Development of a multi-locus sequence typing system helps reveal the evolution of Cardinium hertigii, a reproductive manipulator symbiont of insects
Cardinium is an intracellular bacterial symbiont in the phylum Bacteroidetes that is found in many different species of arthropods and some nematodes. This symbiont is known to be able to induce three reproductive manipulation phenotypes, including cytoplasmic incompatibility. Placing individual strains of Cardinium within a larger evolutionary context has been challenging because only two, relatively slowly evolving genes, 16S rRNA gene and Gyrase B, have been used to generate phylogenetic trees, and consequently, the relationship of different strains has been elucidated in only its roughest form.
We developed a Multi Locus Sequence Typing (MLST) system that provides researchers with three new genes in addition to Gyrase B for inferring phylogenies and delineating Cardinium strains. From our Cardinium phylogeny, we confirmed the presence of a new group D, a Cardinium clade that resides in the arachnid order harvestmen (Opiliones). Many Cardinium clades appear to display a high degree of host affinity, while some show evidence of host shifts to phylogenetically distant hosts, likely associated with ecological opportunity. Like the unrelated reproductive manipulator Wolbachia, the Cardinium phylogeny also shows no clear phylogenetic signal associated with particular reproductive manipulations.
The Cardinium phylogeny shows evidence of diversification within particular host lineages, and also of host shifts among trophic levels within parasitoid-host communities. Like Wolbachia, the relatedness of Cardinium strains does not necessarily predict their reproductive phenotypes. Lastly, the genetic tools proposed in this study may help future authors to characterize new strains and add to our understanding of Cardinium evolution.
KeywordsEndosymbiont Wolbachia Cytoplasmic incompatibility Parthenogenesis induction Feminization MLST Phylogenetics
Markov chain Monte Carlo
Multi-locus sequence typing
Polymerase chain reaction
Quantitative polymerase chain reaction
The life histories and evolution of many multicellular organisms are intimately entwined with the microbes they carry . A large number of arthropods carry maternally inherited, intracellular bacterial symbionts that can affect their host’s reproductive outcomes in both detrimental and beneficial ways [2, 3]. These symbionts come from various bacterial phyla, but are categorized based on their associations with their hosts. Primary (or obligate) symbionts complement their hosts’ diet with essential amino acids or other limiting nutrients, are often housed in specialized structures, and are essential to their host’s reproduction (reviewed in Moran et al ). Secondary (or facultative) symbionts, though largely unnecessary for successful host reproduction, can provide conditional benefits to their host, have no measurable effect, or manipulate their host’s reproduction in ways that increase the spread of the symbiont [4, 5, 6].
Symbiont phylogenies may offer clues to the relationship between the symbionts and their hosts. For instance, primary symbionts, such as Buchnera in their aphid hosts, display congruent phylogenies , indicating the long evolutionary history and cospeciation of these groups. Secondary symbionts generally have shorter associations with their hosts and may occur at intermediate frequencies within the host population . The evolutionary phylogenies of secondary symbionts generally display many host switches and are non-congruent with their host’s phylogenies (e.g. ). Genera of bacteria commonly thought of as secondary symbionts may also include lineages of primary symbionts in their midst, as with Serratia symbiotica in aphids [9, 10]. Even the best-known secondary symbiont, Wolbachia, a notorious host switcher, contains a clade of symbionts that display congruent evolution and co-cladogensis in their obligatory symbiosis with nematodes [11, 12] as well as a lineage that is required for B-vitamin production in bedbugs . These patterns show that different strains within one group of secondary symbionts can differ dramatically in their relationships with their hosts.
While transitions from secondary to obligate symbiosis may be apparent in phylogenies, as shown by host and symbiont phylogenetic congruence, subtler facets of secondary symbiont life histories may also be elucidated by a well-resolved phylogeny. Horizontal transmission of secondary symbionts between hosts is key to the secondary symbiont lifestyle, yet these transmission events are rarely captured in experiments (see exceptions in Huigens et al.  and Caspi-Fluger et al. ), and are likely to happen infrequently in nature. Phylogenies are currently the most powerful tools we have to describe these host switches. Well resolved phylogenies may also elucidate co-cladogenesis over a short evolutionary time scale, which can occur when a reproductive manipulator in essence “hijacks” a key reproductive function of their host, creating host-symbiont dependency [6, 16, 17]. In this paper, we explore evolution of the secondary symbiont of arthropods, Cardinium hertigii (Bacteroidetes), and address questions concerning horizontal transmission and the evolution of reproductive manipulations with a well-resolved phylogeny.
Cardinium hertigii, a member of the phylum Bacteroidetes, infects approximately 7–9% of arthropods [18, 19, 20] as well as at least one lineage of the plant parasitic nematode, Heterodera glycines [21, 22]. Although it infects many insects, particularly members of Hymenoptera and Hemiptera, much of the diversity of this symbiont genus as described so far appears to lie in arachnids, such as mites, spiders, and harvestmen as hosts [18, 23, 24]. Although the phenotype of Cardinium in many hosts is unknown, it has been shown to manipulate host reproduction in insects and mites, and rivals Wolbachia in its versatility. Strains of Cardinium induce at least three reproductive manipulations: parthenogenesis, feminization, and cytoplasmic incompatibility (CI).
In symbiont-induced parthenogenesis, genetic males turn into genetic females during embryogenesis. Parthenogenesis has been shown or associated with Cardinium infection in several parasitoid wasps in the genus Encarsia [20, 25] and with the oleander scale, Aspidiotus nerii . In feminization, as has been shown in Brevipalpus mites, Cardinium causes infected genetic males to be converted into functional females . Finally, Cardinium is able to induce cytoplasmic incompatibility in several wasps, mites, planthoppers and a thrips [28, 29, 30, 31, 32, 33, 34, 35], where infected females produce both male and female offspring, but uninfected females mated with infected males produce few or no offspring (in diploid systems) or few or no daughters (in haplodiploid systems). Of all reproductive manipulators, so far only Cardinium, Wolbachia, and a recently discovered Alphaproteobacterium  have been found to induce CI, although genomic evidence of the Cardinium strain cEper1, found in the parasitic wasp Encarsia suzannae, suggest that at least Wolbachia and Cardinium independently evolved this trait . In addition to the reproductive manipulations, Cardinium has been shown to affect other host fitness traits as well. In the planthopper Sogatella furcifera, Cardinium infection is associated with faster nymphal developmental times  and in the parasitoid wasp Encarsia inaron, Cardinium infection is associated with increased longevity of female wasps .
Despite the diverse impacts Cardinium can have on key aspects of its host’s survival and reproduction, few resources have been devoted towards developing better genetic tools for assessing the evolutionary history of this genus, leaving open some intriguing questions about the symbiont’s evolution and ecological interactions with its hosts. Some of the enduring mysteries involving secondary symbionts, and Cardinium in particular, are how these reproductive manipulations evolved. For example, are the genes coding for these manipulations largely horizontally transmitted between strains or do they evolve independently, perhaps repeatedly, within lineages? Additionally, Cardinium horizontal transmission rate at a genus-wide level is poorly understood. With weakly resolved phylogenies, it is not clear whether Cardinium displays the same low level of host affinity as most other secondary symbionts, or whether the shorter list of host taxa with which it is associated than, for example, the cosmopolitan Wolbachia, is indicative of fewer host switches among host lineages. While a total of six Cardinium genomes have now been sequenced [39, 40, 41, 42, 43, 44] genetic resources that enable broad comparisons among many taxa are still needed. We present four sets of primers from single locus housekeeping genes that each amplify 450-700 bp of DNA in order to more fully resolve the evolutionary relationships of the divergent Cardinium strains. By providing primers for the community of Cardinium researchers to use to diagnose Cardinium and discriminate among as yet uncharacterized strains, the study provides a framework for future studies of this versatile symbiont.
Collection localities of Cardinium strains and their associated reproductive phenotypes
CI (Hunter et al., 2003)
San Diego, USA
PI (Zchori-Fein et al., 2004)
PI association (Zchori-Fein et al., 2001)
Encarsia inaron (IT), high density strain (HIT)
One of two strains co-infecting a host with a CI phenotype (Gebiola et al., 2016). This strain does not cause CI (Stouthamer, et al. unpubl.)
Encarsia inaron, (IT), low density strain (LIT)
One of two strains co-infecting a host with a CI phenotype (Gebiola et al., 2016). This strain causes CI (Stouthamer et al. unpubl.)
Encarsia inaron (USA)
No CI, no PI (White et al., 2009)
University of California, Riverside culture
Associated with parthenogenetic host (Provencher et al., 2005)
Bemisia tabaci, Q1 species
No CI, no PI (Fang et al., 2014)
Ixodes scapularis cell line
Nantucket Island (Massachusetts), USA
CI (Nakamura et al., 2009)
CI (Gotoh et al., 2007)
CI (Gotoh et al., 2003)
No CI, no PI (Gotoh et al., 2003)
Kagoshima Pref. or Okinawa Pref., Japan
Kagoshima Pref, Japan
Yonaguni Isl., Okinawa Pref. Japan
Leahurst Campus, University of Liverpool, UK
Vancouver Island, Canada
Vancouver Island, Canada
Northern California, USA
Northern California, USA
North central California, USA
British Columbia, Canada
North central California, USA
North central California, USA
Washington and Oregon, USA
CI (Roush and Hoy, 1981)
Leiobunum sp 1
Georgetown Island, Maine, USA
Leiobunum sp 2
N. Monmouth, Maine, USA
Ellison Park, Monroe County, New York, USA
Minas Gerais, Brazil
Feminization (Groot and Breeuwer, 2006)
Minas Gerais, Brazil
Feminization (Groot and Breeuwer, 2006)
Tijuana River Valley Park, San Diego, USA
Associated with parthenogenetic host (Stouthamer and Luck, 1991)
CI (Nguyen et al., 2017)
This study aimed to better understand the evolution of the diverse arthropod symbiont Cardinium, and provide genetic tools to better identify individual strains within this group. Phylogenies based on sequences derived from four loci across a representative set of Cardinium strains show a greater resolution of Cardinium clades in this diverse genus than single gene trees using more slowly evolving DNA such as the 16S rRNA gene.
Direct sequencing and analysis of the genes selected for the MLST and phylogenetic analyses suggested that they were almost always single copy genes, although there appeared to be two copies of SufB in one host species, the planthopper Sogatella furcifera. The single gene trees were not entirely congruent with each other (Figs. 3, 4, 5, and 6), as is most common in bacterial multi-locus sequence analyses , and underscores the value of combining data from multiple genes. There are several potential reasons for non-congruence of gene trees and lineage trees . They include lateral gene transfer, which is common in bacterial endosymbionts. However, none of the MLST genes are among the 68 Cardinium genes that showed evidence of horizontal transfer in the Cardinium genome cEper1 . Lineage sorting, where polymorphisms in a gene precede the lineage split is another possible cause of non-congruence, although lineage sorting is more likely when population sizes are large , and endosymbionts typically have small effective population sizes . Lastly, gene duplication and extinction is another process that can lead to non-congruence of gene trees with lineage trees, since different loci may be represented in different terminal taxa, even in single copy genes .
Using 16S rDNA and gyrase B, Nakamura et al.  grouped Cardinium into three groups: A, which contains Cardinium strains infecting insects, mites, and other arthropods, B, which contains the Cardinium strain infecting the plant parasitic nematode, Heterodera glycines; and C, which contains Cardinium infecting biting midges in the genus Culicoides. These groups are supported in the current study using the concatenated sequence of four loci. Chang et al.  suggested that the Cardinium found in in the harvestmen clade (Leiobunum spp., Opiliones) might be an independent group, based on a phylogeny constructed using a partial 16S rRNA sequence. However, because the 16S rRNA gene displays a relatively slow rate of evolution, the phylogeny provided limited support for this idea. The current phylogeny using the concatenated loci provides more robust support for a separate clade of Cardinium found in the Leiobunum Opiliones. Following the convention of Nakamura et al.  this clade is designated group E, with clade D reserved for Cardinium in Daphnia, water fleas .
The monophyly of Cardinium in the oleander scale, Aspidiotus nerii (Diaspididae), Encarsia parasitic wasps, and whiteflies is supported. Species of Encarsia that harbor these Cardinium parasitize either whiteflies (E. hispida, E. suzannae, E. tabacivora, E. inaron (IT and US)) or armored scale insects in the Diaspididae (E. perniciosi). The placement of these Encarsia Cardinium strains with those from scale insects (A. nerii) and whiteflies (A. floccosus, B. tabaci) suggests that horizontal transmission events between host and parasitoids, and perhaps among parasitoids, have occurred, although the directionality of these events can only be discerned with a deeper sampling of both hosts and parasitoids. Additionally, none of the closely related strains of Cardinium residing in whiteflies, and two related species of Encarsia (cBtQ1, cEina1, cEina2) are known not to cause CI or any other reproductive manipulations [48, 49]. This suggests that the ancestral strain of Cardinium in this group either spread with a phenotype other than some type of reproductive manipulation, or lost its ability to manipulate in each new host after it spread. The Cardinium group in this clade is the clearest example of closely related Cardinium strains residing in distantly related hosts, in contrast to the previously observed pattern of closely related Cardinium strains residing in closely related hosts [24, 50, 51, 52], a pattern which is generally supported in the phylogenies of the current study as well.
Interestingly, some patterns appear at least superficially similar between Cardinium and Wolbachia. The reproductive manipulations that Cardinium is able to induce overlap with Wolbachia. Strains that cause the same reproductive manipulations do not clearly form one monophyletic clade, except perhaps in the case of the mite strains causing feminization, but this might change when further examples of feminizing Cardinium are discovered. Additionally, closely related Cardinium strains do not necessarily cause the same reproductive manipulations, as exemplified by the sister strains cEper1, which causes CI, and the parthenogenesis-inducing (PI) strain, cEper2 [25, 30]. Similarly, cEsug1, which causes CI, and cTpue1, which does not cause CI or PI, are sister taxa . This pattern also occurs in Wolbachia; closely related Wolbachia strains in Acraea butterflies have shown multiple transitions between sex ratio distorting and CI-inducing Wolbachia strains . Additionally, in Drosophila, wMel, causing CI, and wAu, having no phenotype, are also very closely related . These similar patterns between Wolbachia and Cardinium trees are not necessarily expected; recently, it has been suggested that the horizontal transfer of the CI phenotype may be linked to the Wolbachia’s WO phage, which can cross-infect Wolbachia strains [55, 56]. So far, sequenced genomes of Cardinium do not show the presence of phage DNA. Unlike Wolbachia, however, many Cardinium strains do harbor plasmids [37, 40], which may serve a similar function in horizontal transmission of reproductive manipulation genes [57, 58].
We fully expect the phylogeny of Cardinium to become better resolved when more Cardinium genomes are published, as have Wolbachia genomes . There are currently six Cardinium genomes published [37, 40, 41, 42, 43, 44], three of them published in the last couple of years, so it is reasonable to expect more in the near future. Indeed, the high genetic diversity within this genus made the design of a single MLST, a scheme designed for strains within a bacterial species , challenging. However, while full genomes are always going to be better for inferring phylogenies and group placements , the cost of sequencing, the sequencing depth necessary for symbionts that may exist at relatively low titer in their hosts, and the expertise in assembling symbiont genomes from metagenome data can still be a limiting factor for many laboratories. For ecological studies and surveys in particular, the ability to relatively quickly type Cardinium strains meets the objective of giving the strain an identity and fitting it into the Cardinium phylogeny. This MLST offers a relatively low-cost way to differentiate between strains of Cardinium and is a starting point for researchers considering the study of Cardinium. In particular, the utility of the Cardinium MLST will be valuable in answering questions concerning relatively recent biogeographic or host switching events.
Characterizing a strain of a symbiont with an MLST allelic profile may be difficult when more than one strain co-infects individual host individuals. If both alleles are amplified, direct sequencing may not be possible, and cloning may be required prior to sequencing. Even more challenging is determining which sequenced allele at a particular locus belongs to which strain. There are a couple of potential solutions to this problem. When multiple strains are present in different combinations among individuals, one can logically examine the sets of alleles in multiply infected and singly infected individuals to allow assignment of allelic profiles to strains, a system known as Allelic Intersection Analysis . This may be particularly relevant in complex situations like that found in the apple maggot, Rhagoletis pomonella, where up to four Wolbachia strains have been found in multiple combinations . Another tool that could be useful when co-infecting symbiont strains are found at different titers is quantitative PCR. It may be possible to design specific qPCR primers for each allele and quantify the relative titer of each. If the titer is consistently higher in one set of alleles than the other, one can presume the alleles in that set belong to the same strain. In the host Encarsia inaron (from Italy) coinfecting strains cEina2 and cEina3 are found at high and low titers, respectively (Table 1).
Cardinium evolution appears to be driven by both ecological opportunity and host specialization. Cardinium has frequently switched between parasitoids and their hosts, even though they are physiologically quite different, causing these strains to form a clade. In contrast, the Cardinium in Cybaeus spiders, Culicoides spp., and Leobinium spp. appear to be quite specialized to particular host lineages, without distantly related hosts breaking up these clades. Similar to Wolbachia, the relatedness of Cardinium strains does not necessarily predict their reproductive phenotypes. Overall, the new genetic tools proposed in this study allow for clearer strain delimitation and a more detailed picture of the evolution of Cardinium, one that will keep unfolding the more the MLST primers are used to characterize strains and add taxa to the Cardinium phylogeny.
Four genes with the highest amino acid identity between the sister group to Cardinium, Amoebophilus asiaticus, and the sequenced Cardinium strain, cEper1, were chosen to develop a Multi Locus Sequence Typing (MLST) system with other strains . We did not attempt to choose genes that are evenly spaced around the Cardinium chromosome. While, in more conserved lineages, linkage among loci is often avoided by choosing MLST genes that are evenly spaced , in Cardinium there is little shared synteny, even between the two related sequenced genomes, cBtQ1 and cEper1 . In addition to making even spacing of chosen genes unworkable across the genus, the low level of synteny suggests frequent gene rearrangements in this lineage, and a low probability of linkage among loci. The genes selected for this study were: Elongation Factor G, a protein responsible for coordinating the movement of tRNA and mRNA during translation ; gyrase B, a topoisomerase that unwinds DNA during DNA replication ; Iron Sulfur Cluster Assembly Protein (SufB), a protein involved in generating Fe-S complexes mainly involved in electron transfer  and the Heat shock protein GroEL, a chaperone protein essential in stress-related responses .
MLST primers and their suggested melting temperatures for PCR
Primer sequence (5′ – 3′)
Gene length (bp)
Amplified nucleotide range of gene (bp)
MLST fragment size (bp)
Primer design, PCR, and sequencing
Primers were iteratively designed as sequenced products from strains were added to sequence alignments. Initially, general primers were designed based on the only two sequenced (and closely related) Cardinium strains (cEper1, cBtQ1) and the sister taxon to Cardinium, Amoebophilus asiaticus 5a2. These initial primers were designed using cEper1 as the reference strand in Primer3 [68, 69] with ambiguities based on the other strains added manually. Amplification of some gene products was not successful from all strains using these initial primers, particularly from strains divergent with respect to cEper1 and cBtQ1, such as those in the biting midges, Culicoides spp. In these instances, strain-specific primers were designed once a small segment of the gene was sequenced. These strain-specific primers were then used in conjunction with the initial degenerate primers to obtain more sequence. When more than three bacterial strains were used for primer design, areas of conservation were manually detected and these potential primer regions were checked for hairpins and tendency to form primer dimers in Primer3  against every strain. All primers were selected by minimizing the number of ambiguities and maximizing the number of conserved base pairs in the 3′ primer region, and M13 tags were added to the primers for ease of sequencing .
Although the melting temperature varied depending on the primer pair (Table 1), PCR conditions were generally as follows: 15 μl reaction volume with New England Biolabs buffer and Taq at 1X concentration, 5 mM dNTPs, 0.76 mM MgCl2, 1.1 μM primers with 2 μl of DNA. From mite extractions, 4 μl of DNA was added (similar to Groot and Breeuwer (2006)). The initial melting temperature was 94 °C for 2 min; this was followed by 40 cycles of 94 °C for 45 s, the annealing temperature (Table 1) for 45 s, and extension at 68 °C for 45 s. The final extension was at 68 °C for 7 min.
DNA sequences were quality-controlled and aligned using CLC Main Workbench 6 (Qiagen) and MUSCLE . jModelTest was used to select the optimum model of evolution based on the Akaike information criterion . Bayesian trees were constructed in MrBayes with one million Markov Chain Monte Carlo (MCMC) generations and sampled every 1000 generations . Maximum likelihood trees were constructed using RaxML with 1000 rapid bootstraps. Both Bayesian and ML methods used the GTR + I + G model of nucleotide evolution with a total of 2145 bp from Gyrase B (gyrB), translation elongation factor G (EF-G), Iron Sulfur cluster assembly protein (sufB), and heat shock protein (groEL) for each taxon, partitioned by gene and codon position. Phylogenetic tree figures were generated in Mesquite .
We are very grateful to a number of people who sent us many Cardinium-infected specimens: Tim Kurtii, Marjorie Hoy, Johannes Breeuwer, Hiroaki Noda, Steve Perlman, Jack Werren, Yuval Gottlieb, Nir Netanel, and Markus Riegler. We are also grateful to Wendy Moore, who provided comments on an earlier version of the MS and consulted on the phylogenetic analysis, as did James Robertson. Garrett Hughes also consulted on tree building.
CMS analyzed sequences, designed primers, performed the experiments and phylogenetic analyses and wrote the first draft of the manuscript. SEK assisted with PCR, helped develop and test methods, and revised drafts of the manuscript. EM consulted on experimental methods, made figures and revised drafts of the manuscript. SS-E consulted on experimental methods, served as a mentor in the design and performance of the research and revised drafts of the manuscript. MSH worked with CMS on the design and execution of the project and worked with CMS on the writing and revision of all drafts of the manuscript. All authors read and approved the final manuscript.
Funding for this study was provided by an NSF grant (IOS-1256905) to MSH and SSE, and by the center for Insect Science at the University of Arizona to CMS. Neither funding body had any role in the design of the study or collection, analysis, and interpretation of data or in writing the manuscript.
Ethics approval and consent to participate
Consent for publication
The authors declare that they have no competing interests.
- 3.O'Neill SL, Hoffmann AA, Werren JH, editors. Influential passengers. New York: Oxford University Press; 1997.Google Scholar
- 24.Nakamura Y, Kawai S, Yukuhiro F, Ito S, Gotoh T, Kisimoto R, et al. Prevalence of Cardinium bacteria in planthoppers and spider mites and taxonomic revision of "Candidatus Cardinium hertigii" based on detection of a new Cardinium group from biting midges. Appl Environ Microbiol. 2009;75(21):6757–63.PubMedPubMedCentralCrossRefGoogle Scholar
- 41.Showmaker KC, Walden KKO, Fields CJ, Lambert KN, Hudson ME. Genome sequence of the soybean cyst nematode (Heterodera glycines) endosymbiont "Candidatus Cardinium hertigii" Strain cHgTN10. Microbiol Resour Announcements. 2018;6(26).Google Scholar
- 42.Siozios S, Pilgrim J, Darby AC, Baylis M, Hurst GDD. The draft genome of strain cCpun from biting midges confirms insect Cardinium are not a monophyletic group and reveals a novel gene family expansion in a symbiont. PeerJ. 2019;7.Google Scholar
- 43.Brown AMV, Wasala SK, Howe DK, Peetz AB, Zasada IA, Denver DR. Comparative genomics of Wolbachia-Cardinium dual endosymbiosis in a plant-parasitic nematode. Front Microbiol. 2018;9.Google Scholar
- 44.Zeng Z, Fu YT, Guo DY, Wu YX, Ajayi OE, Wu QF. Bacterial endosymbiont Cardinium cSfur genome sequence provides insights for understanding the symbiotic relationship in Sogatella furcifera host. BMC Genomics. 2018;19.Google Scholar
- 55.Bordenstein SR, Bordenstein SR. Eukaryotic association module in phage WO genomes from Wolbachia. Nat Comm. 2016;7.Google Scholar
- 59.Gerth M, Gansauge MT, Weigert A, Bleidorn C. Phylogenomic analyses uncover origin and spread of the Wolbachia pandemic. Nature Comm. 2014;5.Google Scholar
- 61.Bleidorn C, Gerth M. A critical re-evaluation of multilocus sequence typing (MLST) efforts in Wolbachia. FEMS Microbiol Ecol. 2018;94(1).Google Scholar
- 74.Maddison W, Maddison D. Mesquite: A modular system for evolutionary analysis. Version 3.51 2018 [Available from: http://www.mesquiteproject.org].
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.