﻿Two new species of Centroptilum Eaton, 1869 from North Africa (Ephemeroptera, Baetidae)

﻿Abstract Based on recently collected larvae from Algeria and Morocco, the species delimitation within the genus Centroptilum Eaton, 1869 in that region is validated. Two new species are described and illustrated, one from north-eastern Algeria, and one from North Morocco, using an integrated approach with morphological and molecular evidence. A table summarising the morphological differences between the new species and Centroptilumluteolum (Müller, 1776) from Central Europe is provided. Further, molecular evidence for additional undescribed species of Centroptilum in other regions of the West Palearctic is provided and discussed.


Introduction
provided a provisional checklist of the mayflies from the Maghreb including 69 species: 41 from Morocco, 50 from Algeria, and 29 from Tunisia. This checklist included 17 species of Baetidae, nine additional species of this family needed to be confirmed. During the last two decades, important improvements were made in the knowledge of North African mayflies. A few new species of Baetidae, The genus Centroptilum was reported from the whole Maghreb. In Tunisia, the genus seems to be extremely rare as Boumaïza and Thomas (1995) only reported a single larva in their extensive survey of the country; they also considered it to be the most sensitive species to ionic concentration. In Algeria, the genus has a very limited distribution as it was recently only collected in the El Kala basin (Samraoui et al. 2021a); it seems to be absent from surrounding basins in East Algeria and other parts of the country (Benhadji et al. 2020;Samraoui et al. 2021b). Its distribution is also limited in Morocco as it was only collected in the northern part of the country (El Alami et al. 2022a). As already previously stated (Samraoui et al. 2021a;El Alami et al. 2022a), the genus Centroptilum needs to be revised in North Africa. In the present study, we use recently collected specimens from north-eastern Algeria and North Morocco to validate the species delimitation, and to describe two new species; we use an integrative approaches combining morphological and molecular evidence.

Materials and methods
The specimens from Algeria were collected between 2018 and 2020 by BS, and the specimens from Morocco in 2014 and 2021 by MEA and collaborators. Comparative material from Switzerland was collected by André Wagner (MZL). The larvae were preserved in 70%-96% ethanol.
The dissection of larvae was done in Cellosolve (2-Ethoxyethanol) with subsequent mounting on slides with Euparal liquid, using an Olympus SZX7 stereomicroscope.
Drawings were made using an Olympus BX43 microscope. To facilitate the determination of the new species and the comparison of important structures with other species, we partly used a combination of dorsal and ventral aspects in the same drawing (see Kaltenbach et al. 2020: fig. 1c).
Photographs of larvae were taken using a Canon EOS 6D camera and processed with Adobe Photoshop Lightroom (http://www.adobe.com) and Helicon Focus v. 5.3 (http://www.heliconsoft.com). Photographs of body parts of the larvae were taken with an Olympus BX51 microscope equipped with an Olympus SC50 camera and processed with Olympus (recently Evident) software Stream Basic v. 1.3. All pictures were subsequently enhanced with Adobe Photoshop Elements 13.
Distribution maps were generated with SimpleMappr (https://simplemappr.net, Shorthouse 2010). The GPS coordinates of the sample locations are given in Table 1. The terminology follows Hubbard (1995) and Kluge (2004). Table 2 of this study was partly developed based on Martynov et al. (2022: table II).
For the molecular part of the study, we first downloaded all Centroptilum cytochrome oxidase subunit 1 (COI) sequences available on GenBank as on 13.04.2022 using a custom script, resulting in 99 records. We then manually removed all sequences from specimens collected outside the Western Palearctic, resulting in 34 European sequences for further analyses. We also examined the sequences available on the BOLDSYSTEMS data portal as on 13.04.2022, but excluded all sequences shared with GenBank, those from specimens collected outside the Western Palearctic, and one sequence that did not blast with Centroptilum (i.e., most probably resulting from a misidentification or a contamination). As a result, no additional sequence could be obtained. We also included three sequences from the European mayfly FREDIE project (unpublished; https://wp.fredie.eu/). Finally, seven specimens were newly sequenced for this study ( Table 1; the nomenclature of gene sequences follows Chakrabarty et al. (2013)), for a total of 44 Centroptilum sequences in our molecular data set. The DNA of the sequenced specimens was extracted using non-destructive methods allowing subsequent morphological analysis (see Vuataz et al. 2011 for details). We amplified a 658 bp fragment of the COI gene using the primers LCO 1490 and HCO 2198 (Folmer et al. 1994, see Kaltenbach and Gattolliat 2020 for details). Sequencing was done with Sanger's method (Sanger et al. 1977). Forward and reverse sequencing reads were assembled and edited in CodonCode Aligner 10.0.2 (Codon-Code Corporation, Dedham, MA), and aligned using MAFFT (Katoh et al. 2019) with default settings as implemented in Jalview 2.11.2.2 (Waterhouse et al. 2009). The best evolutionary model (HKY+ Γ +I) was selected following the second-order Akaike information criterion (AICc; Hurvich and Tsai 1989) implemented in JModelTest 2.1.10 (Darriba et al. 2012) with seven substitution schemes and all other parameters set to default. In order to accommodate different substitution rates among COI codon positions, we analysed our data set in two partitions, one with first and second codon positions and one with third positions (1 + 2, 3). Bayesian inference (BI) gene tree reconstruction was conducted in MrBayes 3.2.7a (Ronquist et al. 2012). Two independent analyses of four MCMC chains run for five million generations with trees sampled every 1'000 generations were implemented, and 500'000 generations were discarded as a burn in after visually verifying run stationarity and convergence in Tracer 1.7.2 (Rambaut et al. 2018). One representative of four species belonging to the same subfamily as Centroptilum (i.e., Cloeoninae sensu Bauernfeind and Soldán 2012) were used as outgroup. The consensus tree was visualised and edited in iTOL 6.5.7 (Letunic and Bork 2021).
To explore COI evolutionary divergence and compare it to our morphological identifications, we applied three single-locus species delimitation methods to our CO1 data set: the distance-based ASAP (Assemble Species by Automatic Partitioning; Puillandre et al. 2020), the tree-based GMYC (General Mixed Yule-Coalescent; Pons et al. 2006;Fujisawa and Barraclough 2013), and mPTP (multi-rate Poisson Tree Processes; Kapli et al. 2017) approaches. The ASAP method, which is an improvement of the widely used ABGD (Automatic Barcode Gap Discovery; Puillandre et al. 2012) approach, has the advantage of providing a score that designates the most likely number of hypothetical species. The GMYC model, which requires a time-calibrated ultrametric tree as input, implements a Maximum Likelihood (ML) approach that defines a threshold separating the branches modelled under speciation events (Yule process) from those described by allele neutral coalescence. The mPTP approach, which is a multi-rate extension of the PTP (Poisson Tree Processes; Zhang et al. 2013), also exploits intra-and interspecies phylogenetic differences, but with the advantage of directly using the number of substitutions from a phylogenetic tree, eliminating the need for time calibration.
ASAP was applied to our COI alignment using the ASAP webserver available at https://bioinfo.mnhn.fr/abi/public/asap/asapweb.html, computing the genetic distances under the Kimura 2-parameter substitution model (K2P; Kimura 1980) with all other settings set to default. Input BI ultra-metric tree for GMYC was generated in BEAST 1.10.4. . To avoid potential biases in threshold estimation, the outgroups were removed, and identical CO1 haplotypes were pruned (see Talavera et al. 2013) using Collapsetypes 4.6 (Chesters 2013). Input BEAST file was created in BEAUTi , implementing the best model of evolution and the partition scheme specified above, and selecting a relaxed molecular clock (uncorrelated lognormal) model, a coalescent (constant size) prior (see Monaghan et al. 2009) and a UPGMA starting tree. Two independent MCMC chains were run for 50 million generations, sampling trees every 1'000 generations. Run convergence was visually verified in Tracer and the independent log and tree files were combined using LogCombiner 1.10.4  after discarding 10% of the trees as burnin. The maximum clade credibility tree, generated in TreeAnnotator 1.10.4 ) with all options set to default, was used as input for GMYC, which was run in R 4.2.0 (R Core Team 2022) using the SPLITS package 1.0-20 (Ezard et al. 2009). We favoured the single-threshold version of the GMYC model because it was shown to outperform the multiple-threshold version (Fujisawa and Barraclough 2013). Input ML tree for mPTP was generated in RAxML-NG 1.1.0 (Kozlov et al. 2019) from our CO1 alignment (outgroup included), selecting the all-in-one (ML search + bootstrapping) option and MRE-based bootstrap convergence criterion. The best model of evolution and the partition scheme specified above, as well as 50 random and 50 parsimony starting trees were implemented. mPTP was conducted on the web service available at https://mptp.h-its.org. Finally, the number of parsimony-informative sites and the mean COI genetic distances between and within species were calculated in MegaX (Kumar et al. 2018;Stecher et al. 2020) under the K2P model. Algeria Louar inf. 36°37'03"N, 08°22'49"E GBIFCH00763735 OP113123 genseq-2 COI Guitna sup. 36°36'42"N, 08°21'19"E GBIFCH00895417 OP113124 genseq-2 COI GBIFCH00895418 OP113125 genseq-2 COI GBIFCH00654969 OP113126 genseq-2 COI Guitna inf. 36°37'05"N, 08°20'47"E GBIFCH00975621 n/a n/a Centroptilum alamiae sp. nov.
Colouration (Fig. 3a,b). Head, thorax and abdomen dorsally brown, with dark grey-brown pattern as in Fig. 3a. Head and thorax ventrally brown, with dark greybrown lateral marks on thorax (Fig. 3b). Abdomen ventrally light brown. Legs light brown, apex of femur and claw darker. Caudalii ecru, brown annulated. Labrum (Fig. 1a). Rectangular, width ca. 1.6× maximum length. Distal margin with broad, angulated, medial emargination. Anterior margin nearly straight. Dorsal surface scattered with long, medium and short, simple setae; setae not arranged in a submarginal arc. Ventrally with marginal row of setae composed of anterolateral long, simple, pointed setae and medial long, apically blunt, pectinate setae; ventral surface with ca. seven short, stout setae near lateral and anterolateral margin.
Right mandible (Fig. 1b, c). Incisor and kinetodontium separated. Incisor with three denticles; kinetodontium with two denticles. Prostheca stick-like, distally with two denticles. Margin between prostheca and mola almost straight, with two tufts of long setae. Tuft of setae at apex of mola present.
Labium (Fig. 1i, j). Glossa nearly as broad and slightly shorter than paraglossa; inner and outer margins with many short, spine-like setae; apex with two medium, robust setae; dorsal surface with long, fine, simple, scattered setae. Paraglossa curved inward; ventrally with many long setae along outer lateral and apical margin, and row of long, stout, pointed, simple setae along inner lateral margin; dorsal surface with long, fine, simple, scattered setae. Labial palp 3-segmented. Segment III nearly trapezoidal with rounded distal corners, distal margin concave; outer lateral margin with short to medium, fine, simple setae, distal margin with short, spine-like and short, fine, simple setae; ventral surface with medium, fine, simple, scattered setae. Segment II with medium, fine, simple, scattered setae along outer lateral margin and on ventral surface; dorsally with 5-7 short, spine-like setae along distal margin. Segment I with medium, fine, simple setae scattered on ventral surface.
Sterna. Posterior margin of sterna I-VI smooth, without spines. Posterior margin of sterna VII-VIII with small, triangular spines.
Tergalii (Figs 2c-i, 3c). Present on segments I-VII. Costal margins with minute denticles and short, fine, simple setae, anal margins almost smooth. Tracheae extending from main trunk to inner and outer margins. Tergalius I as long as length of segments Figure 2. Centroptilum samraouii sp. nov., larva morphology a foreleg b fore claw c tergalius I d tergalius II e tergalius III f tergalius IV g tergalius V h tergalius VI i tergalius VII j paraproct k caudalii, spines on posterior margin of segments. II-IV combined; tergalius IV as long as length of segments V and VI combined; tergalius VII as long as length of segments VIII and IX combined. Paraproct (Fig. 2j). With 17-23 pointed marginal spines of different size, and some additional spines in second row. Cercotractor with minute, irregular, marginal spines.
Caudalii (Fig. 2k). Spines at posterior margins of segments elongated triangular with long points.
Subimago. Judging from subimaginal tarsomeres developing under cuticle of last instar female larvae, all tarsomeres of all legs of female subimago have pointed microlepids on surface (see Kluge 2022).
Imago. Unknown. Etymology. Dedicated to Prof. Boudjéma Samraoui, committed researcher on aquatic insects in Algeria, and collector of the new species; in recognition to his substantial contribution to the knowledge of the ecology and distribution of Algerian mayflies. Biological aspects. Centroptilum samraouii sp. nov. occupies the headwaters of steep, narrow and intermittent streams (Fig. 6c, d;Samraoui et al. 2021b, c). Distribution (Fig. 6e) Differential diagnosis to other species of Centroptilum. Larva. Following combination of characters: A) labrum with anterior margin slightly concave; ratio width vs. length ca. 1.5× (Fig. 7a); B) maxillary palp ca. 1.7× as long as galea-lacinia, segment III apically rounded; segment III ca. 1.6× as long as segment II (Fig. 7g); C) inner distal margin of labial palp segment III slightly concave (Fig. 7k); D) dorsal margin of fore femur with occasional short, spine-like setae; row of stout, pointed setae near margin (Fig. 8a); E) tarsus approx. as long as tibia (Fig. 8a); F) claw with two rows of denticles, each row with ca. 20 small to minute denticles (Fig. 8b); G) paraproct with 30-45 pointed spines, sometimes with split tips, few additional, submarginal spines (Fig. 8j).
Right mandible (Fig. 7b, c). Incisor and kinetodontium separated. Incisor with three denticles; kinetodontium with two denticles. Prostheca stick-like, distally with three denticles. Margin between prostheca and mola almost straight, with two tufts of long setae. Tuft of setae at apex of mola present. Left mandible (Fig. 7d, e). Incisor and kinetodontium separated. Incisor with four denticles; kinetodontium with three denticles. Prostheca stick-like, distolaterally denticulate. Margin between prostheca and mola straight, with large brush-like tuft of long setae. Subtriangular process short, on level of area between prostheca and mola. Tuft of setae at apex of mola absent.
Hypopharynx and superlinguae (Fig. 7f ). Lingua as long as superlinguae. Lingua longer than broad; distal half laterally not expanded; distal margin with short, fine setae, tuft of stout setae short. Superlinguae distally rounded; lateral margins rounded; fine, short to long, simple setae along distal margin. Maxilla (Fig. 7g). Galea-lacinia ventrally with four or five simple, apical setae under canines. Canines long and slender. With three denti-setae, distal denti-seta caninelike, middle and proximal denti-setae slender, bifid and pectinate. Medially with one pectinate, spine-like seta and three simple, spine-like setae (dorsolateral insertions); and ca. six long setae, partly with bifurcated tips (bifurcation often difficult to see; ventrolateral insertions). Maxillary palp 3-segmented, ca. 1.7× as long as length of galealacinia; palp segment III ca. 1.6× length of segment II; setae on maxillary palp fine, simple, scattered over surface of segments I, II, and III; apex of last segment rounded.  Labium (Fig. 7h-k). Glossa nearly as broad and slightly shorter than paraglossa; inner and outer margins with many short, spine-like setae; apex with two medium, robust setae; dorsal surface with long, fine, simple, scattered setae. Paraglossa curved inward; ventrally with many long setae along outer lateral and apical margin, and row of long, stout, pointed, simple setae along inner lateral margin; dorsal surface with long, fine, simple, scattered setae. Labial palp 3-segmented. Segment III nearly trapezoidal with rounded distal corners, distal margin slightly concave; outer lateral margin with short to medium, fine, simple setae, distal margin with short, spine-like and short, fine, simple setae; ventral surface with medium, fine, simple, scattered setae. Segment II with medium, fine, simple, scattered setae along outer lateral margin and on ventral surface; dorsally with seven or eight short, spine-like setae along distal margin. Segment I with medium, fine, simple setae scattered on ventral surface and on outer lateral margin.
Sterna. Posterior margin of sterna I-VI smooth, without spines. Posterior margin of sterna VII-VIII with small, triangular spines.
Tergalii 9d). Present on segments I-VII. Costal margins with minute denticles and short, fine, simple setae, anal margins almost smooth. Tracheae extending from main trunk to inner and outer margins. Tergalius I as long as length of segments II and III combined; tergalius IV as long as length of segments V and VI combined; tergalius VII as long as length of segments VIII and IX combined. Paraproct (Fig. 8j). With irregular row of 30-45 pointed marginal spines of different size, some with split tips, and few additional spines in second row. Cercotractor with minute, irregular, marginal spines.
Caudalii (Fig. 8k). Spines at posterior margins of segments short triangular, pointed. Subimago. Judging from subimaginal tarsomeres developing under cuticle of last instar female larvae, all tarsomeres of all legs of female subimago have pointed microlepids on surface (see Kluge 2022).
Imago. Unknown. Etymology. Dedicated to Prof. Majida El Alami, committed researcher on aquatic insects in Morocco, and collector of some of the specimens; in recognition of her substantial contribution to the knowledge of the systematics, ecology, and distribution of Moroccan mayflies.   Biological aspects. The specimens were collected in calm edge waters, loose substrate, low to moderate current, high temperatures, and sites rich in filamentous algae and mosses (Fig. 6a, El Alami et al. 2022a).

Genetics
The COI ingroup data set was 98% complete and included 34% of parsimony informative sites. The missing data almost exclusively resulted from nine GenBank sequences that lacked 5' and/or 3' end. All main CO1 gene tree relationships were resolved and well supported, except for the placement of the three clades Centroptilum sp. 1, C. sp. 2, and C. luteolum 1 (Fig. 10). The four sequences of C. samraouii sp. nov. were grouped in a well-supported monophyletic clade, supported as a distinct species in the ASAP, GMYC and mPTP species delimitation analyses (Fig. 10). Similarly, the two sequences of C. alamiae sp. nov. were grouped in a well-supported monophyletic clade, supported as a distinct species in all species delimitation analyses. The K2P mean genetic distance within the four C. samraouii sp. nov. and the two C. alamiae sp. nov. sequences were 0.08% and 0%, respectively. The K2P mean genetic distance between C. samraouii sp. nov. and the other six species (or putative species) ranged from 22.1% (mean distance to C. alamiae sp. nov.) to 25.2% (mean distance to C. sp. 1), whereas it ranged from 9.2% (mean distance to C. luteolum 1) to 25.7% (mean distance to C. volodymyri) for C. alamiae sp. nov. The three species delimitation methods were congruent, except for one slightly divergent sequence within the C. luteolum 1 cluster that was isolated by the GMYC, and the three C. volodymyri sequences that were all considered as distinct putative species according to ASAP and GMYC.

Differentiating characters between species of Centroptilum
The characters differentiating the geographically relatively close species Centroptilum luteolum, C. samraouii sp. nov. and C. alamiae sp. nov. are summarised in Table  2. Most important are the spines on posterior margin of abdominal terga and the spines on paraproct margin (see Table 2). Further reliable characters to differentiate both new species from North Africa are the distal margin of the labrum (straight in C. samraouii sp. nov., slightly concave in C. alamiae sp. nov.); the distal margin of labial palp segment III (concave in C. samraouii sp. nov., slightly concave in C. alamiae sp. nov.); the relative length of maxillary palp segment III vs. segment II (1.3× in C. samraouii sp. nov.,1.6× in C. alamiae sp. nov.); and the setation on dorsal margin of femur (only occasional setae in C. samraouii sp. nov., additional row of short, pointed setae near margin in C. alamiae sp. nov.) (see Table 2).
The recently described species C. volodymyri (Georgia, Turkey, Iran) differs from C. samraouii sp. nov. and C. alamiae sp. nov. by several distinct characters: maxillary palp much Figure 10. Bayesian majority-rule consensus tree reconstructed from the CO1 data set. Coloured vertical boxes indicate species delimitation hypothesis according to the ASAP, GMYC and mPTP methods. Tips labelled with GBIF codes indicate newly sequenced specimens, CH007_SR codes designate sequences from the FREDIE project, and other codes correspond to previously published GenBank sequences. For each mPTP species hypothesis, the corresponding species names (where available) and the country of origin is provided. Circles on branches indicate Bayesian posterior probabilities > 0.95. Outgroup branches, tips labels, and species names are presented in grey. longer than galea-lacinia (ca. 2.3×); maxillary palp segment I distinctly wider than segment II (only slightly wider in all other species); labrum much wider than long (1.8-2.0×); claw with more than 60 minute denticles in two rows (ca. 30 per row) (Martynov et al. 2022; for respective character states of C. samraouii sp. nov. and C. alamiae sp. nov. see Table 2). The poorly known species C. pirinense (Pirin Mountains, Bulgaria) differs from C. samraouii sp. nov. and C. alamiae sp. nov. at least in the very wide labrum (ca. 2.0× wider than long; Martynov et al. 2022: table II), whereas in C. samraouii sp. nov. it is ca. 1.6× and in C. alamiae sp. nov. ca. 1.5× (see Table 2).

Microlepids of subimago
Judging from tarsomeres of subimagos developing under cuticle of female last instar larvae, at least female subimagos of both new species of Centroptilum have all their tarsomeres of all legs covered with pointed microlepids. This is in line with C. luteolum, which has pointed microlepids on all tarsomeres of all legs of male and female subimagos (Kluge 2022).

Genetics and biogeography
The two new North African species described here are highly supported by our CO1based analyses. First, the minimum mean genetic distance of 9.2% (mean distance between Centroptilum alamiae sp. nov. to C. luteolum 1) is much higher than the generally accepted intra-/interspecific threshold value of ca. 3% divergence for mayflies (e.g., Ball et al. 2005;Kjaerstad et al. 2012;Gattolliat et al. 2015). Second, both new species are well supported in their own monophyletic clade, and third, all three species delimitation analyses are congruent and support the morphological results. Interestingly, the two new species are not supported as closely related, despite their geographical proximity, suggesting a distinct origin. Rather, C. alamiae sp. nov., and the European species C. sp. 1, C. sp. 2, and C. luteolum 1 are included in the same well-supported clade sister to the others, which possibly indicates a more recent colonisation event from Europe to Morocco. This hypothesis is supported by the presence of C. luteolum 1 in the Pyrenees and in the south of Spain (unpublished sequences from the project FREDIE; not shown in Fig. 10). The type locality of C. alamiae sp. nov. in Morocco is geographically closer to the south of Spain than to the type locality of C. samraouii sp. nov. in Algeria. All examined specimens of Centroptilum in Morocco and Algeria belong to one of the new species and not to C. luteolum or any other species of Centroptilum. The genus Centroptilum seems to be extremely rare in Tunisia, no specimen from this country could be investigated in this study. In conclusion, we cannot formally exclude the presence of C. luteolum in the Maghreb at this point in time, but it seems unlikely.