The Old World sparrows, Passeridae, are a speciose passerine family distributed all over the Afrotropics, the Palearctic and parts of the Oriental Region. Throughout the entire Old World, only the Australian Region and Madagascar are not inhabited by any species of the family – except the human-introduced house sparrow. Several species are highly adapted to extreme environments such as the snowfinches (Montifringilla, Pyrgilauda and Onychostruthus) from the high alpine ecosystems of Eurasian mountain systems (Lei et al. 2014; Päckert et al. 2020). Recent comparison of high-quality genomes provided evidence of divergent adaptation to local selective pressures in each of the three snowfinch genera (Qu et al. 2021). Also, the extremely hot and dry Sahara harbors suitable habitat for specialists like the Desert sparrow, Passer simplex. Areas of highest species richness are located in the African Rift Valley and at the eastern margin of the Qinghai-Tibet Plateau (QTP) (Fig. 1).

Though formerly included in Passeridae (e.g. Dickinson 2003), the sparrow-weavers (genera Plocepasser, Histurgops, Pseudonigrita and Philetarius) had often been affiliated to the Ploceidae based on morphological features like tongue musculature (Bock and Morony Jr 1978; Summers-Smith 2010). Recent phylogenies by de Silva et al. (2017, 2019) confirmed the inclusion of sparrow-weavers in Ploceidae (compare also Jønsson and Fjeldså 2006) in accordance with most taxonomic authorities (Dickinson and Christidis 2014; del Hoyo and Collar 2016; Clements et al. 2019; Gill et al. 2020). The Passeridae are characterized by several synapomorphies of tongue morphology, too (Bock and Morony Jr 1978) and representatives of major genera (Montifringilla, Passer and Petronia) belong to a monophyletic group that was consistent across several recently published phylogenies (e.g. Ericson and Johansson 2003; Zuccon et al. 2012).

To date, the Passeridae are generally classified into eight genera, four of them monotypic (Hypocryptadius, Carpospiza, Petronia and Onychostruthus), with a total number of 43 currently accepted species (according to the IOC World Bird List by Gill et al. 2020). Among these, Passer is the most diverse genus with 28 currently recognized species (del Hoyo and Collar 2016; Gill et al. 2020), of which the house sparrow, Passer domesticus (Fig. 2C), is probably one of the best studied species (reviews in Anderson 2006; Liebl et al. 2015), not least because as a commensal of human civilization it is fairly common all over its range (Sætre et al. 2012). Moreover, past and extant hybridization of the house sparrow with other conspecifics has been intensively studied on a genetic basis with respect to the stabilized hybrid form Passer italiae (Elgvin et al. 2011, 2017; Hermansen et al. 2011, 2014; Eroukhmanoff 2013, 2017; Sætre et al. 2017; Runemark et al. 2018), to distinct genetic lineages in Asia (Ravinet et al. 2018) and to the mosaic hybrid zone with the Spanish sparrow, P. hispaniolensis, in North Africa (Belkacem et al. 2016; Päckert et al. 2019).

In contrast, the phylogenetic relationships among genera and species of Passeridae are poorly studied to date, which is mainly due to a lack of data from the Afrotropics. Recently, it came out as a rather surprising finding, that the Philippine endemic cinnamon ibon, Hypocryptadius cinnamomeus, was sister to a clade of Passeridae species (Fjeldså et al. 2010). Previously, that Philippine endemic had been included in the white-eyes (Zosteropidae), however based on molecular phylogenetic evidence this species is included in Passeridae by several taxonomic authorities today (del Hoyo and Collar 2016; Gill et al. 2020).

A first phylogenetic hypothesis for Passeridae was based on a single mitochondrial gene (Allende et al. 2001) and included only eleven species. Since then, a few molecular studies focused on the phylogenetic relationships of snowfinches (Onychostruthus, Pyrgilauda, Montifringilla [Fig. 2A]), a group of eight high alpine endemic species from the QTP and from other Palearctic mountain systems (Qu et al. 2006; Gebauer et al. 2006; Lei et al. 2014; del Mar Delgado et al. 2019; Päckert et al. 2020). However, to date no multi-locus analysis has ever been performed for a broader taxon sampling across different genera of Passeridae. The most comprehensive phylogenetic hypothesis available for Passeridae by Jønsson and Fjeldså (2006; their Passeroidea clade 8) included 14 species from three genera.

As a contribution to the current discussion on phylogenetic relationships within Passeridae, we provide a new phylogenetic hypothesis for 18 species of Old World sparrows (Passer) and another 11 species of African bush-sparrows (Gymnoris), rock sparrows (Petronia, [Fig. 2B]) and snowfinches (Onychostruthus, Pyrgilauda, Montifringilla) from the Qinghai-Tibet Plateau and other Palearctic mountain systems.

We amplified and sequenced four molecular markers using 65 samples from 22 species of the Passeridae genera Passer, Petronia, Gymnoris, Montifringilla, Pyrgilauda and Onychostruthus. Based on previous evidence of intraspecific diversification from Päckert et al. (2020) we included some additional subspecific taxa of Montifringilla nivalis and Petronia petronia (Table 1), further samples from different island populations of the Cape Verde endemic Iago sparrow (Passer iagoensis; Fig. 2D) and from the range of overlap of two of the smaller snowfinch species (Pyrgilauda blanfordi and P. davidiana; further samples for intraspecific comparison, see supplementary Table S1).

We extracted DNA from frozen blood or tissue samples using the innuPREP DNA Mini Kit (for muscle tissue) or the innuPREP BloodDNA Mini Kit (for blood), respectively (both Analytik Jena AG, Germany) according to the manufacturer’s instructions except for overnight incubation of tissue with proteinase K (instead of one hour).

We amplified and sequenced the mitochondrial cytochrome-b (cyt-b) for all samples available for comparison with the Passer phylogeny by Allende et al. (2001). For multi-locus reconstruction we sequenced one further mitochondrial gene, NADH-dehydrogenase subunit2 (ND2) and two nuclear introns, myoglobin-intron2 (myo) and ornithine-decarboxylase intron7 (ODC). Primers and PCR protocols are documented in Päckert et al. (2020). PCR products were purified using ExoSap-IT (GE Healthcare; adding 0.1 mL ExoSap-IT solution in 4 mL H2O to each sample; 37 °C for 30 min, 94 °C for 15 min). The sequencing of the PCR products was performed with BigDyeTM 3.1 Dye Terminator Cycle Sequencing Kits (Applied Biosystems), according to the manufacturers’ instructions. Cycle sequencing products were purified by salt/ethanol precipitation or by using Sephadex (GE Healthcare, Munich, Germany), and sequenced in both directions on an ABI 3130xl DNA sequencer.

We aligned forward and reverse Sanger sequences for each gene by ClustalW using MEGA 5.1 (Tamura et al. 2011) and we cross-checked the respective electropherograms with Chromas v.2.6.5 (Technelysium Pty Ltd) for possible inaccuracies due to sequencing or reading errors. For each marker per sample, we manually combined sequences of both reading directions to a single consensus sequence. All sequences used for analysis were deposited at GenBank (Table 1).

Newly generated sequences were incorporated in a sequence alignment for Passeroidea from Päckert et al. (2016, 2020), including outgroup taxa from closely related families Ploceidae, Viduidae, Estrildidae and Urocynchramidae (Table 1). The final alignment comprised 3485 base pairs (cyt-b: 1041 bp; ND2: 1041 bp; myo: 732 bp; ODC: 671 bp). We complemented our sequence data set for Passeridae with sequence data from GenBank for eight species missing from our sampling including the cinnamon ibon, Hypocryptadius cinnamomeus (Table 1). Altogether, our final data set comprised 30 species of Passeridae among these 18 out of 28 currently recognized species from genus Passer (del Hoyo and Collar 2016). These are more than two third of all species from this genus (see Table 1) and twice as many species-level taxa compared to the most recent phylogenetic hypothesis for Passeridae (Jønsson and Fjeldså 2006). For hierarchical outgroup rooting we used the waxwing, Bombycilla garrulus (compare Päckert et al. 2020).

We reconstructed multi-locus phylogenies using Bayesian inference of phylogeny BEAST vers. 1.8.1 (Drummond et al. 2012) and Maximum Likelihood (ML) using RAXML (Stamakakis 2006, 2014). We relied on the partitioning scheme applied to the Passeroidea data set by Päckert et al. (2020) who included Passeridae with 26 species. According to their estimates using PARTITIONFINDER (Lanfear et al. 2012) the best-fit partition scheme was a nine-partition scheme by gene and codon: ND2, 1041 bp, three partitions by codon position, GTR +Γ+I model; cytochrome-b, 1041 bp, three partitions by codon position, GTR +Γ+I model; myo, 730 bp, one partition, HKY+Γ model; ODC, 643 bp, one partition, GTR+Γ model.

For inference of divergence times estimates, we applied a molecular clock calibration using mean substitution rate estimates for the two mtDNA markers estimated by Lerner et al. (2011) for Hawaiian honeycreepers (Drepanidinae): cyt-b= 0.014; ND2= 0.029 (both in in substitutions per site per lineage per million years). The cyt-b rate applied here ranges at a similar dimension like the empirical cyt-b rate of 0.0105 evaluated by Weir and Schluter (2008).

We performed three independent runs with BEAST for 30,000,000 generations (parameters were logged and trees sampled every 3,000 generations) under the uncorrelated lognormal clock model for all loci with the “auto-optimize” option activated and a birth-death process prior applied to the tree. We combined log files and tree files from independent BEAST runs with used LOGCOMBINER v.1.8.1 and checked the combined log file in TRACER v. 1.4 (Rambaut and Drummond 2007) to ensure adequate ESS files for all parameters (all ESS > 200). All obtained phylograms were edited in FIGTREE vers. 1.4.2 (Rambaut 2009).

For illustration of intra- and interspecific genetic variation and divergence of selected species, we reconstructed unrooted minimum parsimony networks with PopART (http://popart.otago.ac.nz) using the “tcs network” algorithm (Clement et al. 2000). We calculated uncorrected pairwise p-distances (based on cytochrome-b sequences) using MEGA 5.1.

The Old World sparrows resulted as a strongly supported monophyletic group from all analyses and were sister to another well supported clade including weavers (Ploceidae), Przewalski’s finch (Urocynchramus pylzowi), estrildid finches and wydahs (Estrildidae and Viduidae; Fig. 3). The subclade of Passeridae is shown in Fig. 4. The basal split in Old World sparrows was dated to approximately 17.5 mya and separated the cinnamon ibon (Hypocryptadius cinnamomeus) from all other Passeridae. These were divided into two major clades.

Clade I showed a deep split at about 10 mya between the rock sparrows (genus Petronia; clade Ia) and the snowfinches (Montifringilla, Pyrgilauda, Onychostruthus; clade Ib, Fig. 4). The latter three snowfinch genera started diversifying at about 6.9 mya, a sister-group relationship between Montifringilla and Pyrgilauda (with Onychostruthus as the earliest offshoot) received only poor support (Fig. 4; clade Ib). However, the clade uniting Montifringilla and Pyrgilauda was characterized by a shared 3-bp deletion in myoglobin intron 2, whereas the Pyrgilauda clade was characterized by another 4-bp insertion in the same intron marker (Fig. 4).

Clade II includes the sister genera Passer (IIa) and Gymnoris (IIb) and each of them with strong node support. Members of Clade II shared a 23-bp deletion in ODC intron 7 (Fig. 4) and all members of Passer shared another 4-bp deletion in myo intron 2 (Fig. 4). Contrary to traditional systematic classification, the yellow-throated species of bush sparrows (Gymnoris) and rock sparrows (Petronia) were not closest relatives in the Passeridae phylogeny (Fig. 4).

Old World sparrows of genus Passer were also divided into two strongly supported clades. One entirely Afrotropical clade included seven species from Sub-Saharan Africa. One Sub-Saharan subclade included three species of grey-headed sparrows (Fig. 4: P. griseus, P. diffusus and P. gongoensis). Except for that monophyletic group of grey-headed sparrows, head color pattern does not reflect monophyletic units in the Passer clade (Fig. 4), which is once more in contrast to previous superspecific classifications. The second subclade united two species from South Africa (P. melanurus, P. motitensis) with the small-sized chestnut sparrow (P. eminibey) from East Africa. The Afrotropical Passer clade was sister to a second moderately supported clade that comprised 12 species from the Palearctic and the Oriental Region that started diversifying at about 5.5 mya (Fig. 4). Phylogenetic relationships among members of that clade were ambiguous because of poor support values for many nodes. A basal split separated the Asian russet sparrow (P. cinnamomeus) from the remaining Passer species. Two further ancient offshoots of the Palearctic/Oriental clade, the widespread tree sparrow (P. montanus) and the Central Asian Saxaul sparrow (P. ammodendri) received moderate and poor support, respectively. Another poorly supported Afro-Arabian clade of four sparrow species united the Saharan desert sparrow (P. simplex), the Sudan golden sparrow (P. luteus) from the Sahel Region, the Dead Sea sparrow (P. moabiticus) from the Near East and the Middle East and the Cape Verde endemic Iago sparrow (P. iagoensis)(Fig. 4). In the latter, no clear phylogeographic structure among island populations could be observed in the maximum parsimony network of five cyt-b haplotypes (Fig. 5C). Finally, a well-supported terminal clade united four closely related species that started diversifying in the early Pleistocene: the house sparrow (P. domesticus), the Spanish sparrow (P. hispaniolensis), the Italian sparrow (P. italiae) and the Socotra sparrow (P. insularis) (Fig. 4). The sister-group relationship of the Southeast Asian plain-backed sparrow (P. flaveolus) to that terminal clade received moderate support.

High intraspecific differentiation with split ages estimated at 2.3–2.7 Ma was found in two species of clade I: Both Petronia petronia and Montifringilla nivalis showed a deep split between European and Asian lineages (Fig. 4). Paraphyly of M. nivalis with respect to its Tibetan congener M. adamsi was only poorly supported. European and Asian populations of the white-winged snowfinch (M. nivalis) appeared as two distinct clusters in the cyt-b haplotype network separated by a minimum of 22 substitutions (Fig. 5A). Uncorrected pairwise distances between the European and the Asian mitochondrial lineage ranged between 4.9–5.1% (cyt-b) at the same p-distance level like interspecific comparison between M. nivalis and M. adamsi (4.5–4.8%; cyt-b). Similarly, uncorrected p-distances between rock sparrow populations from Spain (P. p. petronia) and from China (P. p. brevirostris) were as high as 4.8% (cyt-b; compare the deep split in Fig. 4). In the small-sized species of genus Pyrgilauda, one specimen of Père David’s snowfinch, P. davidiana, was sister to a syntopic P. blanfordi specimen instead to a conspecific specimen from the northern allopatric part of the breeding range (Fig. 4; however this grouping was not supported in the RAXML tree that united both P. davidiana sequences in a poorly supported clade). The haplotype network for a larger set of Pyrgilauda samples showed that regardless of phenotypic species identification all specimens from the region of sympatry at Koko Nor in northern Qinghai belonged to one haplotype cluster that was separated from another distantly related P. davidiana haplotype (shared by two specimens of unknown origin) by 37 substitutions (Fig. 5B). The Koko Nor cluster had a star-like structure with eight tip haplotypes and a central haplotype shared by eleven individuals of both species (P. davidiana and P. blanfordi).

To date, there is no comprehensive phylogeny of Old World sparrows (Passeridae) available except for a single-locus tree covering about 40% of the currently accepted species (Allende et al. 2001) and a Passeridae clade from a supertree by Jønsson and Fjeldså (2006) which was largely based on the same sequence information (see below). Though our new phylogeny still misses ten out of 28 Passer species we are covering 30 of 43 currently accepted species of Passeridae (about 70%) and several clear conclusions can be drawn from this new (though still incomplete) phylogenetic hypothesis. Most importantly, our results confirm the monophyly of the genera Gymnoris, Passer, Montifringilla and Pyrgilauda (the remaining genera are monotypic). This is particularly relevant with respect to the taxonomic treatment of bush sparrows and rock sparrows.

Despite from being far from taxon-complete, this updated phylogeny contributed further evidence for clarification of taxonomic controversy, e.g. the status of Petronia and Gymnoris as separate genera, the monophyly of grey-headed sparrows (but not of all grey-crested Passer species) or a lack of phylogenetic justification for recognizing the genera Sorella and Auripasser. We failed to include the enigmatic pale rock sparrow, Carpospiza brachydactyla, from the Middle East and Central Asia that was long regarded as a member of Fringillidae. Based on shared traits of tongue morphology inclusion in Passeridae was recommended by Bock (2004), generally Carpospiza has been affiliated with Petronia sensu lato (including Gymnoris), however, it lacks the yellow throat patch (being a rather uninformative trait as shown in our phylogeny). Also, phylogenetic relationships of missing Passer species from India (P. pyrrhonotus), Central Asia (P. zarudny), the Socotra archipelago (P. hemileucus) and East Africa (P. chordofanicus, P. euchlorus, P. rufocinctus, P. shelleyi, P. suahelicus, P. swainsoni, P. castanopterus) will remain unresolved so far. Recently, the narrow-range endemic Somali sparrow, P. castanopterus, has attracted ornithologists’ attention for its putative hybridization with the house sparrow, P. domesticus, in the areas of range overlap in Somalia (Summers-Smith 2020), Ethiopia (Gedeon et al. 2015), Kenya (Turner 2016) and Djibouti (Cohen et al. 2011; Hering et al. 2020).