NATURAL HISTORY COLLECTIONS ARE NEEDED TO RESOLVE HOST SAMPLING GAPS IN PARASITOLOGY: INSIGHTS FROM AVIAN HAEMOSPORIDIANS
ABSTRACT
The field of parasitology, and thus biodiversity research more broadly, is faced with the unfortunate reality that our understanding of parasite biodiversity can be only as good as our ability to sample parasite host species. Although the sampling of many host species is trivial, there typically remains a subset of species across any host group of interest that is difficult to sample due to rarity, habitat, body size, or some other trait. The result is a blind spot in our understanding of parasite biodiversity that is centered around parasite species that infect hosts that are rarely sampled by humans. However, for many groups of hosts, the daunting task of obtaining host samples has already been done, and these samples exist in the form of natural history collections at institutions across the world. With avian malaria parasites and other haemosporidians as an example, I demonstrate that significant host sampling gaps exist in the United States and Canada. Bird species that have not been sampled for molecular haemosporidian research typically are associated with aquatic habitats, significantly greater masses, and more restricted geographic distributions than are bird species that have already been sampled. These unsampled host species are likely to be infected with a high richness of previously undiscovered avian haemosporidian genetic lineages. However, natural history collections in the United States can be used to nearly completely address these sampling gaps with tissue samples currently housed in these institutions. The result of this analysis indicates that the future of parasite biodiversity research is dependent on the use and support of natural history collections and other biorepositories.
An accurate accounting of global parasite diversity has never been more important. As interest in disease ecology and evolution has grown, the realization that our understanding of parasite diversification and community diversity is incomplete because of gaps in taxon sampling has come sharply into focus (Poulin, 2010; Carlson et al., 2020). From a macroevolutionary perspective, missing taxa in parasite phylogenies impact our ability to reconstruct evolutionary history (Hillis et al., 2003) and understand trait evolution (Ackerly, 2000), diversification rates (Heath et al., 2008), and historical biogeography (Turner et al., 2009). Although not well studied, missing parasite taxa likely also have a profound impact on studies of the process of parasite diversification such as host switching and cospeciation because these processes mirror biogeographic events of free-living organisms (Ronquist, 1997). Ecologically, incomplete surveys of organismal communities can mislead efforts to understand the drivers of macroecological diversity patterns (Menegotto and Rangel, 2018). A complete inventory of parasite diversity is of the utmost importance for understanding how and why novel diseases emerge in human, livestock, and wildlife populations (Cook et al., 2020).
Although species description is one important component of characterizing parasite biodiversity, a more proximate issue is the recognition of the existence of diversity in the first place, often through molecular surveillance. A common approach for molecular parasite surveys consists of sampling host organisms and preserving some form of host tissue that is then tested for the presence of the target parasite group by using polymerase chain reaction (PCR) gene amplification followed by DNA sequencing to generate genetic data that can be used for basic and applied research. Unfortunately, the success of molecular surveillance of parasite diversity is limited by our ability to adequately sample host diversity. Often, host surveys are performed through passive means such as mist netting for birds or trapping for small mammals, which limits the taxonomic breadth of host captures (Tattoni and LaBarbera, 2022). By focusing parasite surveys on a biased subset of the host community, we are certainly missing a substantial component of the parasite community.
Fortunately, biodiversity repositories such as natural history collections can address some of the challenges posed by difficult-to-sample host species. Most relevant for parasite detection and discovery are specimens and tissues stored in alcohol or a buffer solution or that are frozen (cryo collections) (Dunnum et al., 2017; Thompson et al., 2021), which enable parasitologists to use molecular methods to test these samples for the presence of a parasite of interest. Through field expeditions that engage in active collecting and through salvaged specimens, tissue collections are continuously growing representations of earth’s biodiversity, including both hosts and the symbionts preserved with them (Miller et al., 2020). The use of tissue collections has been advocated for in recent years as an approach for studying emerging infectious diseases (DiEuliis et al., 2016; Colella et al., 2021), although their utility also clearly extends to basic parasitological and ecological research (Fecchio et al., 2019; Wood and Vanhove, 2023).
There may be no better illustration of the utility of museum collections for parasite biodiversity research than the apicomplexan order Haemosporida, which encompasses a diverse clade of vector-transmitted parasites that use vertebrate tissues for asexual reproduction. Hundreds of species of haemosporidians have been described from vertebrate blood primarily by using parasite morphology for species commonly infecting mammals, squamates, chelonians, and birds (Perkins, 2014). However, our understanding of the global scope of haemosporidian biodiversity has undergone a revolution over the last quarter century. Two developments ratcheted haemosporidian biodiversity research into high gear, particularly over the last 2 decades. In quick succession, several highly effective and robust protocols were developed to amplify and sequence haemosporidian cytochrome b (cytb) mitochondrial DNA barcode sequences (e.g., Hellgren et al., 2004; Waldenström et al., 2004), techniques that were rapidly adopted by haemosporidian researchers. Researchers then began using available blood and tissue samples of vertebrate species of interest to quickly screen large numbers of samples for the presence of haemosporidian DNA. In tandem, these advances revealed a massive diversity of cryptic genetic lineages among haemosporidians, particularly within the avian haemosporidians (Bensch et al., 2009), revealing the potential for many thousands of undescribed species globally and the existence of an emerging model for the study of such diverse topics as coevolution, parasite biogeography, and parasite community ecology.
A downside to the haemosporidian molecular revolution has been the daunting realization that there is an incredible amount of work yet to be done to simply discover the global extent of haemosporidian diversity. One of the biggest challenges in this regard is host sampling—put simply, samples from many known and likely hosts of haemosporidians are difficult to obtain. Fortunately, biodiversity collections can facilitate a profound leap forward for the study of haemosporidian diversity and parasite biodiversity more broadly. In this study, I argue that collaboration between parasitologists and biodiversity collections is absolutely essential for the future of biodiversity research. Using birds and their haemosporidians across the United States and Canada as a case study, I identified host species gaps in the sampling record for molecular haemosporidian research and have summarized how biodiversity collections can effectively address these gaps and rapidly accelerate our understanding of haemosporidian diversity.
MATERIALS AND METHODS
Identification of host sampling gaps in avian haemosporidian research
Although avian haemosporidians have been studied for over a century, only within the last 2 decades have we begun to understand the scope of avian haemosporidian diversity through the use of molecular tools to characterize genetic lineages. This present analysis was focused specifically on gaps in host sampling for studies that use amplification and DNA sequencing to detect haemosporidians and to characterize their genetic diversity. Hosts that have been sampled to study haemosporidians strictly based on microscopy were not considered.
I focused this analysis geographically on the United States and Canada because of my familiarity with the bird species that occur in this region and my ability to reconcile the taxonomies used by the various databases that I accessed for this study. Currently, the United States is the most well-sampled country for molecular surveys of avian haemosporidians (12,681 hosts sampled as of 27 March 2024 according to the MalAvi database; Bensch et al., 2009), and Canada is currently the eighth-most sampled country (3,456 hosts sampled). Therefore, sampling biases that exist in the United States and Canada are likely to be reflected (and possibly accentuated) in other regions of the world where sampling has been less comprehensive.
I used 2 databases as the basis for identifying host sampling gaps in the United States and Canada. To characterize the total pool of avian host species that could potentially be sampled for haemosporidian research in this region, I used the list of bird species recorded in the United States and Canada during the North American Breeding Bird Survey between 1966 and 2022 (2023 release; Ziolkowski et al., 2023). The North American Breeding Bird Survey is an annual assessment of the presence and abundance of birds across over 4,000 survey routes primarily in the United States and Canada. I used the American Ornithological Society checklist of North American birds (Chesser et al., 2023) to remove vagrant and introduced species from the list of bird species under consideration.
I then used the MalAvi database (version 2.5.8; Bensch et al., 2009), which is a comprehensive database of avian haemosporidian cytb haplotypes that have been sampled globally. Here, I refer to haemosporidian cytb haplotypes as genetic lineages, following the standard in the field that has been set by studies showing that haemosporidian cytb haplotypes that differ by a single base pair can exhibit signatures of reproductive isolation (Bensch et al., 2004; Nilsson et al., 2016).
I used the MalAvi “hosts and sites table,” which summarizes the sampling location and host in which each parasite genetic lineage has been found. Although the MalAvi database has not incorporated every published study on avian haemosporidian genetic research, it is the most comprehensive database of avian haemosporidian molecular surveys in existence and so probably closely reflects the state of host sampling in this field. However, the MalAvi database does not record host species that have been surveyed for haemosporidians but for which no infections were detected and sequenced. To address this limitation, I manually searched Google Scholar for each bird species from the United States and Canada that was not recorded as a host species on MalAvi. I used the search phrase “host species AND haemosporidian” (where “host species” was replaced with the taxonomic name of the host species of interest) and limited this search to studies published since 2000. This search revealed 20 studies that were not found on the MalAvi database in which a bird species was surveyed for haemosporidians in the United States and Canada by using molecular methods. The host data from these studies were combined with the MalAvi database for analysis.
I used a custom R script (available from Figshare repository, https://doi.org/10.6084/m9.figshare.26414911.v1) to compare the MalAvi database with the Breeding Bird Survey species list and summarize which bird species have been surveyed for haemosporidians and if surveyed how many of each host species have been tested for infection.
Data analysis
I tested whether the bird species that have not been surveyed for haemosporidians in the United States and Canada differed morphologically and ecologically from those that have been sampled. To assemble trait data for breeding bird species in the United States and Canada, I used the AVONET dataset (Tobias et al., 2022), which is a global trait database for over 11,000 bird species. I specifically extracted the variables of mass, habitat, and range size for each bird species in the database of breeding species in the United States and Canada. I assessed the effect of bird species mass on sampling status to test for an effect of the overall size of a bird species on the probability that it was sampled. Habitat was included to determine whether bird ecological preferences influence whether it has been sampled for haemosporidian research; for this variable, I simplified the categories to terrestrial, terrestrial aquatic (including wetlands and rivers), and coastal/marine to improve model fit. I included range size to test whether bird species with larger geographic ranges were more likely to be sampled for haemosporidians than species with smaller geographic ranges. I used a phylogenetically controlled logistic regression with the phylolm package (Tung Ho and Ané, 2014), with the binary variable of “has been sampled for haemosporidians” as the dependent variable and the 3 AVONET variables described above as the predictor variables. Both mass and range size were log transformed for analysis. Because this analysis included bird phylogeny in the model, I used a phylogeny for all bird species in the dataset generated from www.birdtree.org (Jetz et al., 2012). I downloaded 100 trees using the “Ericson All Species” option and selected a topology from this distribution at random to be included in the regression model.
I also tested whether unsampled avian host species in the United States and Canada are likely to be infected with undiscovered haemosporidian lineages using the codependent package in R (Carlson et al., 2019). The “copredict” function uses bipartite species association networks to estimate the total richness of host affiliates (in this case, parasites), given the total number of host species, by fitting a power law function and extrapolating to the unsampled host species in the network. This approach follows from those of Strona and Fattorini (2014) and Carlson et al. (2020), who found that global helminth richness can be fit to a power law curve and that this approach is superior to the assumption of a linear relationship between host and parasite diversity. Here, I focused on estimating the richness of haemosporidian genetic lineages because species limits among avian haemosporidians are not widely characterized. However, given that haemosporidian genetic lineages that differ by as little as 1 base pair can exhibit signatures of reproductive isolation, lineage richness is likely strongly correlated with species richness in this group of parasites. For this analysis, I estimated haemosporidian lineage richness in 7 bird orders that have been relatively well sampled for avian haemosporidians in the United States and Canada (a minimum of 25% of species in the order sampled and at least 5 haemosporidian lineages documented that are exclusive to that bird order) to test the hypothesis that there are undiscovered haemosporidian lineages even in bird groups that have been the focus of previous research. All haemosporidians from the genera Haemoproteus, Leucocytozoon, and Plasmodium were grouped together for this analysis. Because the purpose of this analysis was not to estimate the total haemosporidian richness that infects these bird orders but rather to document the existence of parasite richness that could be discovered with additional host species sampling, I restricted networks to include only haemosporidian lineages that are specific to the bird order being analyzed (i.e., generalist, multiorder parasites were not accounted for). As a result, this analysis was limited in scope to the estimated number of haemosporidian lineages that are expected to be discovered by exclusively sampling each separate bird order.
Assessment of the potential for natural history museum collections to fill host sampling gaps in avian haemosporidian research
I surveyed 5 natural history museum tissue collections in the United States to test whether current avian haemosporidian host sampling gaps could be addressed using samples that exist in these institutions. I focused on the samples in tissue collections at these institutions because haemosporidians can be reliably detected and genetically characterized from avian tissues (e.g., muscle, liver, and heart; Fecchio et al., 2019).
The 5 institutions were selected to represent important natural history collections in distinct regions of the United States: the American Museum of Natural History (New York, New York), the Field Museum of Natural History (Chicago, Illinois), the Louisiana State University Museum of Natural Science (LSUMZ; Baton Rouge), the University of New Mexico Museum of Southwestern Biology (Albuquerque), and the University of Washington Burke Museum (Seattle). I used the advanced search options on VertNet (Constable et al., 2010; www.vertnet.org) to search each institution for samples in their collections that fit the following criteria: the samples were collected from birds (class = Aves), the samples were collected in the United States (country = United States), and the filter option “has tissues” was selected. These options were selected to identify bird samples that could potentially be used to test for the presence of avian haemosporidians in associated tissue samples that are archived in these institutions.
Using the output of the searches, I reconciled the avian taxonomy between each institutional database and the Breeding Bird Survey species list that I used to identify host sampling gaps. Next, I summarized the tissue collection holdings of these 5 institutions at the species level to identify which species from the Breeding Bird Survey list are represented in their tissue collections. I then compared the list of host species that have not been surveyed for avian haemosporidians with the tissue collection holdings of the 5 institutions and identified bird species that could potentially be screened for haemosporidians using these collections to fill existing host sampling gaps. All data manipulation of the natural history museum tissue collection databases was performed using a custom R script (available from Figshare repository; https://doi.org/10.6084/m9.figshare.26414911.v1).
RESULTS
Summary of host sampling gaps in avian haemosporidian research
I identified significant gaps in the host species that have been sampled for molecular avian haemosporidian research in the United States and Canada (Fig. 1; Table I). Overall, just 41.6% (262 of 630) of the bird species that were recorded by the Breeding Bird Survey in the United States and Canada between 1966 and 2022 have been sampled for molecular haemosporidian research according to the MalAvi database and an additional search of the literature. Across the 262 sampled bird species in the United States and Canada, 19,894 samples have been tested for avian haemosporidian research (Table I; this total includes the samples reported on MalAvi plus those in the additional literature search). The proportion of bird species that have been sampled for haemosporidian research was highly heterogeneous across avian orders. Comparing birds in the order Passeriformes (i.e., songbirds, which make up approximately half of all known bird diversity) with birds not in the order Passeriformes, I found that 62.2% of Passeriformes have been sampled, whereas just 24% of non-Passeriformes have been sampled in the United States and Canada. The majority of total bird samples from the United States and Canada used for haemosporidian research were from Passeriformes (64%, 12,799 of 19,894 samples).
Citation: The Journal of Parasitology 111, 6; 10.1645/24-57
Within the Passeriformes, several families were noteworthy for being well sampled, such as Cardinalidae (14 of 14 species sampled), Corvidae (13 of 17 species sampled), and Turdidae (11 of 12 species sampled). However, several large passeriform families have been poorly sampled, such as Passerellidae (24 of 42 species sampled) and Tyrannidae (15 of 34 species sampled). Across the non-Passeriformes, sampling proportion also differed dramatically. There was no non-Passeriformes order with more than 50% of its members sampled (aside from the Cathartiformes and Ciconiiformes, which are represented by 3 species and 1 species in the United States and Canada, respectively). The Strigiformes (8 of 19 sampled species) and Galliformes (10 of 20 sampled species) were the most thoroughly sampled orders among the non-Passeriformes. In contrast, several orders were nearly or entirely unsampled. The Coraciiformes, Podicipediformes, Procellariiformes, Suliformes, and Trogoniformes were entirely unsampled in the United States and Canada, although each of these orders contains 11 or fewer species breeding in this region. The Anseriformes were the most well sampled non-Passeriformes order by total samples (4,136 samples), whereas total sample number for the remaining non-Passeriformes orders ranged from 0 to 850.
Using a phylogenetically controlled logistic regression, I found that bird species that have been sampled for molecular haemosporidian research in the United States and Canada have significantly lower masses (β = −0.35, P < 0.001) and larger geographic range sizes (β = 0.47, P < 0.001) and are more likely to be found in terrestrial environments (reference: marine/coastal; β = 2.81, P < 0.001) than species that have not yet been sampled (Suppl. Data, Table S1).
I estimated the richness of avian haemosporidian genetic lineages that infect a single host order for 7 of the most well sampled avian orders in the United States and Canada (Table II). For each avian order, estimated haemosporidian richness was at least 50% higher than the currently recognized haemosporidian richness. For instance, using the most well sampled avian order in the United States and Canada, the Passeriformes, I estimated a total Passeriformes-specific haemosporidian lineage richness of 1,086 lineages (95% confidence interval: 1,074–1,099) relative to the currently known richness of 688 lineages.
Capacity for natural history museum collections to fill host sampling gaps
Of the 368 avian host species in the United States and Canada that have not been surveyed for haemosporidians, 339 were represented by tissue samples in at least 1 of the 5 natural history museum collections that I surveyed (Table III). Across the 339 species that were present in natural history museum collections as tissues, the total number of samples that exist across all 5 institutions was highly variable (range: 1–1,267 samples; median: 20 samples), although just 38 of these species were represented by a single tissue sample across all 5 collections. Across the 5 collections that I surveyed, each collection held tissues of at least 47.5% of the unsampled bird species, with 1 collection alone (LSUMZ) holding tissue samples of nearly 83% of bird species in the United States and Canada that are unsampled for haemosporidian research (Table III).
The 29 species that were not represented by tissues collected in the United States and Canada across the 5 natural history collections that I surveyed could generally be categorized into 4 groups. Most of the species not represented in the surveyed tissue collections do regularly breed in the United States and Canada but occur in this region at the edges of their much larger distributions (e.g., Coccyzus minor, Pachyramphus aglaiae, and Luscinia svevica). Smaller groups of species not represented in the surveyed tissue collections include those that do not regularly breed in the United States and Canada (e.g., Chroicocephalus ridibundus and Sula nebouxii), are federally protected (e.g., Setophaga chrysoparia and Vireo atricapilla), or occur only over a small total geographic area (e.g., Polioptila californica and Peucaea carpalis). A summary of tissue holdings for unsampled bird species is available from Figshare repository (https://doi.org/10.6084/m9.figshare.26414911.v1).
DISCUSSION
I have demonstrated here that in the most thoroughly sampled region on earth for molecular avian haemosporidian research, there still exist substantial and striking gaps in the host species that have been surveyed. These unsampled host species are likely to be infected by a high richness of novel haemosporidian genetic diversity. This analysis provides a path forward for addressing existing host gaps by providing targets for future sampling in the United States, Canada, and elsewhere and indicating clearly that these gaps are unlikely to be reconciled without collaboration with natural history collections.
The most striking finding from this analysis is that haemosporidian researchers are far from having a complete understanding of haemosporidian diversity in the United States and Canada because a minority of bird species in this region have been sampled for the study of haemosporidians with modern molecular techniques. My analysis revealed several avian clades and avian habitats more broadly that should be prioritized for future haemosporidian biodiversity surveys. In particular, bird species that are closely associated with coastal and marine aquatic habitats have been undersampled relative to birds found in terrestrial environments. Several avian orders, namely the Anseriformes, Charadriiformes, Pelecaniformes, and Suliformes, are nearly or entirely restricted to aquatic habitats and have been severely undersampled for haemosporidians (the most well sampled of these groups is the Anseriformes, with 36.4% of species sampled in the United States and Canada). The gap in host sampling seen in aquatic bird groups most likely reflects the simple difficulty of sampling birds where traditional mist netting approaches are largely ineffective.
Although there are clear host sampling gaps in avian haemosporidian research, it is relevant to ask whether further host sampling is likely to uncover additional haemosporidian genetic diversity to an appreciable extent. Several lines of evidence suggest that the answer is most certainly a resounding yes. We can consider whether the avian host groups that have been undersampled in the United States and Canada are even known to host haemosporidians in the first place. For virtually every avian order that occurs in the United States and Canada, haemosporidians are not only known to infect that order but there are multiple morphologically described haemosporidian species for which the type vertebrate host belongs to that order (see Table 6 of Valkiūnas, 2004). Given that morphologically described haemosporidian species often represent clusters of genetically distinct and likely reproductively isolated lineages (Hellgren et al., 2007), novel haemosporidian diversity may be found in the avian host orders that have been undersampled. There are several extreme examples of a mismatch between the current state of sampling for a bird order in the United States and Canada and the documented existence of haemosporidians that infect that host group. For example, 10 morphologically described haemosporidian species appear to be restricted to the bird order Columbiformes (Valkiūnas, 2004), yet just 8 total samples have been taken from birds in this host group in the United States and Canada. Host groups such as the Columbiformes that are undersampled yet known to be host to a diversity of morphologically distinct haemosporidians should be prime candidates for future surveys.
We also can ask whether the rate at which previously undocumented genetic lineages of haemosporidians have been found is indicative of additional diversity that is yet to be discovered. Using an approach to extrapolate parasite diversity from host diversity with a power law function, I found that in 7 of the most well sampled avian orders in the United States and Canada there is still substantial haemosporidian genetic diversity left to be discovered. Even in the most well-sampled avian order, the Passeriformes, there still appears to be hundreds of haemosporidian lineages that are currently undiscovered in the United States and Canada. In other avian orders such as the Anseriformes, we could currently be undersampling haemosporidian richness by as much as a factor of 5. Although every remaining haemosporidian genetic lineage to be discovered probably does not reflect a reproductively isolated species, these analyses broadly suggest that our understanding of haemosporidian diversity in the Nearctic is far from complete.
The role of natural history collections in the future of haemosporidian research
Natural history collections represent extraordinary concentrations of resources for the study of the natural world and can profoundly expand the sampling ability of any single researcher or research group. The potential for parasitologists to take advantage of this host material is of paramount importance for the future of biodiversity research for many parasite groups. For avian haemosporidians in the United States and Canada, existing host sampling gaps could be almost completely addressed from collaboration with just 5 major natural history museum collections. Not that future efforts should rely on these specific institutions; the collections that I focused on here were chosen to demonstrate that expanding host sampling is feasible through a small number of existing collections in the United States. For the few unsampled bird species that were not represented in the tissue holdings of the 5 institutions that I surveyed, many species do have tissue samples archived in additional institutions that I did not assess in this study. For most bird species, the holdings of these institutions can both address basic gaps in haemosporidian research and provide adequate sample sizes from geographically widespread locations to estimate parasite prevalence at the host species level.
One limitation of surveying parasites from host tissue collections should be noted. Although molecular surveys from archived tissues have the potential for discovery of novel genetic lineages, species, and even higher groups of parasites (e.g., Yabsley et al., 2018), the morphological identification and description of the isolated organisms may rely on additional sampling or tissue preparations. For instance, the morphological description of haemosporidians is dependent on high-quality blood smears; thus, morphological identification of haemosporidians is impossible from preserved tissue samples. This limitation speaks to the importance of encouraging the collection and curation of multipart specimens (Webster, 2017) and the development of digital resources to link these specimens in an easily accessible format (Lendemer et al., 2020).
Parasitologists should not expect that access to samples from natural history collections will be a trivial matter. Researchers should have a well-defined plan and strong justification for access to these resources and must recognize that it is up to the discretion of the natural history collection to provide access to the tissues that they curate. However, the curators of natural history collections should also recognize that the use of tissue samples to study symbiotic organisms is a valid, important, and increasingly common use of the resources that they manage. As my analysis demonstrates, increasing communication and collaboration between natural history museum collections and parasitologists is certain to be a fruitful endeavor that has the potential to transform our understanding of haemosporidians and other parasite groups.
Phylogenetic tree obtained from www.birdtree.org for 630 bird species that were recorded during breeding bird surveys in the United States and Canada between 1966 and 2022. Black marks surrounding the phylogeny indicate bird species that have been sampled for molecular haemosporidian research (lack of a mark indicates that the species has not been previously sampled). Branches of the phylogeny marked in light gray represent bird species that are not in the order Passeriformes.
Contributor Notes