Manuscript Click here to access/download;Manuscript;kauai_endangereds_revisions_nofi Click here to view linked References 2 Genetic structure and population history in two critically endangered Kaua‘i honeycreepers 3 4 5 Loren Cassin-Sackett1,2, Michael G. Campana1, Nancy Rotzel McInerney1, Haw Chuan Lim1,3, Natalia 6 A.S. Przelomska1,4,5, Bryce Masuda6, R. Terry Chesser7,8, Eben H. Paxton9, Jeffrey T. Foster10, Lisa H. 7 Crampton11, Robert C. Fleischer1 8 9 1 – Center for Conservation Genomics, Smithsonian Conservation Biology Institute, Washington DC, 10 USA 11 2 – Current affiliation: Department of Biology, University of Louisiana, Lafayette LA, USA 12 3 – Department of Biology, George Mason University, Fairfax VA, USA 13 4 – Department of Anthropology, National Museum of Natural History, Washington, DC, USA 14 5 – Comparative Plant and Fungal Biology, Royal Botanic Gardens, Kew, Richmond, United Kingdom 15 6 – Hawai‘i Endangered Bird Conservation Program, Recovery Ecology, Institute for Conservation 16 Research, San Diego Zoo Global, Volcano, HI, USA 17 7 – U.S. Geological Survey, Patuxent Wildlife Research Center, Laurel, MD, USA 18 8 – Department of Vertebrate Zoology, National Museum of Natural History, Washington, DC, USA 19 9 – U.S. Geological Survey Pacific Island Ecosystems Research Center, Hawai‘i National Park, HI, USA 20 10 – Pathogen and Microbiome Institute, Northern Arizona University, Flagstaff, AZ, USA 21 11 – Kaua‘i Forest Bird Recovery Project, Pacific Cooperative Studies Unit, Hanapepe, HI, USA 22 23 Corresponding Author: Loren Cassin-Sackett, Cassin.Sackett@gmail.com 24 25 ORCIDs: 26 Loren Cassin-Sackett: 0000-0002-6000-4789 27 Michael G. Campana: 0000-0003-0461-6462 28 Nancy Rotzel McInerney: 0000-0002-6519-7671 29 Haw Chuan Lim: 0000-0002-6420-1667 30 Natalia A.S. Przelomska: 0000-0001-9207-4565 31 R. Terry Chesser: 0000-0003-4389-7092 32 Eben H. Paxton: 0000-0001-5578-7689 33 Jeffrey T. Foster: 0000-0001-8235-8564 34 Lisa H. Crampton: 0000-0002-5420-4338 35 Robert C. Fleischer: 0000-0002-2792-7055 1 36 37 38 Keywords: bottleneck, drepanids, population structure, islands, Hawaii, conservation breeding 39 40 Acknowledgements 41 Blood samples from individuals in the conservation breeding population were collected following 42 San Diego Zoo Global IACUC 18-022. We are thankful to Carter Atkinson, U.S. Geological Survey, 43 Pacific Island Ecosystems Research Center, for sharing blood samples collected in the Alaka‘i during the 44 1990s and to Katy Parise for DNA extractions. We are grateful to the Kaua‘i Forest Bird Recovery 45 Project field crews and the San Diego Zoo Global crews for their tremendous dedication to catching all 46 the wild birds and to finding, collecting, and hatching the eggs. The establishment of the conservation 47 breeding population was funded by the U.S. Fish and Wildlife Service, State of Hawai‘i, San Diego Zoo 48 Global, and the Mohamed Bin Zayed Species Conservation Fund. Funding to collect samples from wild 49 individuals and for genetic analysis was in part provided by the U.S. Fish and Wildlife Service, the State 50 of Hawai‘i, and Friends of the National Zoo’s Small Grant program. We appreciate comments on the 51 manuscript from the @CrawLab and Cadena Lab at Universidad de los Andes. Any use of trade, product, 52 or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government. 53 54 55 2 56 Abstract 57 Population sizes of endemic songbirds on Kaua‘i have decreased by an order of magnitude over 58 the past 10–15 years to dangerously low numbers. The primary cause appears to be the ascent of invasive 59 mosquitoes and Plasmodium relictum, the agent of avian malaria, into elevations formerly free of 60 introduced malarial parasites and their vectors. Given that these declines in native bird populations appear 61 to be continuing, last resort measures to save these species from extinction, such as conservation 62 breeding, are being implemented. Using 200–1439 SNPs from across the genome, we assessed kinship 63 among individuals, levels of genetic variation, and extent of population decline in wild birds of the two 64 most critically endangered Kaua‘i endemic species, the ‘akikiki (Oreomystis bairdi) and ‘akeke‘e (Loxops 65 caeruleirostris). We found relatively high genomic diversity within individuals and little evidence of 66 spatial population genetic structure. Populations displayed genomic signatures of declining population 67 size, but individual inbreeding coefficients were universally negative, likely indicating inbreeding 68 avoidance. Diversity within the founding conservation breeding population largely mirrored that in the 69 wild, indicating that genetic variation in the conservation breeding population is representative of the wild 70 population and suggesting that the current breeding program captures existing variation. Thus, although 71 existing genetic diversity is likely lower than in historical populations, contemporary variation has been 72 retained through high gene flow and inbreeding avoidance. Nonetheless, current effective population size 73 for both species was estimated at fewer than 20 individuals, highlighting the urgency of management 74 actions to protect these species. 75 76 Declarations 77 Funding: U.S. Fish and Wildlife Service, State of Hawai‘i, San Diego Zoo Global, Mohamed Bin Zayed 78 Species Conservation Fund, Friends of the National Zoo 79 Conflicts of Interest: The authors declare that they have no conflict of interest. 80 Ethics approval: Blood sample collection protocols from individuals in the conservation breeding 81 population were approved by San Diego Zoo Global IACUC 18-022. 82 Consent to participate: All authors participated in the project and have seen this manuscript. 83 Consent for publication: All authors consent to publication of this manuscript. 84 Availability of data and material: Sequence reads have been published on GenBank as individual fastq 85 files for each individual under BioProject PRJNA527134. 86 Code availability: All software is publicly and freely available, and all custom scripts are published on 87 GitHub. 88 Author contributions: This study was conceived and designed by LCS, MGC, BM, RTC, LHC and RCF. 89 Material preparation, data collection, and data analysis were performed by LCS, MGC, NRM, HCL, 3 90 NASP, BM, RTC, EHP, JTF, LHC and RCF. The first draft of the manuscript was written by LCS with 91 MGC and all authors commented on previous versions of the manuscript. All authors read and approved 92 the final manuscript. 93 94 Introduction 95 The contemporary rate of biodiversity loss, with extinction rates 1000 times higher than 96 background rates (Pimm et al. 2014), has been unparalleled since the mass extinction of non-avian 97 dinosaurs (Jablonski 1986). Species declines and extinctions may occur due to external forces such as 98 environmental catastrophes or emerging infectious diseases (Lande 1993; Spiller et al. 1998; Courchamp 99 et al. 2006; Prowse et al. 2013), demographic stochasticity (Shaffer 1983; Lande 1988; Wootton & Pfister 100 2013; Mashayekhi et al. 2014), genetic processes such as adaptive diversity loss or fixation of deleterious 101 mutations (Lande 1994, 1998; Palkopoulou et al. 2015; Rogers & Slatkin 2017), or the interaction of 102 these forces (Robert 2011). External forces are often the most straightforward to document, but it is also 103 important to understand how these processes influence demographics and genetics of declining species. 104 With changes in climate and other anthropogenic effects, emerging infectious diseases are a 105 primary driver of global biodiversity loss (La Marca et al. 2005; Lips et al. 2006; Smith et al. 2006). 106 Because climatic variables can influence pathogen vector abundance and in turn pathogen distribution 107 (Padilla et al. 2017), climate is likely to influence population dynamics of host species susceptible to 108 infectious diseases (Samuel et al. 2015). The interaction between infectious diseases and climate is 109 complex (Paull et al. 2012, Mordecai et al. 2017), and their combined influence on genetic variation of 110 declining species is less well understood. Especially vulnerable are island species, which are threatened 111 by introduced predators and pathogens, and often require specialized habitats or specific climatic regimes 112 (Fortini et al. 2015; Glad and Crampton 2015; Harter et al. 2015; Liao et al. 2015). 113 Hawaiian honeycreepers (Passeriformes: Fringillidae: Carduelinae), which are endemic to the 114 Hawaiian Islands, have experienced population declines and extinctions since humans arrived on the 115 islands in approximately 1000 C.E. (Kirch 2011). These declines accelerated in the late 19th century, 116 likely due to introduced avian pox virus and introduced predators, and in the 20th century honeycreeper 117 populations began to crash (Foster et al. 2004; Camp et al. 2009; Gorresen et al. 2009). Plasmodium 118 relictum, the causative agent of avian malaria that was introduced by the 1940s (Fisher & Baldwin 1947) 119 and the previously introduced mosquito vectors (Culex quinquefasciatus) were historically restricted by 120 temperature to low elevations. As a result of climate warming (Diaz et al. 2011), mosquitoes have 121 expanded their elevational range so that bird populations are within the range of malaria-infected 122 mosquitoes for most of the year (Atkinson et al. 2014). This expansion has contributed to the extinction of 123 several avian species and pushed most remaining honeycreeper species to the brink of extinction (Paxton 4 124 et al. 2016, 2018). On Kaua‘i in particular, avian species no longer have high-elevation refuge from 125 malaria (Atkinson et al. 2014). The past decade has witnessed the near complete collapse of the native 126 bird community on Kaua‘i, and several endemic species have likely gone extinct (Paxton et al. 2016). 127 Particularly alarming are the population declines and range contractions of two Kaua‘i endemic species: 128 the ‘akeke‘e (Loxops caeruleirostris, 98% decline from 2000–2012) and the ‘akikiki (Oreomystis bairdi, 129 71% decline from 1981–2012; Paxton et al., 2016). These population crashes directly correspond to the 130 ascent of introduced malaria and its invasive mosquito vector Cx. quinquefasciatus to even the highest 131 elevations on the island (Atkinson et al. 2014), in part due to the influence of warming temperatures on 132 disease dynamics (Samuel et al. 2011), and potentially to the replacement of a warm-adapted mosquito 133 lineage by a cold-adapted lineage (Fonseca et al. 2006). 134 Both the ‘akeke‘e and the ‘akikiki are listed as Critically Endangered by the International Unio 135 for the Conservation of Nature (IUCN; BirdLife International 2018a, 2018b), Endangered by the U.S. 136 Fish and Wildlife Service (USFWS 2010), and of Greatest Conservation Need by the State of Hawai‘i 137 (Hawai‘i Division of Forestry and Wildlife 2015). ‘Akikiki are estimated to number ~450 individuals and 138 occupy ~25km2 of habitat; ‘akeke‘e are estimated to number ~1160 and occupy ~40km2 (Paxton et al., 139 2020). Both species are currently restricted to the remote interior high elevation ‘ōhi’a (Metrosideros 140 polymorpha, Myrtaceae) forests in the Nā Pali Forest Reserve and Alaka’i Wilderness Preserve (Figure 141 1), which has long been a refuge from mosquito-borne avian diseases because temperatures were 142 historically too low for mosquito and parasite development. However, mean temperatures on Kaua‘i have 143 risen in the last several decades (Fortini et al. 2015, Atkinson et al. 2014), allowing the incursion of 144 mosquitoes and malaria into this former refuge (Atkinson et al. 2014) and threatening the survival of 145 these avian species in the absence of intervention. 146 Given the catastrophic declines documented in ‘akeke‘e and ‘akikiki populations and the lack of 147 means to control mosquitoes across the landscape, two key conservation strategies are gaining knowledge 148 about the distribution and degree of genetic variation in the species and establishing conservation 149 breeding populations for each species. As a critical element of the ‘akeke‘e and ‘akikiki conservation 150 management programs, egg collections were initiated in 2015 by the state and San Diego Zoo 151 International to establish a conservation breeding (also known as captive propagation, captive breeding, 152 ex situ management, or managed care) population for each species. The ultimate goal of conservation 153 breeding programs is to ensure species survival (Rodrigues 2006, Farhadinia et al. 2020), and in several 154 well-known species these programs likely have been the primary or only factor preventing extinction 155 (e.g., California condor (Gymnogyps californianus), black-footed ferrets (Mustela nigripes), ‘alalā 156 (Corvus hawaiiensis), whooping cranes (Grus americana), Butchard et al. 2006, Santymire et al. 2014). 157 The viability, productivity, and success of a conservation breeding population depends largely on the 5 158 genetic diversity of the founding individuals and how well it represents the neutral and adaptive genetic 159 variation contained in wild populations. Maximizing the degree of genetic diversity and the extent of 160 outbreeding in a conservation breeding population minimizes the risk of inbreeding depression that occurs 161 due to the unmasking of deleterious recessive alleles and reduces the risk of mortality from the expression 162 of lethal equivalents (Ralls et al. 1988; Roelke et al. 1993). Moreover, species’ persistence in the wild is 163 likely affected by the degree of genetic diversity contained in wild populations (Palkopoulou et al. 2015). 164 Therefore, genomic methods provide a powerful means to assess the patterns of neutral and adaptive 165 diversity within and among individuals (Cassin-Sackett et al. 2019b). Here, in the first study to 166 characterize the genetics of ‘akeke‘e and ‘akikiki, we assess the kinship and genetic diversity of wild 167 individuals and those that were used to initiate the conservation breeding population for each species. Our 168 goals were to (1) characterize the degree and distribution of genetic variation in wild populations of each 169 species, (2) evaluate whether their genomes show evidence of recent declines, and (3) determine the 170 genetic characteristics and inbreeding levels of the initial conservation breeding population of each 171 species, including whether these populations adequately represent the genomic variation currently present 172 in the wild. 173 174 Materials and Methods 175 Sampling 176 For wild birds, we sampled from five sites where ‘akeke‘e and ‘akikiki occur in the higher 177 elevations of Kaua‘i (Figure 1). Field sites were as follows: Pihea (PIH), Kawaikōī (KWK), Upper 178 Kawaikōī (UUK), Mohihi (MOH), and Halepa‘akai (HPK). Mist nets were set intermittently from 2012– 179 2018 in the canopy, both actively (i.e., with audio playback targeting each species separately) and 180 passively (without playback). In addition, some samples from wild birds were obtained from fieldwork in 181 the region dating back to the mid-1990s. Wild birds were banded and blood was collected as described 182 elsewhere (Atkinson et al. 2014); wild samples included 32 ‘akeke‘e and 52 ‘akikiki. For the conservation 183 breeding populations, eggs were collected from nests from 2015–2018 in UUK, MOH and HPK, the 184 locations with the highest density of birds. Behavioral clues were used to find nests, and eggs were 185 collected 10–15 days after the clutch was completed. To reduce the risk of inbreeding in the conservation 186 breeding population, color band patterns and plumage differences were used to avoid collecting from the 187 same pair more than once. All eggs (1–4 per nest) from sampled nests, which were accessed by ladder, 188 were collected to encourage the pair to lay an additional clutch. Eggs were transported to a conservation 189 breeding facility for subsequent artificial incubation and hand rearing aviculture. All individuals in the 190 conservation breeding population for this study were collected from the wild (i.e., none were F1 offspring 191 of founding individuals). As of late 2018 (when our analyses were conducted), these populations included 6 192 10 ‘akeke‘e and 46 ‘akikiki founding individuals; a subset of these individuals with blood samples with 193 sufficiently high DNA quantity and quality were included in this manuscript. 194 Upon first capture (for wild birds) or after fledging (for birds in managed care; hereafter 195 ‘managed’), birds were fitted with a unique combination of color bands and/or a Federal bird band with a 196 unique identifier. For wild birds, approximately 50 µL of blood was collected via brachial venipuncture in 197 heparinized capillary tubes, while for birds in managed care, approximately 20 µL of blood was collected 198 via jugular venipuncture. Blood was then transferred to Queen’s Lysis buffer to preserve the DNA and 199 shipped on dry ice for storage in the Smithsonian Cryo-Collection at the National Zoo in Washington, 200 D.C. 201 202 DNA Preparation and Sequencing 203 DNA was extracted from blood with a DNeasy Blood & Tissue kit (Qiagen, Hilden, Germany) 204 and sheared using a Qsonica Q800R (Newton, CT); libraries were constructed on sheared DNA using 205 KAPA library preparation kits (Kapa Biosystems, Wilmington, MA). Each sample was dual indexed with 206 unique adapters, and following library prep, low-cycle number PCRs were run in duplicate or triplicate 207 and pooled to minimize carryover of PCR artifacts. Post-PCR libraries were subsequently pooled in 208 groups of eight samples and then hybridized for 24–48 hours to a custom-designed and filtered (Arbor 209 Biosciences, Ann Arbor, MI) set of 40,000 oligonucleotide baits targeting single nucleotide 210 polymorphisms (SNPs) distributed randomly across non-repetitive portions of the genome (Cassin- 211 Sackett et al. 2019a). Baits were designed using the genome of the Hawai‘i ‘amakihi (Chlorodrepanis 212 virens; Callicrate et al. 2014), which diverged from ‘akeke‘e 2.47 million years ago (mya) and from 213 ‘akikiki 4.73 mya (Lerner et al. 2011), corresponding to approximately 1.2 million generations for 214 ‘akeke‘e and 2.3 million generations for ‘akikiki (Hammond et al. 2015). After hybridization, pools were 215 combined and size-selected with a Pippin Prep (Sage Science, Beverly, MA) prior to sequencing. 216 Libraries were sequenced on 150 bp paired-end runs on an Illumina HiSeq at Johns Hopkins University or 217 Brigham Young University. 218 219 SNP filtering and processing 220 Reads were trimmed using Trimmomatic 0.36 (Bolger et al. 2014) with the following parameters: 221 ILLUMINACLIP:NexteraPE-PE.fa:2:30:10, LEADING:3, TRAILING:3, SLIDINGWINDOW:4:20, 222 MINLEN:36. Trimmed reads were subsequently aligned to the ‘amakihi genome (Callicrate et al. 2014) 223 using BWA-MEM 0.7.17 (Li 2013). Reads with MAPQ < 20 were removed from the alignments using 224 SAMtools 1.6. (Li et al. 2009). PCR duplicates were marked using Picard 2.9.4 MarkDuplicates (Broad 225 Institute, http://broadinstitute.github.io/picard). Reads were realigned around indels using the Genome 7 226 Analysis Toolkit (GATK) 3.7.0 IndelRealigner (McKenna et al. 2010). Single nucleotide variants and 227 small indels were called for each species using the GATK HaplotypeCaller; non-variant sites were not 228 included in downstream analysis. Variants within the baited regions were extracted using VCFtools 229 0.1.15 (Danecek et al. 2011), and filtered for depth (DP ≥ 4) and quality in GATK (ReadPosRankSum ≥ - 230 8.0, MQRankSum ≥ -12.5, FS ≤ 60.0, QD ≥ 2.0) and minor allele frequency (maf ≥ 0.01) in VCFtools 231 0.1.15. Removing sites with low minor allele frequency aims to eliminate SNPs generated from 232 sequencing error; the number of remaining SNPs after applying these filters was 16,778 for ‘akeke‘e and 233 22,482 for ‘akikiki. We also filtered for missingness in VCFtools (individuals missing data at >95% of 234 loci were removed; in the remaining dataset we retained loci genotyped in at least 80% or 100% of 235 individuals, depending on whether the analysis allowed for missing data). Due to our interest in specific 236 individuals, particularly in the conservation breeding populations, we prioritized retaining individuals 237 over retaining loci even when samples from those individuals were of poor quality (leading to higher 238 missing data). 239 240 Kinship 241 We removed Z chromosome variants using VCFtools 0.1.15; no baits targeted chromosome W. 242 We included only biallelic SNPs for kinship estimation. Final datasets for kinship analysis consisted of 243 matrices with individuals missing fewer than 95% of loci and loci genotyped in at least 80% of 244 individuals. Bootstrapped kinship analysis (100 bootstrap replicates) using maximum likelihood 245 estimation of identity by descent was performed in SNPRelate 1.14.0 (Zheng et al. 2012) in R 3.5.0 (R 246 Core Team 2018) using kinshipUtils (Campana, M.G.: https://github.com/campanam/kinshipUtils) 247 following Cortes-Rodriguez et al. (2019). We pruned SNPs in linkage disequilibrium using a threshold of 248 0.2 for the absolute value of the correlation coefficient (|r|). 249 250 Population Structure and Diversity 251 Population structure was assessed in several ways. First, we estimated the number of ancestral 252 populations using three replicate runs in ADMIXTURE 1.3 (Alexander et al. 2009), allowing K to vary 253 from 1 to 20. Second, we inferred ancestry coefficients of each individual in Structure 2.3.4 (Pritchard et 254 al. 2000, Hubisz et al. 2009), assuming correlated allele frequencies and allowing admixture. We 255 performed three replicate runs consisting of 250,000 burn-in and 1,000,000 iterations. We allowed K to 256 vary from 1 to 8 and ensured convergence among runs with the same assumed K. Next, we tested for 257 isolation by distance by performing a Mantel test on linearized FST against the log of geographic distance 258 in the vegan package (Dixon 2003) for R. Finally, using a dataset with no missing data, we used the 259 vegan package to perform a Principal Coordinate Analysis (PCoA) with Prevosti’s Distance method. A 8 260 PCoA was preferred over Principal Component Analysis because both datasets were characterized by 261 more SNP sites than individuals (Rohlf 1972). 262 We used VCFtools to estimate several diversity metrics within species: (1) inbreeding 263 coefficients for each individual, (2) mean relatedness of an individual to each other member of the 264 population, and (3) nucleotide diversity (π, Nei and Li, 1979) in non-overlapping 1-million base pair bins 265 within species (these larger than typical bins were used because of the small number of loci). In addition, 266 we performed an AMOVA on each species in the ade4 package (Dray and Dufour 2007) for R, assuming 267 two hierarchical population clusters: the lowest-level cluster separated all sampling locations and the 268 higher cluster grouped HPK separately from the remaining sampling sites, a division consistent with 269 geography (Figure 1). We calculated individual genome-wide observed heterozygosity across sites using 270 the adegenet package (Jombart 2008) for R and expected heterozygosity (He) as measured by gene 271 diversity in Genepop (Rousset 2008). Finally, we calculated pairwise FST values between sampling sites 272 using the Weir and Cockerham method in the adegenet package in R. Due to the observed distribution of 273 sequencing depth in ‘akikiki loci (Figure S2), we repeated diversity analyses after removing loci with a 274 depth greater than the mean plus two standard deviations (see Supplementary Materials). 275 Next, we aimed to assess whether the population in managed care adequately represents genomic 276 variation in the wild. First, we compared the three diversity metrics above between wild and managed 277 populations. Second, we constructed median joining networks to visualize the proportion of the network 278 covered by managed individuals. To do so, we generated a fasta alignment file from each filtered vcf 279 using SNiPlay (Dereeper et al. 2015) and constructed a median joining network (Bandelt et al. 1999) for 280 each species in the pegas v. 0.14 (Paradis 2010) R package. We then plotted the networks to determine 281 whether individuals in managed care were found throughout the network. 282 283 Detection of Bottlenecks 284 Because both species have suffered dramatic population declines in recent years, we used several 285 approaches to test whether we could detect genetic signatures of bottlenecks. First, we examined the 286 degree of heterozygosity excess relative to expectations based on allelic diversity at each site (Cornuet 287 and Luikart 1996). We calculated both He and Ho in the Genepop package for R and conducted a sign test 288 with 95% confidence intervals (Cornuet and Luikart 1996) using the BSDA package (Arnholt and Evans 289 2017) for R. Second, we used VCFtools to calculate Tajima’s D in 5 million base pair bins. Third, we 290 used GADMA (Genetic Algorithm for Demographic Analysis, Noskova et al. 2020) to explore 291 demographic history separately in each species, assuming a single population and allowing for up to four 292 time intervals (i.e., where each interval was allowed a different pattern of population growth). This 293 approach generates a site frequency spectrum using SNPs, and performs simulations using the spectrum 9 294 to infer demographic history beginning at the emergence of the species. Simulations were run using the 295 moments scheme (Jouganous et al. 2017). Candidate models contained either three or four time intervals 296 with linear, exponential or sudden population growth or decline in each interval. Models were run in 297 triplicate and the best models (evaluated with log likelihood and Akaike Information Criterion) were 298 visualized. The timing of the onset of each interval was inferred by the model, but because of the bias 299 inherent in selecting sites for analysis that are known to be variable, we did not attempt to estimate 300 precise timing of any demographic events (e.g., population declines); rather, we were solely interested in 301 patterns of population growth and decline as well as relative timing (e.g., old vs. recent). We repeated the 302 analyses with different values of theta to ensure our results were robust to changes in this parameter. 303 Fourth, we estimated historical trends in effective population size (Ne) over time using SNeP (Barbato et 304 al. 2015), which uses linkage disequilibrium to estimate Ne in the more recent past. Because changes in 305 most run parameters did not appreciably change the numerical estimates of Ne, we used the default 306 settings for all parameters except the minimum distance between SNPs (set to minimum of 1 base pair, 307 thus using all SNPs in the calculations of LD) and the minimum allele frequency for inclusion in the 308 analysis (set to minimum of 0.01). These parameter settings were designed to maximize the number of 309 loci used in the calculations. Finally, because SNeP may underestimate very recent and current Ne 310 (Barbato et al. 2015), we estimated current effective population sizes in both species using the molecular 311 coancestry method implemented in NeEstimator v2.1 (Do et al. 2014) and generated jackknife confidence 312 intervals. To validate these estimates, we repeated the analysis with the dataset containing no missing 313 genotypes. 314 315 Results 316 Samples 317 Final post-filtering datasets included 37 ‘akeke‘e (29 wild, eight managed) and 64 ‘akikiki (36 318 wild, 28 managed) individuals (Supplementary Tables S1–S2). Because our aim was to maximize the 319 number of individuals in the dataset, and several individuals had large proportions of missing data (e.g., 320 six wild and three managed retained ‘akeke‘e, as well as four wild and eight managed retained ‘akikiki, 321 were missing genotypes at >90% of SNPs), datasets for kinship analyses and population structure (80% 322 complete) contained 1021 (‘akeke‘e) and 1439 (‘akikiki) SNPs (Supplementary Tables S3–S5; resulting 323 levels of missing data among individuals shown in Figure S1). The distribution and mean coverage depth 324 of loci was similar before and after filtering was applied (Figure S2). Final datasets for PCoA did not 325 permit missing data and included 218 (‘akeke‘e) and 246 (‘akikiki) SNPs; this dataset was also used for 326 estimating demographic history. Sequences are available on GenBank (BioProject PRJNA527134). 327 10 328 Kinship 329 Kinship values are based on the genetic similarity of individuals, which usually reflects a close 330 kin relationship (identity by descent). Mean empirical kinship of all ‘akeke‘e individuals was 0.088 (range 331 0.003–0.215; Figure 2, top left; Supplementary Tables S6–S7), and was similar in wild (0.095; 95% CI 332 0.092–0.098) and managed (0.075; 95% CI 0.060–0.090) individuals. Mean kinship of all ‘akikiki 333 individuals was 0.065 (range 0.003–0.362; Figure 2, bottom right; Supplementary Tables S8–S9), and 334 was nearly identical in wild (0.066; 95% CI 0.063–0.068) and managed (0.066; 95% CI 0.062–0.069) 335 individuals. 336 We examined eggs removed from the same nest, which we expected to be siblings or half siblings 337 in the case of extra-pair mating (n = nine ‘akikiki pairs, one ‘akeke‘e pair and one ‘akeke‘e trio). In 338 ‘akikiki, these pairings showed expected kinship levels (~0.25 for siblings), but in ‘akeke‘e the values 339 were lower than expected for full or half siblings (i.e., ‘akeke‘e nestmates SB1, SB2 and SB3 all had 340 pairwise kinships below 0.14). These low kinship values were not related to the quality or coverage of the 341 sequence data – instead, they may reflect extra-pair mating, intraspecific brood parasitism, or simply 342 independent assortment (but likely a lower level than we found). The nestmate pairing kinship value was 343 0.141 for the single remaining ‘akeke‘e nest and averaged 0.204 (range 0.178–0.244) for ‘akikiki nests, a 344 range expected for half or full siblings (especially given that independent assortment can cause high 345 variation in sibling kinship estimates relative to parent-offspring values). 346 347 Population Structure and Genetic Diversity 348 For both species, coancestry analysis in ADMIXTURE 1.3 indicated the highest support for a 349 single ancestral population (K) containing all extant individuals (estimated by the lowest cross-validation 350 error), with decreasing support for each increase in number of ancestral populations. Log-likelihood 351 scores also supported a single population. However, for ‘akikiki, cross-validation errors were similar for 352 one, two and three ancestral populations. Structure results also indicated support for a single population in 353 both species. 354 The principal coordinates analysis recovered groupings that were related to sampled locations, 355 with one site, HPK, encompassing nearly all of the variation and most other sampling locations 356 containing genetic diversity that also existed in HPK (Figure 3). A second site, MOH, also contained 357 some unique variation in both species, as did KWK in ‘akeke‘e. Mantel tests detected no isolation by 358 distance in either species (p ≥ 0.5, Figure S3). The AMOVA did not recover significant genetic variance 359 partitioned among sampling locations in ‘akeke‘e (0.33% of the variation, p=0.19). However, in ‘akikiki, 360 there was a small but significant amount of genetic variance partitioned among sampling locations (0.70% 361 of the variation, p=0.04). In both species, the remaining explained variation existed not among individuals 11 362 but within individuals. In other words, individuals did not contain many unique SNPs that were not 363 present in other individuals; instead, individuals were heterozygous at many SNPs. In line with this result, 364 pairwise FST values between sites were low, particularly in akeke’e (Table 1). No private alleles were 365 detected within any sampling location in ‘akeke‘e; the frequency of private alleles in ‘akikiki was 0.17. 366 The mean nucleotide diversity (in 1 million base pair bins) in ‘akeke‘e was 1.61 x 10-6 and in 367 ‘akikiki was 1.07 x 10-6, and was not significantly different between managed and wild individuals 368 (Tables S1, 2). The mean inbreeding coefficient was negative in both species (‘akeke‘e -0.478, ‘akikiki - 369 0.373, Table S1), and all individuals were characterized by negative inbreeding coefficients. Managed 370 individuals had a slightly less-negative (i.e., higher) inbreeding coefficient than wild individuals, 371 particularly in ‘akeke‘e (Table 2). Observed heterozygosity within individuals was high: 0.623 in 372 ‘akeke‘e and 0.543 in ‘akikiki. 373 Median joining networks showed that the populations in managed care largely represent existing 374 genomic variation (Figure 4). Nonetheless, if additional egg collections are deemed necessary, the 375 networks indicate slightly underrepresented regions of the genomic space that could be targeted to 376 augment diversity in the managed population (e.g., left side of the ‘akeke‘e network). When plotting the 377 median joining networks that included samples collected in the 1990s (insets in Figure 4), the managed 378 population of ‘akeke‘e appears to miss a notable proportion of diversity. However, when plotting only the 379 individuals that may still be alive, this pattern disappears. This suggests that extant wild ‘akeke‘e 380 populations are missing some diversity that was present in the 1990s. 381 382 Detection of Bottlenecks 383 Both species (with individuals pooled across sampling locations), exhibited a marked excess of 384 heterozygosity relative to expectations based on allelic diversity. In ‘akeke‘e, 89.5% of loci were 385 characterized by higher heterozygosity than expected (sign test p<2.2e-16; median He – Ho = -0.199), 386 while in ‘akikiki, 73.9% of loci displayed higher heterozygosity than expected (sign test p<2.2e-16; 387 median He – Ho = -0.065). Tajima’s D was strongly positive in both species (mean 1.81 in ‘akeke‘e, mean 388 2.15 in ‘akikiki; Table S1), consistent with population bottlenecks (Nei et al. 1975, Gattepaille et al. 389 2013). GADMA indicated consistent support for exponential population decline in both species (Figure 390 S4), with similar patterns in the models for each replicate run. Within replicates, visualizations indicated 391 ostensibly identical patterns of decline among all well-supported models, differing only in the relative 392 timing of the onset of population size reduction. Replicate runs resulted in 1–15 statistically 393 indistinguishable ( AIC < 2) models; when there were multiple indistinguishable models, at least three 394 were visualized to ensure inferences were consistent. In all ‘akeke‘e models, the declines occurred over 395 relatively long time intervals (estimated not in actual timing but duration of the existence of the species, 12 396 on the order of the last 20–25% of the species’ existence; Figure S4). ‘Akikiki population decline 397 occurred more rapidly and more recently (on the order of the last 4–9% of the species’ existence; Figure 398 S4). Effective population sizes estimated in SNeP historically numbered in the tens of thousands, but 399 exhibited trends of linear decline during the past several hundred generations in both species (Figures 5, 400 S4). Approximately 150 generations ago (generation time ~2 years), effective population sizes numbered 401 in the thousands; in the last five generations, Ne fell from 79.5 to 10 in ‘akeke‘e and from 158 to 15.3 in 402 ‘akikiki (Figure S5). Because the number of bins used can influence the magnitude of inferred Ne in very 403 recent time periods (Barbato et al. 2015), this number should be interpreted with caution. Estimates of 404 current effective population size from NeEstimator were similarly small: for datasets allowing up to 20% 405 missing data, Ne was estimated to be 18.5 (95% CI 15.4–21.8) in ‘akeke‘e and 13.4 (95% CI 11.4–15.5) 406 in ‘akikiki. Estimates were slightly higher with larger confidence intervals for datasets with no missing 407 data (‘akeke‘e: 25.3, CI 13.0–41.4; ‘akikiki: 16.5, CI 11.0–23.1). 408 409 Discussion 410 Using multiple approaches, we explored the distribution of genetic variation in wild and founding 411 conservation breeding populations of two endangered Hawaiian honeycreepers. We detected high 412 heterozygosity, little to no spatial structure among sampled locations, and genetic signatures of severe 413 population declines in both species. We also found that individuals were not inbred, with a high 414 proportion of variation contained within individuals and universally negative inbreeding coefficients. The 415 high levels of observed relative to expected heterozygosity in both species (e.g., 5–10x higher than in 416 Hawai‘i ‘amakihi; Cassin-Sackett et al. 2019a) are likely indicative of both recent severe population 417 bottlenecks, which appear coincident with human settlement of the island, and of linkage disequilibrium 418 between variable sites (as LD increases after bottlenecks). If disassortative mating occurs based on a few 419 loci (e.g., the major histocompatibility complex, Juola and Dearborn 2012) in high LD with other loci, 420 then high heterozygosity across the genome can persist for multiple generations after bottlenecks. Both 421 species also showed evidence of long-term population declines (potentially due to climatic fluctuations or 422 changes in island size; Figures 4, S5). Kinship between nestmates was generally in line with predictions, 423 although a few hypothesized ‘akeke‘e siblings demonstrated lower relatedness than expected, suggesting 424 potential intraspecific brood parasitism or multiple paternity. These scenarios support recent observations 425 of multiple adults attending nests (L.H. Crampton, written communication, 2020). Finally, for both 426 species the genetic diversity of founding individuals for the conservation breeding population is largely 427 representative of what remains in nature. 428 The lack of genetic structure among sampled sites, along with high heterozygosity and negative 429 inbreeding coefficients, suggests that wild ‘akeke‘e and ‘akikiki move relatively freely among sampling 13 430 locations and either avoid inbreeding, experience selection against inbred individuals, or both (Keller et 431 al. 1994, Hemmings et al. 2012). As a result of the species’ apparent inbreeding avoidance and movement 432 among sites, the egg collections to establish the conservation breeding populations appear to encompass 433 existing genetic diversity (Sutton 2014) well for ‘akikiki and reasonably well for ‘akeke‘e, even without 434 sampling all sites harboring individuals of these species. This conclusion is supported by the similarity in 435 diversity measures between wild and managed individuals in both species (Table 2). Representation of 436 wild genomic variation in captive ‘akeke‘e may be lower because the species occupies a larger portion of 437 its range on the Alaka‘i Plateau relative to ‘akikiki (Behnke et al. 2016, Fricker et al. in press). Therefore, 438 if additional egg collections are undertaken for ‘akeke‘e, attempts to sample eggs from individuals 439 containing diversity that is not encompassed in the managed population (e.g., Figure 4), if such nests can 440 be found and accessed, will increase overall genetic diversity in the conservation breeding population. 441 This strategy would amplify the probability of encapsulating all unique genetic variation and adding new 442 unrelated founders to the managed population, thus maximizing the long-term viability of both species. 443 The strategy of avoiding pairings of individuals from the same nest can be used in conjunction with the 444 kinship and network data presented here to maximize the amount of genetic variation within managed 445 individuals relative to the available pool of diversity. Thus, managers can use genomic data to guide 446 future breeding efforts (Galla et al. 2020), and these data may enable capturing a higher proportion of 447 genetic variation than strategies not informed by genomics. The high proportion of genetic diversity both 448 within and among individuals highlights the need to protect as many wild individuals as possible (Muya 449 et al. 2011). 450 The range of both ‘akeke‘e and ‘akikiki has been drastically restricted in recent years, due to the 451 combined forces of introduced predators, habitat disturbance from humans (Behnke et al. 2016), and the 452 arrival of introduced mosquitoes (Glad and Crampton 2015) and avian malaria (Atkinson et al. 2014) to 453 high elevation forests as a result of climate change and the introduction of cold-adapted mosquitoes 454 (Fonseca et al. 2006). With this range contraction, ongoing loss of genetic variation is expected in the 455 absence of intervention (Frankham et al. 2002). In line with this prediction, we observed evidence of 456 bottlenecks in both species, including heterozygosity excess and strongly positive Tajima’s D. 457 Conservation actions should aim to protect and restore the wild populations in the way that best 458 eliminates the current threats, which may include selecting existing or novel reintroduction and 459 translocation destination sites (Fortini et al. 2017) that have high quality forest habitat and low abundance 460 of mosquitoes. Large-scale mosquito and predator control efforts should be considered (Liao et al. 2017), 461 as reintroduction programs cannot succeed until the original threats are eliminated. Because these species 462 each exist as single functional populations, they are vulnerable to stochastic extinction (Griffen & Drake 463 2008); thus, intervention measures such as establishing novel sites (e.g., Warren et al. 2019) on higher 14 464 elevation islands, such as Maui or Hawai‘i, may be warranted as a last resort to prevent extinction 465 (Fricker et al. in press). Finally, continuing ongoing efforts to prioritize pairings of managed individuals 466 from different nests and the least-related individuals (Figure 2) will help to maintain maximum within- 467 individual variation. The extremely small effective population sizes (<20 birds) in both species reveal 468 their vulnerability to mutational meltdown (Lynch et al. 1995, Bank et al. 2016) and underscore the 469 importance of ongoing management to preserve existing genomic variation and to prevent these forest 470 bird species from going extinct. 471 Population bottlenecks caused by species introductions, climate change, and habitat modification 472 can lead to diversity loss among populations due to genetic drift, which can erode adaptive variation— 473 including alleles that may confer adaptation to these very selection pressures (Cassin-Sackett et al. 474 2019a). The high heterozygosity observed in ‘akeke‘e and ‘akikiki likely has both biological and 475 technical origins. For instance, the combination of ascertainment bias (selecting only variable sites) and 476 linkage disequilibrium resulting from bottlenecks results in high average heterozygosity. In addition, 477 these species appear to avoid inbreeding (consistent with observations in many other species, e.g., 478 Clutton-Brock 1989, Brouwer et al. 2011), which may slow the loss of genetic diversity within 479 individuals. Despite the high levels of measured diversity, a global loss of allelic variation is likely 480 inevitable in populations that have experienced severe bottlenecks. Nonetheless, the high heterozygosity 481 within ‘akeke‘e and ‘akikiki suggests that inbreeding depression and homozygosity at lethal alleles are 482 not imminent threats to these species; more pressing concerns are introduced avian malaria, introduced 483 predators, and the possibility of environmental catastrophes (e.g., hurricanes). Thus, unless specific alleles 484 conferring enhanced survival from malaria can be identified, the best breeding strategy is likely to 485 continue to pair the least-related individuals. 486 Under novel selection regimes, such as those imposed by introduced species, native species may 487 be pushed to the brink of extinction (Fortini et al. 2015). Other species on Kaua‘i, such as the ‘anianiau 488 (Magumma parva), Kaua‘i ‘amakihi (Chlorodrepanis stejnegeri), ‘apapane (Himatione sanguinea) and 489 i‘iwi (Drepanis coccinea), have also experienced declines (Paxton et al., 2016) as the temperature warms 490 and mosquito-free refugia are lost (Fortini et al. 2015). The changes on Kaua‘i hint at future similar 491 scenarios on other Hawaiian Islands (Liao et al. 2015) if large-scale integrative conservation efforts to 492 reduce malaria transmission are not undertaken (Liao et al. 2017). 493 Islands contain some of the world’s most imperiled species, which often face heightened pressure 494 from introduced species and human-induced environmental change. In addition, their populations are 495 often small owing to small geographic distributions, and thus are subject to elevated demographic 496 stochasticity. Nonetheless, many of these at-risk species may demonstrate genetic resilience that can be 15 497 leveraged in conservation. Island species can serve as models for species around the globe whose habitats 498 are becoming increasingly fragmented, causing their populations to operate as functional islands. 499 500 References 501 Alexander, D.H., Novembre, J., Lange, K., 2009. Fast model-based estimation of ancestry in unrelated 502 individuals. Genome Res. 19, 1655–1664. doi:10.1101/gr.094052.109.vidual 503 Arnholt, A.T., and Evans, B., 2017. BSDA: Basic Statistics and Data Analysis. R package version 1.2.0. 504 Atkinson, C.T., Utzurrum, R.B., Lapointe, D.A., Camp, R.J., Crampton, L.H., Foster, J.T., Giambelluca, 505 T.W., 2014. Changing climate and the altitudinal range of avian malaria in the Hawaiian Islands – 506 an ongoing conservation crisis on the island of Kaua’i. Global Change Biology 20, 2426–2436. 507 doi:10.1111/gcb.12535 508 Bandelt, H.J., Forster, P. and Röhl, A., 1999. Median-joining networks for inferring intraspecific 509 phylogenies. Molecular Biology and Evolution, 16(1), pp.37-48. 510 Bank, C., Renzette, N., Liu, P., Matuszewski, S., Shim, H., Foll, M., Bolon, D.N., Zeldovich, K.B., 511 Kowalik, T.F., Finberg, R.W. and Wang, J.P., 2016. An experimental evaluation of drug‐ induced 512 mutational meltdown as an antiviral treatment strategy. Evolution, 70(11), pp.2470-2484. 513 Barbato M, Orozco-terWengel P, Tapio M and Bruford MW, 2015. SNeP: a tool to estimate trends in 514 recent effective population size trajectories using genome-wide SNP data. Front. Genet. 6:109. doi: 515 10.3389/fgene.2015.00109 516 Behnke, L.A.H., Pejchar, L., Crampton, L.H., 2016. Occupancy and habitat use of the endangered Akikiki 517 and Akekee on Kauai Island, Hawaii. Condor 118, 148–158. doi:10.1650/CONDOR-15-80.1 518 BirdLife International. 2018a. Loxops caeruleirostris. The IUCN Red List of Threatened Species 2018: 519 e.T22720832A130851810. https://dx.doi.org/10.2305/IUCN.UK.2018- 520 2.RLTS.T22720832A130851810.en. 521 BirdLife International. 2018b. Oreomystis bairdi. The IUCN Red List of Threatened Species 2018: 522 e.T22720809A130843089. https://dx.doi.org/10.2305/IUCN.UK.2018- 523 2.RLTS.T22720809A130843089.en. 524 Bolger, A.M., Lohse, M., Usadel, B., 2014. Trimmomatic: a flexible trimmer for Illumina sequence data. 525 Bioinformatics 30, 2114–2120. doi:10.1093/bioinformatics/btu170 526 Brouwer, L., Van De POL, M.A., Atema, E.L.S. and Cockburn, A., 2011. Strategic promiscuity helps 527 avoid inbreeding at multiple levels in a cooperative breeder where both sexes are 528 philopatric. Molecular Ecology, 20(22), pp.4796-4807. 529 Butchart, S.H., Stattersfield, A.J. and Collar, N.J., 2006. How many bird extinctions have we 530 prevented?. Oryx, 40(3), pp.266-278. 16 531 Callicrate, T., Dikow, R., Thomas, J.W., Mullikin, J.C., Jarvis, E.D., Fleischer, R.C. and NISC 532 Comparative Sequencing Program, 2014. Genomic resources for the endangered Hawaiian 533 honeycreepers. BMC Genomics, 15(1), p.1098. 534 Camp, R.J., Gorresen, P.M., Pratt, T.K. and Woodworth, B.L., 2009. Population trends of native 535 Hawaiian forest birds, 1976–2008: the data and statistical analyses. Hawai‘i Cooperative Studies 536 Unit Technical Report HCSU-012. University of Hawai‘i at Hilo, Hilo. 537 Cassin-Sackett, L., Callicrate, T.E., Fleischer, R.C., 2019a. Parallel evolution of gene classes, but not 538 genes: Evidence from Hawai’ian honeycreeper populations exposed to avian malaria. Mol. Ecol. 28, 539 568–583. doi:10.1111/mec.14891 540 Cassin-Sackett, L., Welch, A.J., Venkatraman, M.X., Callicrate, T.E., Fleischer, R.C., 2019b. The 541 Contribution of Genomics to Bird Conservation, in: Avian Genomics in Ecology and Evolution. 542 Springer, Cham, pp. 295–330. 543 Clutton-Brock, T.H., 1989. Female transfer and inbreeding avoidance in social mammals. Nature 544 337(6202), pp. 70-72. 545 Cornuet, J. M. , & Luikart, G. (1996). Description and power analysis of two tests for detecting recent 546 population bottlenecks from allele frequency data. Genetics, 144(4), 2001-2014. 547 10.1093/oxfordjournals.jhered.a111627 548 Cortes-Rodriguez N, Campana MG, Berry L, Faegre S, Derrickson SR, Ha RR, Dikow RB, Rutz C, 549 Fleischer RC. 2019. Population genomics and structure of the critically endangered Mariana Crow 550 (Corvus kubaryi). Genes. 10:187. 551 Courchamp, F., Angulo, E., Rivalan, P., Hall, R.J., Signoret, L., Bull, L. and Meinard, Y., 2006. Rarity 552 value and species extinction: the anthropogenic Allee effect. PLoS biology, 4(12). 553 Dereeper A, Homa F, Andres G, Sempere G, Sarah G, Hueber Y, Dufayard JF, Ruiz M. SNiPlay3: a web- 554 based application for exploration and large scale analyses of genomic variations.. Nucleic Acids 555 Res. 2015 Jul 1;43(W1) 556 Danecek P, Auton A, Abecasis G, Albers CA, Banks E, DePristo MA, Handsaker RE, Lunter G, Marth 557 GT, Sherry ST, et al.; 1000 Genomes Project Analysis Group. 2011. The variant call format and 558 VCFtools. Bioinformatics. 27:2156–2158. 559 Diaz, H.F., Giambelluca, T.W., Eischeid, J.K., 2011. Changes in the vertical profiles of mean temperature 560 and humidity in the Hawaiian Islands. Glob. Planet. Change 77, 21–25. 561 doi:10.1016/j.gloplacha.2011.02.007 562 Dixon, P., 2003. VEGAN, a package of R functions for community ecology. J. Veg. Sci. 14, 927–930. 17 563 Do, C., Waples, R.S., Peel, D., Macbeth, G.M., Tillett, B.J. and Ovenden, J.R., 2014. NeEstimator v2: 564 re‐ implementation of software for the estimation of contemporary effective population size (Ne) 565 from genetic data. Molecular ecology resources, 14(1), pp.209-214. 566 Dray, S., Dufour, A.-B., 2007. The ade4 package: implementing the duality diagram for ecologists. J. 567 Stat. Softw. 22, 1–20. 568 Farhadinia, M.S., Johnson, P.J., Zimmermann, A., McGowan, P.J., Meijaard, E., Stanley‐ Price, M. and 569 Macdonald, D.W., 2020. Ex situ management as insurance against extinction of mammalian 570 megafauna in an uncertain world. Conservation Biology 988 - 996. 571 Fisher, H.I., Baldwin, P.H. 1947. Notes on the Red-billed Leiothrix in Hawaii. Pacific Science 45-51. 572 Fonseca, D.M., Smith, J.L., Wilkerson, R.C., Fleischer, R.C., 2006. Pathways of expansion and multiple 573 introductions illustrated by large genetic differentiation among worldwide populations of the 574 southern house mosquito. Am. J. Trop. Med. Hyg. 74, 284–9. 575 Fortini, L.B., Kaiser, L.R., Vorsino, A.E., Paxton, E.H., Jacobi, J.D., 2017. Assessing the potential of 576 translocating vulnerable forest birds by searching for novel and enduring climatic ranges. Ecol. 577 Evol. 7, 9119–9130. doi:10.1002/ece3.3451 578 Fortini, L.B., Vorsino, A.E., Amidon, F.A., Paxton, E.H., Jacobi, J.D., 2015. Large-scale range collapse 579 of Hawaiian forest birds under climate change and the need for 21st century conservation options. 580 PLoS One 10, e0140389. doi:10.1371/journal.pone.0140389 581 Foster, J.T., Tweed, E.J., Camp, R.J., Woodworth, B.L., Adler, C.D., Telfer, T., 2004. Long-term 582 population changes of native and introduced birds in the Alaka’i Swamp, Kaua’i. Conserv. Biol. 18, 583 716–725. 584 Frankham, R., Ballou, J.D. and Briscoe, D.A., 2002. Conservation genetics. Cambridge. 585 Fricker, GA, LH Crampton, EM Gallerani, JM Hite, R Inman, TW Gillespie. In press. Application of 586 LiDAR for Critical Endangered Bird Species Conservation on the island of Kauai, Hawaii. 587 Ecosphere. 588 Galla, S.J., Moraga, R., Brown, L., Cleland, S., Hoeppner, M.P., Maloney, R.F., Richardson, A., Slater, 589 L., Santure, A., Steeves, T.E., 2020. A comparison of pedigree, genetic, and genomic estimates of 590 relatedness for informing pairing decisions in two critically endangered birds: Implications for 591 conservation breeding programmes worldwide. Evolutionary Applications 13(5): 991-1008. 592 Gattepaille, L. M., Jakobsson, M., & Blum, M. G. (2013). Inferring population size changes with 593 sequence and SNP data: lessons from human bottlenecks. Heredity, 110(5), 409–419. 594 https://doi.org/10.1038/hdy.2012.120 595 Glad, A., Crampton, L.H., 2015. Local prevalence and transmission of avian malaria in the Alakai Plateau 596 of Kauai, Hawaii, U.S.A. J. Vector Ecol. 40, 221–229. 18 597 Gorresen, P.M., Camp, R.J. and Reynolds, M.H., 2009. Status and trends of native Hawaiian songbirds. 598 Chapter 5 in Conservation Biology of Hawaiian Forest Birds: Implications for Island Avifauna. TK 599 Pratt, CT Atkinson, PC Banko, JD Jacobi, BL Woodworth. 600 Griffen, B.D., Drake, J.M. 2008. A review of extinction in experimental populations. Journal of Animal 601 Ecology 77(6): 1274-1287. 602 Hammond, R.L., Crampton, L.H. and Foster, J.T., 2015. Breeding biology of two endangered forest birds 603 on the island of Kauai, Hawaii. The Condor: Ornithological Applications, 117(1), pp.31-40. 604 Harter, D.E. V, Irl, S.D.H., Seo, B., Steinbauer, M.J., Gillespie, R., Triantis, K.A., Beierkuhnlein, C., 605 2015. Perspectives in Plant Ecology , Evolution and Systematics Impacts of global climate change 606 on the floras of oceanic islands – Projections , implications and current knowledge 17, 160–183. 607 doi:10.1016/j.ppees.2015.01.003 608 Hawai‘i Department of Land and Natural Resources. 2015. Hawai‘i’s State Wildlife Action Plan. 609 Prepared by H. T. Harvey and Associates, Honolulu, Hawai‘i. 610 Hemmings, N.L., Slate, J. and Birkhead, T.R., 2012. Inbreeding causes early death in a passerine 611 bird. Nature Communications, 3(1), pp.1-4. 612 Hubisz, M.J., Falush, D., Stephens, M. and Pritchard, J.K., 2009. Inferring weak population structure with 613 the assistance of sample group information. Molecular ecology resources, 9(5), pp.1322-1332. 614 Jablonski, D., 1986. Background and mass extinctions: the alternation of macroevolutionary 615 regimes. Science, 231(4734), pp.129-133. 616 Jombart, T., 2008. adegenet: a R package for the multivariate analysis of genetic markers. Bioinformatics 617 24, 1403–5. doi:10.1093/bioinformatics/btn129 618 Jouganous, J., Long, W., Ragsdale, A.P. and Gravel, S., 2017. Inferring the joint demographic history of 619 multiple populations: beyond the diffusion approximation. Genetics, 206(3), pp.1549-1567. 620 Juola, F.A. and Dearborn, D.C., 2012. Sequence-based evidence for major histocompatibility complex- 621 disassortative mating in a colonial seabird. Proceedings of the Royal Society B: Biological 622 Sciences, 279(1726), pp.153-162. 623 Keller, L.F., Arcese, P., Smith, J.N., Hochachka, W.M. and Stearns, S.C., 1994. Selection against inbred 624 song sparrows during a natural population bottleneck. Nature, 372(6504), pp.356-357. 625 Kirch P.V. 2011. When did the Polynesians settle Hawaii? A review of 150 years of scholarly inquiry and 626 a tentative answer. Hawaiian Archaeology 12: 3-26 627 La Marca, E., Lips, K.R., Lotters, S., Puschendorf, R., Ibanez, R., Rueda-Almonacid, J.V., Schulte, R., 628 Marty, C., Castro, F., Manzanilla-Puppo, J., Garcia-Perez, J.E., Bolanos, F., Chaves, G., Pounds, 629 J.A., Toral, E., Young, B.E., 2005. Catastrophic Population Declines and Extinctions in Neotropical 630 Harlequin Frogs (Bufonidae: Atelopus)1. Biotropica 37, 190–201. doi:10.1111/j.1744- 19 631 7429.2005.00026.x 632 Lande, R., 1988. Genetics and demography in biological conservation. Science, 241(4872), pp.1455-1460. 633 Lande, R., 1993. Risks of population extinction from demographic and environmental stochasticity and 634 random catastrophes. The American Naturalist, 142(6), pp.911-927. 635 Lande, R., 1994. Risk of population extinction from fixation of new deleterious 636 mutations. Evolution, 48(5), pp.1460-1469. 637 Lande, R., 1998. Risk of population extinction from fixation of deleterious and reverse 638 mutations. Genetica, 102, pp.21-27. 639 Lerner, H.R., Meyer, M., James, H.F., Hofreiter, M. and Fleischer, R.C., 2011. Multilocus resolution of 640 phylogeny and timescale in the extant adaptive radiation of Hawaiian honeycreepers. Current 641 Biology, 21(21), pp.1838-1844. 642 Li H. 2013. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. arXiv. 643 1303.3997. 644 Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Marth, G., Abecasis, G., Durbin, R., 645 Data, G.P., Sam, T., 2009. The Sequence Alignment / Map format and SAMtools. Bioinformatics 646 25, 2078–2079. doi:10.1093/bioinformatics/btp352 647 Liao, W., Atkinson, C.T., Lapointe, D.A., Samuel, M.D., 2017. Mitigating Future Avian Malaria Threats 648 to Hawaiian Forest Birds from Climate Change. PLoS One 12, e0168880. 649 doi:10.1371/journal.pone.0168880 650 Liao, W., Timm, O.E., Zhang, C., Atkinson, C.T., LaPointe, D.A., Samuel, M.D., 2015. Will a warmer 651 and wetter future cause extinction of native Hawaiian forest birds ? Glob. Chang. Biol. 21, 4342– 652 4352. doi:10.1111/gcb.13005 653 Lips, K.R., Brem, F., Brenes, R., Reeve, J.D., Alford, R. a, Voyles, J., Carey, C., Livo, L., Pessier, A.P., 654 Collins, J.P., 2006. Emerging infectious disease and the loss of biodiversity in a Neotropical 655 amphibian community. Proc. Natl. Acad. Sci. U. S. A. 103, 3165–70. doi:10.1073/pnas.0506889103 656 Lynch, M., Conery, J. and Burger, R., 1995. Mutation accumulation and the extinction of small 657 populations. The American Naturalist, 146(4), pp.489-518. 658 Mashayekhi, M., MacPherson, B. and Gras, R., 2014. A machine learning approach to investigate the 659 reasons behind species extinction. Ecological Informatics, 20, pp.58-66. 660 McKenna A, Hanna M, Banks E, Sivachenko A, Cibulskis K, Kernytsky A, Garimella K, Altshuler D, 661 Gabriel S, Daly M, et al. 2010. The Genome Analysis Toolkit: a MapReduce framework for 662 analyzing next-generation DNA sequencing data. Genome Res. 20:1297–1303. 663 Mordecai, E.A., Cohen, J.M., Evans, M.V., Gudapati, P., Johnson, L.R., Lippi, C.A., Miazgowicz, K., 664 Murdock, C.C., Rohr, J.R., Ryan, S.J. and Savage, V., 2017. Detecting the impact of temperature on 20 665 transmission of Zika, dengue, and chikungunya using mechanistic models. PLoS neglected tropical 666 diseases, 11(4), p.e0005568. 667 Muya, S.M., Bruford, M.W., Muigai, A.T., Osiemo, Z.B., Mwachiro, E., Okita-Ouma, B. and Goossens, 668 B., 2011. Substantial molecular variation and low genetic structure in Kenya’s black rhinoceros: 669 implications for conservation. Conservation genetics, 12(6), pp.1575-1588. 670 Nei, M., Li, W.-H., 1979. Mathematical model for studying genetic variation in terms of restriction 671 endonucleases. Proc. Natl. Acad. Sci. USA 76, 5269–5273. 672 Nei M, Maruyama T, Chakraborty R (1975). The bottleneck effect and genetic variability in populations. 673 Evolution 29: 1–10. 674 Noskova, E., Ulyantsev, V., Koepfli, K., Brien, S.J.O., 2020. GADMA: Genetic algorithm for inferring 675 demographic history of multiple populations from allele frequency spectrum data. GigaScience 9, 1- 676 18. 677 Padilla, D.P., Illera, J.C., Gonzalez-Quevedo, C., Villalba, M., Richardson, D.S., 2017. Factors affecting 678 the distribution of haemosporidian parasites within an oceanic island. Int. J. Parasitol. 47, 225–235. 679 doi:10.1016/j.ijpara.2016.11.008 680 Palkopoulou, E., Mallick, S., Skoglund, P., Enk, J., Rohland, N., Li, H., Omrak, A., Vartanyan, S., Poinar, 681 H., Götherström, A. and Reich, D., 2015. Complete genomes reveal signatures of demographic and 682 genetic declines in the woolly mammoth. Current Biology, 25(10), pp.1395-1400. 683 Paradis E (2010). “pegas: an R package for population genetics with an integrated–modular 684 approach.” Bioinformatics, 26, 419-420. 685 Paxton, E.H., Brinck, K.W., Crampton, L.H., Hite, J. and Costantini, M., 2020. 2018 Kaua'i forest bird 686 population estimates and trends. 687 Paxton, E.H., Camp, R.J., Gorresen, P.M., Crampton, L.H., Leonard, D.L.J., Vanderwerf, E.A., 2016. 688 Collapsing avian community on a Hawaiian island. Sci. Adv. 2, e1600029. 689 doi:10.1126/sciadv.1600029 690 Paxton, E.H., M. Laut, J.P. Vetter, and S.J. Kendall. 2018. Research and management priorities for 691 Hawaiian forest birds. Condor 120:557-565 692 Paull, S.H., LaFonte, B.E. and Johnson, P.T., 2012. Temperature‐ driven shifts in a host‐ parasite 693 interaction drive nonlinear changes in disease risk. Global Change Biology, 18(12), pp.3558-3567. 694 Pimm, S.L., Jenkins, C.N., Abell, R., Brooks, T.M., Gittleman, J.L., Joppa, L.N., Raven, P.H., Roberts, 695 C.M. and Sexton, J.O., 2014. The biodiversity of species and their rates of extinction, distribution, 696 and protection. Science, 344(6187), p.1246752. 697 Pritchard, J.K., Stephens, M., Donnelly, P., 2000. Inference of population structure using multilocus 698 genotype data. Genetics 155, 945–959. 21 699 Prowse, T.A., Johnson, C.N., Lacy, R.C., Bradshaw, C.J., Pollak, J.P., Watts, M.J. and Brook, B.W., 700 2013. No need for disease: testing extinction hypotheses for the thylacine using multi‐ species 701 metamodels. Journal of Animal Ecology, 82(2), pp.355-364. 702 R Core Team. 2018. R: A Language and Environment for Statistical Computing; R Foundation for 703 Statistical Computing: Vienna, Austria. 704 Ralls, K., Ballou, J.D. and Templeton, A., 1988. Estimates of lethal equivalents and the cost of inbreeding 705 in mammals. Conservation biology, 2(2), pp.185-193. 706 Robert, A. 2011. Find the weakest link. A comparison between demographic, genetic and demo-genetic 707 metapopulation extinction times. BMC Evol Biol 11, 260. https://doi.org/10.1186/1471-2148-11-260 708 Rodrigues, A.S., 2006. Are global conservation efforts successful?. Science, 313(5790), pp.1051-1052. 709 Roelke, M. E., Martenson, J. S., & O’Brien, S. J. (1993). The consequences of demographic reduction 710 and genetic depletion in the endangered Florida panther. Current Biology, 3(6), 340–350. 711 Rogers, R.L. and Slatkin, M., 2017. Excess of genomic defects in a woolly mammoth on Wrangel 712 island. PLoS genetics, 13(3). 713 Rohlf, F.J., 1972. An empirical comparison of three ordination techniques in numerical 714 taxonomy. Systematic zoology, 21(3), pp.271-280. 715 Rousset, F. (2008). Genepop'007: a complete reimplementation of the Genepop software for Windows 716 and Linux. Mol. Ecol. Res. 8: 103-106. 717 Samuel, M.D., Hobbelen, P.H.F., DeCastro, F., Ahumada, J.A., LaPointe, D.A., Atkinson, C.T., 718 Woodworth, B.L., Hart, P.J., Duffy, D.C., 2011. The dynamics, transmission, and population 719 impacts of avian malaria in native Hawaiian birds: a modeling approach. Ecol. Appl. 21, 2960– 720 2973. 721 Samuel, M.D., Woodworth, B.L., Atkinson, C.T., Hart, P.J., LaPointe, D.A., 2015. Avian malaria in 722 Hawaiian forest birds: infection and population impacts across species and elevations. Ecosphere 6, 723 1–21. 724 Santymire, R.M., Livieri, T.M., Branvold-Faber, H. and Marinari, P.E., 2014. The black-footed ferret: on 725 the brink of recovery?. In Reproductive sciences in animal conservation (pp. 119-134). Springer, 726 New York, NY. 727 Shaffer, M.L., 1983. Determining minimum viable population sizes for the grizzly bear. Bears: Their 728 biology and management, pp.133-139. 729 Smith, K.F., Sax, D.F., Lafferty, K.D., 2006. Evidence for the role of infectious disease in species 730 extinction and endangerment. Conserv. Biol. 20, 1349–57. doi:10.1111/j.1523-1739.2006.00524.x 731 Spiller, D.A., Losos, J.B. and Schoener, T.W., 1998. Impact of a catastrophic hurricane on island 732 populations. Science, 281(5377), pp.695-697. 22 733 Sutton, JT. 2014. Recommendations for the number of founders, the target captive population size, and 734 choosing the number of harvest sites for maintaining genetic diversity in captive populations of 735 ‘Akikiki and ‘Akeke‘e. Report submitted to the Kauaʻi Forest Bird Recovery Group, 7pp. 736 United States Fish and Wildlife Service 2010. Endangered and Threatened Wildlife and Plants; 737 Determination of Endangered Status for 48 Species on Kauai and Designation of Critical Habitat. 738 Federal Register, 50 CFR Part 17. 739 Warren, C., Mounce, H., Berthold, L., Farmer, C., Leonard, D. and Duvall, F., 2019. Experimental 740 restoration trials in Nakula Natural Area Reserve in preparation for reintroduction of Kiwikiu 741 (Pseudonestor xanthophrys). Pacific Cooperative Studies Unit Technical Report, pp.1-102. 742 Wootton, J.T. and Pfister, C.A., 2013. Experimental separation of genetic and demographic factors on 743 extinction risk in wild populations. Ecology, 94(10), pp.2117-2123. 744 Zheng, X.; Levine, D.; Shen, J.; Gogarten, S.M.; Laurie, C.; Weir, B.S. 2012. A high-performance 745 computing toolset for relatedness and principal component analysis of SNP data. Bioinformatics 28, 746 3326–3328. 747 748 749 Figure captions: 750 Fig. 1 Map of Kaua‘i study area and field sites, with tan outline (extending across approximately 40 km2) 751 encompassing the geographic range of both ‘akeke‘e and ‘akikiki, which occurs at the highest elevations 752 on the island. PIH = Pihea, KWK = Kawaikōī, UUK = Upper Kawaikōī, MOH = Mohihi, HPK = 753 Halepa‘akai 754 755 Fig. 2 Kinship matrixes showing estimated pairwise relatedness among ‘akeke‘e (top left) and ‘akikiki 756 (bottom right) individuals. Darker shading represents higher kinship; triangles surround individuals in the 757 conservation breeding population. The primarily light colors in pairwise relationships among individuals 758 in the conservation breeding program (triangles) indicate the low degree of relatedness among individuals 759 760 Fig. 3 Principal Coordinates Analysis of (top) ‘akeke‘e genotypes from four sampled sites and (bottom) 761 ‘akikiki genotypes from five sampled sites on the Alaka’i Plateau 762 763 Fig. 4 Median joining networks of ‘akeke‘e (left) and ‘akikiki (right) genomic variation. Wild individuals 764 are denoted in dark/purple circles, managed individuals in light/green circles, and median vectors 765 (unobserved intermediate branching nodes) in white circles; hash marks represent mutations. In both 766 species, the main figure comprises potentially extant individuals, while the inset figure also contains 23 767 individuals sampled in the 1990s. In ‘akeke‘e, the managed population captures most extant diversity but 768 misses a small portion of existing wild diversity (left side of network), while in ‘akikiki, managed 769 individuals are present throughout the network. In ‘akeke‘e, some diversity was present in the 1990s that 770 is not contained among extant individuals 771 772 Fig. 5 Effective population size (Ne) over time as estimated in SNeP (Barbato et al. 2015); plots show 773 only the most recent 150 generations to facilitate visualization of recent trends. Triangle point is the 774 estimate of current effective size inferred from NeEstimator (Do et al. 2014). Longer time scales are 775 shown in Figure S5. Top: ‘akeke‘e, photo by Lucas Behnke; bottom: ‘akikiki, photo by Justin Hite 776 777 778 24 Figure 1 Click here to access/download;Figure;Figure1.jpg Figure 2 Click here to access/download;Figure;Figure2_kinship.pdf 36 SB5 35 SB3 SBV260 4 34 SBSB2V2 6 3 SBV168 33 SB 2 S1VB1 6 31 32 AE KV266800CE_VM8579 4 -1 31 AE A_VI_ AI_15 468 440 30 V57AEA0IV_40536 AI_42 29 AE0V055 AVI_54144 28 AIV_4503 25 A2VI_05372-b8 9311 27 22A8IV_3541 AI_206- 37705 26 AV50228IV020459- 37663 AI024 25 228V48AIV020437- 37662 24 AI22 A8V 022 I004216- 3 7660 23 AI V V01 4 49 5 22 4A8IV0107-43 37522 22 21A7IV012 AIV01 1412-41 77661 21 21A7I0110- 77660 20 AI V V00 490 21A7IV001 38938- 77659 19 AI007 21 A4VI013057- 6 8354 18 21A4I V V00 3 34 6 2631-5-3617065 34 17 21242VV81 3 30- 4 3-36777345 24 16 22V830-37336 21242V8130- 2 1-06937005 20 15 2280-09190 19241V9031-0-2030V29 6764 17 14 19224 1V7021-141-8- 707640 6836364 13 13 19148 V2201- 7 -201411 18V12 2 21-62145075 93 19148V2021-5-20145034 91 11 18V221-421402 18128V2521-3-22130757 40 10 18V221-221369 18128V221-1-22131684 08 9 V 181282 21-0-2213166 8 18V211-921365 3 39 18128V21-821364 18V211-7-22131632 32 7 18128V211-6-2213162 6 18V215 155 1-21361 18128V211-4-22131601 5 46 18V21-321356 4 18128V21-2-2130547 84 18128V21-121347 3 18V211-0-22130267 48 18128V2191--22110876 2 V8 60 1821-21176 1 18 128V2171--22110546 54 18V261-20775 1518V2051--24070740 61 18V241-20737 18V231-20694 15V820-40035 X1580.40044 1 X151805-4800023.54 0061 X1821-20181218-2201 639 743. 4 72 0654 X181218-2201757.42 0660 1821-20775 X181 6 218-2211175.42 0748 X181218-221118761821-21198.72 0784 X181218-22111302.62 1146 1821-2113147 X181218-22111325.42 1155 X18181 218-22111335.6 21-21134602 1232 X181218-22111356.12 1339 1821-2113X 6 62 181218-22111376.32 1408 X181218-2211538641821-211396.52 0740 X181219-24210306.60 5491 1821-2213168 X181219-24210326.90 5593 X181219-22133751821-24210440.20 6413 X181219-24210450.30 6417 1821-2214607 X182211-24211471.16 7520 X212411-64283836 2171-7271694.06 7524 X212911-24331076 2280-09109.06 7534 X222801-0433913 1 20.06 8354 X222801-37371333.67 2280-37734 7659 X26231-3 361406 AI0014 7 315.77660 XAI20015 7 31637.77661AI007 XAI20028 8 330 8 9.37522 XAI20029 8 AI010 400.37660 XAI20121 8 41 AI012 4 02.37662 XAI20127 8 403.37663 XAI20129 8 44 AI021 405.37705 XAI20252 2 406 AI023 47 .89311 AI024 48AE004 AI025 49 AI_26 5 0AE006 AI_34 51AE_10 AI_37b5 2 AI_40 53 AE_8 AI_41 54 AI_42 55 SB1 AI_43 56 SB2 AI_44 5 7 AI_46 58 SB3 CEM745-19 K2680 60 SB13 61 SB18 SB622 SB20 63 Figure 3 Click here to access/download;Figure;Figure3_PCoA.pdf Principal Coordinates Analysis of 'Akeke'e Principal Coordinates Analysis of 'Akikiki UUK KWK HPK UUK KWK PIH HPK MOH MOH -0.05 0.00 0.05 -0.10 -0.05 0.00 0.05 PCoA 1 PCoA 1 method = "Provesti" method = "Provesti" PCoA 2 -0.08 -0.06 -0.04 -0.02 0.00 0.02 0.04 PCoA 2 -0.08 -0.06 -0.04 -0.02 0.00 0.02 0.04 Figure 4 Click here to access/download;Figure;Figure4_AKEK_AKIK_network_final.pdf Figure 5 Click here to access/download;Figure;Figure5_Ne.pdf 0 50 100 150 FALSE 0 50 100 150 Generations Before Present Ne Ne 0 1000 3000 5000 0 500 1000 1500 2000 Table 1 Click here to access/download;Table;Table 1.docx Table 1. Pairwise FST between sampling sites of ‘akeke‘e (upper triangle, in bold) and ‘akikiki (lower triangle). HPK KWK UUK PIH MOH HPK 0.00440 0.00549 0.01413 KWK -0.00260 0.01207 0.00650 UUK 0.02069 0.04210 0.02285 PIH -0.00133 -0.01336 0.03814 MOH 0.01889 0.01918 0.03014 0.01661 Table 2 Click here to access/download;Table;Table 2.docx 2 Table 2. Mean relatedness among and diversity within wild individuals and managed individuals; 3 statistics were calculated in VCFtools. ‘All years’ refers to all sampled wild individuals including those 4 sampled in the 1990s; ‘current’ refers to individuals sampled recently enough that the birds may still be 5 alive (i.e., since 2014). This distinction was designed to evaluate the recent loss of diversity. Estimates 6 may differ from whole-species estimates in Table S1 due to the automatic exclusion of non-informative 7 loci in data subsets (e.g., quality filtered managed AKEK comprised a dataset of only nine individuals). 8 Sample sizes are as follows: AKEK wild all N=29, AKEK wild current N=12, AKEK managed N=9, 9 AKIK wild all N=31, AKIK wild current N=25, and AKIK managed N=30. Statistic Species Wild Wild Managed (all years) (current) Relatedness AKEK 0.0161 0.0334 0.0519 AKIK 0.0194 0.0236 0.0192 AKEK 1.64 x 10-6 1.75 x 10-6 1.55 x 10-6 Nucleotide diversity (π) AKIK 1.16 x 10-6 1.17 x 10-6 1.03 x 10-6 Inbreeding coefficient (F) AKEK -0.4853 -0.4818 -0.3631 AKIK -0.3644 -0.3652 -0.3605 10 11 12 13 14 15 1 Supplementary Tables Click here to view linked References Click here to access/download Supplementary Material supplementary_tables.xlsx Supplementary Figs/ Methods/Results Click here to view linked References Click here to access/download Supplementary Material kauai_endangereds_supplement_revised.docx