Corymbophanes andersoni

Lujan, Nathan K., Armbruster, Jonathan W., Werneke, David C., Teixeira, Túlio Franco & Lovejoy, Nathan R., 2020, Phylogeny and biogeography of the Brazilian-Guiana Shield endemic Corymbophanes clade of armoured catfishes (Loricariidae), Zoological Journal of the Linnean Society 188, pp. 1213-1235 : 1215-1220

Corymbophanes andersoni

GUYANA, Region 8 (Potaro–Siparuni), Potaro River– Essequibo River drainage: FMNH 52675 , holotype , 65.5 mm SL, Aruataima Falls , Upper Potaro; AUM 28149 , 3 , 25.6 –57.0 mm SL, plus 1 cleared and stained, INHS 49586 , 2 , 17.3 , 64.9 mm SL, topotypes ,

Aruataima (Chenapou) Falls, 23.7 km southwest of Menzies Landing, 05.00139°, -059.62583°.

PHYLOGENETIC TAXON SAMPLING

In addition to both Corymbophanes andersoni and C. kaiei , we included in our phylogenetic analysis several populations of Corymbophanes from different parts of the Kuribrong River basin, giving us the opportunity to test hypotheses related not only to species relationships, but also population-level divergence and pathways by which Corymbophanes may have dispersed into and throughout the upper Kuribrong River drainage. The geographic distribution of these samples is illustrated in Figure 1. We also included both specimens of the enigmatic recently collected species from the upper Ireng River.

As outgroups, we included representatives of six other genera found to be included in the tribe Ancistrini in the multilocus analysis by Lujan et al. (2015): Ancistrus , Dekeyseria , Guyanancistrus , Hopliancistrus , Lasiancistrus and Pseudolithoxus , plus Cryptancistrus , which was found to be sister to Corymbophanes in a multilocus analysis by Fisch-Muller et al. (2018), and Araichthys , which was found to be sister to Hopliancistrus in a phylogenomic analysis by Roxo et al. (2019). To root our trees, we included Lithogenes villosus , which has been found to be either sister to all other Loricariidae based on morphological data ( Schaefer, 2003) or part of a basal polytomy with Delturinae based on multilocus analyses ( Lujan et al., 2015) and Cteniloricaria platystoma , which is a member of the subfamily Loricariinae that is sister to Hypostominae + Hypoptopomatinae.

Drainage Oyapock Oyapock Xingu Tapajos Marañon Orinoco Ireng Ireng Country French Guiana French Guiana Brazil Brazil Peru Venezuela Guyana Guyana Cat # 2725.100 2722.089 193087 39853 45548 57674 1722 F 67193 Voucher MHNG MHNG ANSP AUM AUM AUM CSBD AUM 2 RAG X X X X X X X X 1 RAG X X X X X X X X 2 ND X X X Cytb X X X X X X S 16 X X X X X X X X of # loci 4 4 4 4 4 5 4 4 Type species X X X X Type specimen X X # - 204 - 185 10277 10303 Tissue GF 99 99 GF 2167 B 9018 T P 6123 09376 T AUF AUF longispinis niger Xingu

n

. sp. tricornis tigris

.

Table

2

Continued

Taxa

Guyanancistrus Guyanancistrus Hopliancistrus Hopliancistrus Lasiancistrus schomburgkii Pseudolithoxus primus Yaluwak Yaluwak primus

TISSUE AND DNA SOURCES

Newly generated sequence data ( Table 2) were obtained from tissue samples or DNA extracts collected by the authors or provided by the Academy of Natural Sciences of Drexel University in Philadelphia , PA, USA ( ANSP), the Auburn University Museum Fish Collection in Auburn , AL, USA ( AUM), the Laboratório de Biologia e Genética de Peixes , Departamento de Morfologia , Instituto de Biociências , Universidade Estadual Paulista ‘ Júlio de Mesquita Filho’ , Campus de Botucatu , São Paulo, Brazil ( LBP), the Royal Ontario Museum in Toronto , Canada ( ROM), the Muséum d’Histoire Naturelle, Geneva, Switzerland ( MHNG) or obtained via the ornamental fish trade. Voucher specimens ( Table 2) were identified either by direct examination or in collaboration with museum workers at different institutions. Most of the taxa in our analysis were represented in either this or previous analyses by multiple individuals, but Corymbophanes andersoni was represented in our analysis by only a single degraded tissue collected in 1998. Institutional abbreviations follow Sabaj (2016) .

MOLECULAR MARKERS, DNA EXTRACTION, AMPLIFICATION AND SEQUENCING

Molecular phylogenetic methods followed those of Lujan et al. (2015) with the exception that the mitochondrial gene region NADH dehydrogenase 2 (ND2) was added to this analysis and the nuclear gene region MyH6 was not examined in this study. In brief, we amplified and sequenced a fragment of the mitochondrial 16S (538 bp), cytochrome b (865 bp) and ND2 (1040 bp) genes, as well as fragments of the nuclear RAG1 (807 bp) and RAG2 (873 bp) genes for a total of 4123 aligned base pairs. Most gene regions were sequenced from most taxa ( Table 2), with the exception that only the 16S gene region could be amplified and sequenced from Corymbophanes andersoni .

Gene regions were amplified using combinations of previously published primers ( Arroyave et al., 2013; Lujan et al., 2015). Whole genomic DNA was extracted from fin or muscle tissues preserved in 95% ethanol following either manufacturer’s instructions for the DNeasy Blood & Tissue Kit (Qiagen N.V., Venlo, Netherlands) or standard laboratory protocols for salt extraction followed by ethanol precipitation. Fragment amplifications were performed following the methods of Arroyave et al. (2013) and Lujan et al. (2015).

Post-PCR clean-up of all loci was achieved by either running the entire volume of PCR product on a 1% agarose gel with 0.01% SYBR Safe DNA gel stain (LTI: Life Technologies Inc., Carlsbad, CA) or by adding ExoSap-IT (Applied Biosystems Co., Foster City, CA) and following manufacturer’s instructions. For samples that were gel purified, the band corresponding to the target locus was cut from the gel and the target PCR product extracted by centrifuge filtration through the top of a P-200 pipette filter tip in a labelled 1 mL snaptop tube (5 min at 15 000 rpm) followed by precipitation and washing of the DNA to remove salts. Forward and reverse sequencing reactions either followed the manufacturer’s recommendations for sequencing on an Applied Biosystems 3730 DNA Analyzer (LTI) at the Royal Ontario Museum or were conducted by staff at The Centre for Applied Genomics at The Hospital for Sick Children (SickKids) in Toronto, ON, Canada.

SEQUENCE ASSEMBLY, ALIGNMENT AND PHYLOGENETIC INFERENCE

Sequence data were assembled, edited, aligned and concatenated following the methods of Lujan et al. (2015). PartitionFinder (v.1.1.1, Lanfear et al., 2012) was used to determine codon-position specific models of molecular evolution for each gene under the Bayesian information criterion (BIC).

For the Bayesian analysis, an HKY model with rate heterogeneity being modelled by a gamma distribution (HKY+G) was determined to be the best model of molecular evolution for third codon positions of ND2 and Cytb and first codon positions of RAG1 and RAG2. A GTR model with a proportion of invariable sites estimated and with rate heterogeneity being modelled by a gamma distribution (GTR+I+G) was determined to be the best model for 16S and the first codon positions of ND2 and Cytb. An HKY model with a proportion of invariable sites estimated (HKY+I) was determined to be the best model for the second codon positions of ND2 and Cytb. A K80 model with a proportion of invariable sites estimated (K80+I) was determined to be the best model for the second codon positions of RAG1 and RAG2. And a K80 model with rate heterogeneity being modelled by a gamma distribution (K80+G) was determined to be the best model for the third codon positions of RAG1 and RAG2. All data partitions were unlinked with rates free to vary across partitions and Lithogenes villosus designated as the outgroup.

For the Bayesian analysis, a Markov chain Monte Carlo (MCMC) search of tree space was conducted using MrBayes (v.3.2.3; Ronquist & Huelsenbeck, 2003) programmed to run for 10 million generations using two sets of eight chains (one cold, seven hot, with default temperature parameter), sampling every 666 trees with the first 5000 trees (~33%) being discarded as burn-in, thus generating a total of 10 000 trees from which posterior probabilities were calculated. The Bayesian search was determined to have reached stationarity when likelihood values of the cold chains began randomly fluctuating within a stable range and when effective sample sizes for all metrics exceeded 2000 as determined by the program TRACER (v.1.6; Rambaut et al., 2007).

For the maximum likelihood analysis, the concatenated alignment was also partitioned by genes and codon positions, but the same model (GTR+G) was used for all partitions. Maximum likelihood analysis was conducted using RAxML (v.8.0.0; Stamatakis, 2014) run locally, with a 200 generation GTR+G search for a best tree and a 2000 generation GTR+G bootstrap.

PRESENTATION OF PHYLOGENETIC RESULTS

Complete results of the Bayesian and maximum likelihood analyses are presented as Supporting Information ( Figs S1, S 2). Manuscript figures were trimmed of select outgroup taxa and were based on results of the Bayesian analysis. Node support values from both the Bayesian and maximum likelihood analyses are provided in Table 3. We also provide Bayesian posterior probability (i.e. Bayesian inference = BI) and maximum likelihood (ML) bootstrap support values for each node discussed in the text.