interactions Habitat size, recruitment, and longevity as factors limiting population size in stage-structured species Evolutionary ecology at the extremes of species' ranges Detecting the macroevolutionary signal of species interactions Comparative demography elucidates the longevity of parasitic and symbiotic relationships Bayesian inference of population size history from multiple loci The confounding effect of population structure on Bayesian skyline plot inferences of demographic history Skyline-plot methods for estimating demographic history from nucleotide sequences Evolution in action: climate change, biodiversity dynamics and emerging infectious disease A macroevolutionary mosaic: episodic host-switching, geographical colonization and diversification in complex host-parasite systems Distribution of Borrelia burgdorferi s.l. spirochaete DNA in British ticks (Argasidae and Ixodidae) since the 19th Century, assessed by PCR The coalescent process in models with selection and recombination A genome triplication associated with early diversification of the core eudicots Genetic and phylogenetic consequences of island biogeography Exotic plant invasions and the enemy release hypothesis The macroecology of population dynamics: taxonomic and biogeographic patterns in population cycles The Neutral Theory of Molecular Evolution Genetic drift and effective population size Maximum likelihood estimation of population growth rates based on the coalescent Ecosystem energetic implications of parasite and free-living biomass in three estuaries Inference of human population history from individ ual whole-genome sequences PSMC (pairwise sequentially Markovian coalescent) analysis of RAD (restriction site associated DNA) sequencing data Region-wide and ecotype-specific differences in demographic histories of threespine stickleback populations, estimated from whole genome sequences The Theory of Island Biogeography Detection of Borrelia burgdorferi DNA in museum specimens of Peromyscus A combined mitochondrial and nuclear multilocus phylogeny of the genus Phytophthora Nutrient versus pollination limitation in Platanthera bifolia and Dactylorhiza incarnata (Orchidaceae) Demographic inference using genetic data from a single individual: Separating population size variation from population structure. Theor On the importance of being structured: instantaneous coalescence rates and human evolution-lessons for ancestral population size inference? Polar and brown bear genomes reveal ancient admixture and demographic footprints of past climate change Parasites and the fossil record PSMC analysis of effective population sizes in molecular ecology and its application to black-and-white Ficedula flycatchers Energy flow in ecosystems: a historical review Effective/census population size ratio estimation: a compendium and appraisal Evolution of insect pollination and angiosperm diversification Ticks parasitised feathered dinosaurs as revealed by Cretaceous amber assemblages Detection of Borrelia burgdorferi DNA in museum specimens of Ixodes dammini ticks Latitudinal gradients in species diversity: a review of concepts Trends in the evolution of insect parasitism by nematodes as inferred from fossil evidence Evidence of intestinal parasites of dinosaurs An integrated framework for the inference of viral population history From reconstructed genealogies The causes and consequences of HIV evolution Latitudinal gradients in species diversity: the search for the primary cause Comparative population genomics in animals uncovers the determinants of genetic diversity Genealogical trees, coalescent theory and the analysis of genetic polymorphisms Inferring human population size and separation history from multiple genome sequences Methods and models for unravelling human evolutionary history Community ecology theory as a framework for biological invasions Census vs. effective population size in chinook salmon: large-and small-scale environmental perturbation effects Prisoners of war -host adaptation and its constraints on virus evolution Pairwise comparisons of mitochondrial DNA sequences in stable and exponentially growing populations Inference of population history using coalescent HMMs: review and outlook Performance of generation time approximations for extinction risk assessments Robust and scalable inference of population history from hundreds of unphased whole genomes Biological Invasions in Marine Ecosystems: Ecological, Management, and Geographic Perspectives, Ecological Studies Introduced species and their missing parasites Release from parasites as natural enemies: increased performance of a globally introduced marine crab Repetitive DNA and next-generation sequencing: computational challenges and solutions Drastic effects of El Niño on Galapagos pinnipeds Life history and environmental variation interact to determine effective population to census size ratio Environmental conditions and prey-switching by a seabird predator impact juvenile salmon survival Natural selection favoring more transmissible HIV detected in United States molecular transmission network Latitudinal gradients of biodiversity: pattern, process, scale, and synthesis A Primer of Population Biology, 1 edition Population dynamics of flaviviruses revealed by molecular phylogenies Population genomics of finless porpoises reveal an incipient cetacean species adapted to freshwater Whole-genome sequencing of the snub-nosed monkey provides insights into folivory and evolutionary history The authors of the manuscript "Assessing the evolutionary persistence of ecological relationships: a review and preview" hereby declare that we have no financial interests in the contents of the submitted manuscript.