The Bactrocera dorsalis complex (Tephritidae) comprises 85 species of fruit flies, including five highly destructive polyphagous fruit pests. Despite significant work on a few key pest species within the complex, little has been published on the majority of non-economic species in the complex, other than basic descriptions and illustrations of single specimens regarded as typical representatives. To elucidate the species relationships within the B. dorsalis complex, we used 159 sequences from one mitochondrial (COI) and two nuclear (elongation factor-1α and period) genes to construct a phylogeny containing 20 described species from within the complex, four additional species that may be new to science, and 26 other species from Bactrocera and its sister genus Dacus. The resulting concatenated phylogeny revealed that most of the species placed in the complex appear to be unrelated, emerging across numerous clades. This suggests that they were placed in the B. dorsalis complex based on the similarity of convergent characters, which does not appear to be diagnostic. Variations in scutum and abdomen color patterns within each of the non-economic species are presented and demonstrate that distantly-related, cryptic species overlap greatly in traditional morphological color patterns used to separate them in keys. Some of these species may not be distinguishable with confidence by means other than DNA data.

Bactrocera , dorsalis , intraspecific variation, phylogenetics

Most of the Dacine fruit flies (Tephritidae: Dacini) are in the genera Bactrocera (651 described species) and Dacus (270 species), with many species (73 Bactrocera and 11 Dacus) bred from commercial/edible fruit and fleshy vegetables (Vargas et al. 2015). Species of Bactrocera thrive in the endemic rainforest habitats of South-East Asia and Australasia, with a high degree of host specialization and a large number of cryptic species (Drew and Hancock 2000, Drew 2004).

Among the pest species, Bactrocera dorsalis (Hendel) (= the Oriental fruit fly) is the most destructive and polyphagous species (Vargas et al. 2015), belonging to a large complex of similar-looking species: the Bactrocera dorsalis complex (hereafter referred to as the OFF complex). The first reference to the OFF complex was by Hardy (1969), who recognized and provided a key to B. dorsalis and 15 other non-economic species. Subsequently, Drew and Hancock (1994) revised the group from South-East Asia, describing 40 new species and splitting B. dorsalis into four distinct species, resulting in a total of 52 species, plus 16 species in Australasia (Drew 1989). Among the combination of character states defining the complex, they included a mostly black scutum and abdomen terga III–V with a medial longitudinal band forming a “T-shaped” pattern with the transverse band at base of tergum III, and with variable dark patterns on lateral margins of terga III–V. Currently, 85 species are recognized, taking into account the recent revision (Drew and Romig 2013) and synonymization (Schutze et al. 2014). Six of the species (B. carambolae Drew & Hancock, B. caryeae (Kapoor), B. dorsalis, B. kandiensis Drew & Hancock, B. occipitalis (Bezzi), and B. trivialis (Drew)) in the complex are significant pests of cultivated fruit (Vargas et al. 2015).

While literature abounds on the taxonomy, genetic diversity, biology and management of the economic species (Clarke et al. 2005, Schutze et al. 2014a), very little is known about most of the other species in the OFF complex, other than basic taxonomic descriptions. Identification to species level is challenging for many species, due to uniform appearance and extensive intraspecific morphological variation. Morphological diagnostic tools were developed for the economic species, based of wing morphometrics (Schutze et al. 2012) and ovipositor and aedeagus lengths (Iwaizumi et al. 1997, Drew et al. 2008, Krosch et al. 2013, White 2000). Some of the species, especially B. dorsalis, display a broad range of color patterns and length of aedeagus and ovipositor, that have resulted in the description of geographic variants as new species, which were subsequently argued to be conspecific (Schutze et al. 2012, 2014b), and synonymized (Schutze et al. 2014a). The range of color variation in the scutum and abdomen was characterized to some extent for B. dorsalis and B. carambolae (e.g. Nishida and Vargas 1992, Iwahashi 1999, Drew et al. 2005, Leblanc et al. 2013), but no information has been published for the other 83 species.

Species descriptions and illustrations in published monographs (Drew 1989, Drew and Hancock 1994, Drew and Romig 2013) are based on the most commonly encountered morphological variants, and little information is presented on intraspecific variation. The dichotomous key in Drew and Hancock (1994) is based on these most common variants, hence difficult to use to identify more atypical specimens. An attempt to account for variation in an interactive CD-ROM key (Lawson et al. 2003) yielded limited success (Clarke et al. 2005). In addition to the described species, there may likely exist cryptic species, hard to distinguish by morphological means, which can be separated with the help of genetic sequencing (e.g. Carew et al. 2011, Dujardin and Kitthawee 2013).

Clarke et al. (2005), when reviewing the data available at the time, stated that phylogenetic studies using limited taxa and genes may not demonstrate the monophyly of the complex. However, recent molecular phylogenies which include the OFF complex have found that most species form a well-defined monophyletic clade (Krosch et al. 2012, Virgilio et al. 2015). However, these studies only included methyl eugenol-attracted species and were limited to six economic species and six, mainly Australian, non-pest species such as B. cacuminata (Hering) and B. opiliae (Drew and Hardy). An alternate, polyphyletic complex was indicated by a phylogeny based on one mitochondrial and two nuclear genes by San Jose et al. (2013), but sampling was limited.

Our goal was to examine the B. dorsalis species complex more broadly than the few frequently targeted pest species. This is accomplished by reporting and analyzing novel molecular and morphological data on 22 non‐pest species in the complex, in the context of the main pest species and selected outgroups. These data are used to: (i) determine through phylogenetic analysis if the complex is monophyletic or polyphyletic; (ii) provide diagnostic molecular data for over 25 species for which such data is currently lacking; and (iii) determine the utility of thoracic and abdominal color/pattern variation as species level diagnostic characters.

Sequences listed on Table 1, as well as COI sequences for all specimens included on all figure plates, were deposited into GenBank KT591129 to KT591164 and KT594783 to KT595006.

Topological differences between the individual gene trees were not supported with high bootstrap values and posterior probabilities (<50% BS <0.9 PP) and overall individual gene trees were poorly resolved, with COI providing more signal for the more recent divergences (Suppl. material 1) and the nuclear genes providing signal for deeper relationships (Suppl. material 2, 3). However the concatenated analysis produced a well-resolved tree (Figure 1) which is consistent with previous studies (Krosch et al. 2012, San Jose et al. 2013, Virgilio et al. 2015). In the concatenated phylogeny, the Zeugodacus group of subgenera (as defined by Drew and Hancock 2000) is sister to Dacus and the Bactrocera+Notodacus+Daculus clades, which themselves are sister taxa. This renders Bactrocera paraphyletic with respect to Dacus, as suggested previously (White 2006, Krosch et al. 2012, Virgilio et al. 2015). However the relationship is not strongly supported in the tree and additional genes and taxa are necessary to fully resolve this relationship. The subgenus Bactrocera is monophyletic in the concatenated phylogeny (100 BS, 1.0 PP). The inclusion of many non-economic OFF complex species in our study shows with high support that despite a similar appearance, the complex is a highly polyphyletic group. Multiple, well-supported, clades (75–100% BS values) in the subgenus Bactrocera contain a mix of species previously thought to belong to the OFF complex and non-OFF complex species. One clear example is the inclusion of non-OFF complex B. bryoniae, B. latifrons, B. limbifera, with B. kohkongiae, which fits in the OFF complex (Figure 15 A–C) in a strongly supported (100%, 100% PP value) clade (Figure 1). This indicates that, despite low support for the backbone topology in the subgenus Bactrocera, the polyphyletic nature of the OFF complex is still well supported. The main pest species in the complex (B. carambolae and B. dorsalis, now including B. papayae and B. invadens, see Schutze et al. 2014a) form a monophyletic unit with very little genetic differentiation (<1.3% in COI) between them, and rest within a well defined clade that includes several other species attracted to methyl eugenol (B. occipitalis, B. cacuminata, B. raiensis). Three species, B. melastomatos (Figure 9F–O), B. osbeckiae (Figure 10) and B. rubigina (Figure 14F), were genetically indistinguishable using COI (0.1% pair-wise difference) in the phylogeny, appearing together in a single lineage, despite having very distinctive color patterns. Interestingly, they were slightly more distinct in the nuclear genes (1.1% EF-1α and 1% period pair-wise difference), which was not the case for most species. Population genetic statistics, based on COI sequences, showed high levels of haplotype diversity for most of the non-economic species in the B. dorsalis complex (Table 2).

Figure 1. 

Maximum likelihood tree, concatenated, based three gene (COI, period, EF-1α) dataset. Support values above branches are Maximum Likelihood Bootstrap values / Bayesian Posterior Probabilities. Scale bar indicates the number of substitutions per site. Species in the Oriental fruit fly complex are outlined in red.

Color patterns of scutum and/or abdomen (Figures 2–15) varied extensively within some of the species (Figures 2A–J, 3A–I, 4A–J, 5D–H, 6K–O, 7, 8, 9F–O, 10, 11B–E, 12A–J, 13F–O), and were relatively uniform in others (Figures 2K–O, 3J–M, 4K–M, 5A–C, 6A–J, 9A–E, 11A, 12K–O, 13A–E). Scutum color pattern was highly polymorphic in B. bhutaniae (Figure 2), B. bivittata (Figure 3), B. kohkongiae (Figure 7), B. melastomatos (Figure 9F–O), B. osbeckiae (Figure 10), and B. propinqua (Figure 12). Abdomen pattern was confusingly polymorphic, yet scutum remained uniform in B. thailandica (Figure 13).

Figure 2. 

Variation in color pattern of scutum and abdomen in Bactrocera bhutaniae Drew and Romig (321 specimens examined and 36 sequenced). Voucher codes are: A ms3593 B ms3531 C ms3533 D ms4321 E ms2034 F ms1166 G ms2031 H ms3527 I ms3580 J ms1168 K ms2030 L ms3578 M ms4329 N ms3527 O ms1168.

Figure 3. 

Variation in color pattern of scutum and abdomen in Bactrocera bivittata Li and Wang (47 specimens examined and 10 sequenced). Voucher codes are: A ms3606 B ms1305 C ms1304 D ms3607 E ms3605 F ms3604 G ms3609 H ms3608 I ms1790 J ms3605 K ms3606 L ms3609 M ms3604.

Figure 4. 

Variation in color pattern of scutum and abdomen in Bactrocera cacuminata (Hering) (> 300 specimens examined and 12 sequenced). Voucher codes are: A ms2003 B ms2005 C ms1998 D ms2008 E ms1999 F ms1997 G ms2010 H ms2004 I ms2009 J ms2002 K ms2005 L ms2008 M ms2009.

Figure 5. 

Variation in color pattern of scutum and abdomen in Bactrocera fuscitibia (Drew and Hancock) (33 specimens examined and 6 sequenced). Voucher codes are: A ms1178 B ms1177 C ms1297 D ms1175 E ms1176 F ms1177 G ms1178 H ms1297.

Figure 6. 

Variation in color pattern of scutum and abdomen in Bactrocera kanchanaburi Drew and Hancock (47 specimens examined and 16 sequenced). Voucher codes are: A ms3599 B ms1300 C ms3598 D ms1303 E ms3725 F ms1302 G ms3597 H ms3596 I ms3728 J ms3603 K ms3599 L ms3728 M ms1300 N ms1301 O ms3729.

Figure 7. 

Variation in color pattern of scutum in Bactrocera kohkongiae Leblanc (210 specimens examined and 22 sequenced). Voucher codes are: A ms1149 B ms1144 C ms1142 D ms1780 E ms1148 F ms1145 G ms1307 H ms1143 I ms1141 J ms1146 K ms1151 L ms1785 M ms1140 N ms1781 O ms1150.

Figure 8. 

Variation in color pattern of abdomen in Bactrocera kohkongiae Leblanc. Voucher codes are: A ms1149 B ms1147 C ms1145 D ms1785 E ms1146 F ms1139 G ms1137.

Figure 9. 

Variation in color pattern of scutum and abdomen in Bactrocera latilineola Drew and Hancock (A–E) (11 specimens examined and 4 sequenced) and B. melastomatos Drew and Hancock (F–O) (46 specimens examined and 8 sequenced). Voucher codes are: A ms1114 B ms2025 C ms2025 D ms2024 E ms1299 F ms1415 G ms1416 H ms1410 I ms1412 J ms1411 K ms1416 L ms1417 M ms1413 N ms1410 O ms1411.

Figure 10. 

Variation in color pattern of scutum and abdomen in Bactrocera osbeckiae Drew and Romig (100 specimens examined and 39 sequenced). Voucher codes are: A ms1161 B ms3559 C ms3558 D ms3553 E ms3555 F ms3561 G ms1163 H ms3785 I ms3764 J ms3768 K ms1153 L ms3758 M ms3554 N ms1180 O ms1138 P ms3555 Q ms3784 R ms3560 S ms1154 T ms3768.

Figure 11. 

Variation in color pattern of scutum and abdomen in Bactrocera paraarecae Drew and Romig (10 specimens examined and 5 sequenced). Voucher codes are: A ms1295 B ms1296 C ms2040 D ms1294 E ms1110.

Figure 12. 

Variation in color pattern of scutum and abdomen in Bactrocera propinqua (Hardy and Adachi) (49 specimens examined and 24 sequenced). Voucher codes are: A ms4324 B ms4331 C ms4322 D ms3568 E ms3571 F ms3833 G ms3572 H ms3567 I ms2041 J ms1170 K ms4331 L ms3765 M ms3757 N ms3572 O ms3566.

Figure 13. 

Variation in color pattern of scutum in Bactrocera thailandica Drew and Romig (712 specimens and 56 sequenced). Voucher codes are: A ms3587 B ms3588 C ms3586 D ms3525 E ms1952 F ms3576 G ms3586 H ms3736 I ms3585 J ms3539 K ms3581 L ms3538 M ms3695 N ms3582 O ms1949.

Figure 14. 

Scutum and abdomen of: A Bactrocera dongnaiae Drew and Romig (ms1158; 7 specimens examined and 3 sequenced) B B. laithieuiae Drew and Romig (ms3762; 1 specimen examined and sequenced) C B. lombokensis Drew and Hancock (ms1548; 1 specimen examined and sequenced) D B. quasiinfulata Drew and Romig (ms3455; 4 specimens examined and sequenced) E B. raiensis Drew and Hancock (ms1331; 2 specimens examined and 1 sequenced) F B. rubigina (Wang and Zhao) (ms3543; 259 specimens examined and 27 sequenced) G B. usitata Drew and Hancock (ms2039; 27 specimens examined and 6 sequenced).

Figure 15. 

Scutum, abdomen and wing costal region of: A–C Bactrocera species 54 (ms1798 (wing, scutum), ms3777 (abdomen); 7 specimens examined and sequenced) D–F B. species 55 (ms3575; 7 specimens examined and sequenced) G–I B. species 59 (ms1164; 1 specimen examined and sequenced) J–L B. species 60 (ms3730; 3 specimens examined and sequenced).

Scutum color and variation followed three basic patterns among species for which series of specimens were examined. In B. bhutaniae (Figure 2), B. bivittata (Figure 3), B. kohkongiae (Figure 7), and B. osbeckiae (Figure 10), scutum was predominantly red–brown with a highly variable dark lanceolate pattern. The pattern was composed of a medial and two lateral bands, generally interrupted at the level of the transverse suture, in B. bhutaniae and B. bivittata (medial band usually narrower and lateral bands very broad). The lanceolate pattern was highly variable in B. kohkongiae, from extensively pale with a narrow medial band to almost entirely dark with light markings restricted to the transverse suture, and B. osbeckiae, from mostly dark fuscous, with red-brown markings at level of postpronotal lobes and along transverse suture, to extensive lanceolate red–brown pattern with a broad medial longitudinal band, which can be faint or absent. In B. cacuminata (Figure 4), scutum was red–brown with a single medial dark band widened at apex of scutum and anteriorly narrowed to a point, and with two short lateral bands pointed anteriorly. A similar pattern was frequently observed in B. propinqua (Figure 12), in which the scutum varied from B. cacuminata-like to uniformly dark with light markings at level of transverse suture and inside postpronotal lobes. Scutum was generally uniformly black, with at most small red–brown markings anterior to lateral postsutural vittae, inside postpronotal lobes and sometimes at the level of prescutellar setae, in B. fuscitibia (Figure 5), B. latilineola (Figure 9), B. paraarecae (Figure 11), B. thailandica (Figure 13), and B. usitata (Figure 14H), and frequently with more extensive red–brown markings along transverse suture in B. kanchanaburi (Figure 6) and B. melastomatos (Figure 9 F–J). The shape and width of lateral postsutural vittae was relatively constant for all species except B. thailandica (Figure 13).

Abdomen color for almost all species and variants followed the basic “T-shaped” pattern typical of the B. dorsalis complex, i.e. a black band across the base of tergum III, a narrow to broad medial longitudinal black band covering the entire length of terga III to V, and narrow to broadly expanded lateral black markings on terga III to V. Medial band was broad and lateral markings generally broad along margins of tergum III and narrower on terga IV and V in B. bhutaniae (Figure 2) and B. propinqua (Figure 12), or the markings on terga III and IV expanded and pointed at apex in B. bivittata (Figure 3). Medial band was broad and extended to the base of tergum II and lateral markings broad on terga III, IV, and base of tergum V in B. latilineola (Figure 9C–E). Medial band was narrow (broad in B. usitata) and lateral markings usually broad along terga III–IV and basal half of tergum V in B. cacuminata (Figure 4), B. kanchanaburi (Figure 6), and B. usitata (Figure 14 G). Medial band was broad and lateral markings moderately to very broad on tergum III and IV, and shining spots on tergum V usually black (fuscous to dark fuscous in most other species) and continuous with lateral black markings in B. fuscitibia (Figure 5). Medial band was narrow (broad in B. paraarecae) and lateral markings moderately to very broad but diffuse, rather than well defined (as in previous species), in B. kohkongiae (Figure 8), B. melastomatos (Figure 9 K–O), B. osbeckiae (Figure 10), and B. paraarecae (Figure 11). In B. thailandica, medial band was narrow and the extent of lateral markings varied considerably, from very limited to almost entirely covering the terga except traces of red–brown on tergum V, on either side of medial band (Figure 13).

The concatenated tree demonstrates that the OFF complex is a highly polyphyletic assemblage of unrelated species. Consistent with other published studies, the methyl eugenol responsive B. dorsalis, B. carambolae, B. occipitalis, B. cacuminata, and B. raiensis, form a well-defined monophyletic unit (Krosch et al. 2012, San Jose et al. 2013, Boykin et al. 2013, Virgilio et al. 2015). Because the phylogeny is based on a relatively limited (24%) proportion of all the species included in the OFF complex, adding more species and using multiple genes may reveal scattered clusters of related species, but the proportion of unrelated clades including OFF complex species is likely to remain high.

The widespread conformity of unrelated species to the dorsalis-like appearance is unclear. Color patterns in Dacine fruit flies are assumed to mimic wasps (White 2000), though few actual wasp mimic examples exist in Dacine fruit flies, and the OFF complex appearance is not a particularly convincing wasp imitation when compared to other groups of mimics (sesiid moths, syrphid flies, etc.). Whether the similarity represents convergent evolution or a retained ancestral state requires further investigation.

Except for a handful of well-studied species (e.g. B. dorsalis, B. carambolae, B. cacuminata), the definitions and concepts for the majority of the OFF complex species were based on morphology (mainly color patterns), lure response, and generally limited host fruit records. Only now are we starting to better characterize these species with molecular tools. Most of the non-economic species described by Drew and Hancock (1994) included in our study appear to be valid, confirmed by molecular data and comparison of morphological intraspecific variation with large series of specimens in QDAF (L.L., unpublished observations).

Attraction of B. osbeckiae to cue-lure is a new lure record. Morphological variation in our cue-lure trapped specimens (Figure 10) closely matched that observed in the QDAF and Bishop Museum (Honolulu, Hawaii, USA) series, which consist of host-reared specimens without male lure records.

Four species, consistent in appearance with the definition of the OFF complex, could not clearly be identified and are referred to here as numbered species. Species 54 (from Chiang Mai, Thailand) and 55 (Luang Nam Tha, Laos and Jinghong, China) look very similar (Figure 15A–F), yet are genetically distinct (8.85% COI pair-wise difference). They both key to B. irvingiae in Drew and Hancock (1994), but neither can be confidently matched to that species, even after comparison with series of pinned specimens of B. irvingiae and other OFF complex species in the QDAF collection. Also, B. irvingiae was collected further south in Thailand (Khao Yai) than the samples we have. Until fresh host-reared specimens of B. irvingiae can be obtained from the type locality and sequenced, we will defer from describing new species that may in the future turn out to be synonyms. Species 59 (Luang Nam Tha, Laos) and 60 (Jinghong, China) (Figure 15G–L) could not be definitely determined to species using available resources (Drew and Hancock 1994, Drew and Romig 2013), and did not match any of the OFF complex species examined in QDAF. They are likely new species, but not described here, due to the lack of distinctive characters and the very small number of specimens available (1 of species 59 and 3 of species 60). With additional survey work and genetic sequencing, a number of additional cryptic species likely will appear.

Bactrocera dorsalis, B. invadens and B. papayae were recently declared conspecific, and are genetically indistinguishable (Schutze et al. 2014a), despite what some consider diagnosable differences (Drew and Romig 2013). We have found a similar genetically indistinguishable situation for B. osbeckiae, B. melastomatos and B. rubigina (Wang and Zhao) in our phylogeny, despite them being very distinct from each other in color pattern (Figures 9F–O, 10, 14F). This suggests that these species, if distinct, may or may not differ at gene loci other than those sequenced in our study. Bactrocera osbeckiae and B. rubigina are sympatric in Thailand and Southern China (Leblanc, unpublished) and differ in color patterns and wing costal band expansion (Drew and Romig 2013), while B. melastomatos is confined to Peninsular Malaysia, Borneo, Java and Sumatra. Bactrocera osbeckiae and B. melastomatos are biologically close and peculiar, both breeding on flowers rather than fruits of Melastomataceae (Drew and Hancock 1994, Allwood et al. 1999), while the host range of B. rubigina is not well documented. Similarly, sympatric B. tryoni and B. neohumeralis in Australia are currently genetically inseparable, yet are likely valid biological species, isolated by time of mating (Clarke et al. 2011).

The high degree of intraspecific variation in color pattern severely limits the reliability of dichotomous and interactive keys. The range of variation differs considerably among species, with extreme cases like the scutum of B. dorsalis (Leblanc et al. 2013), and the abdomen of B. thailandica (Figure 13). Also, variants of unrelated species, such as B. bhutaniae (Figure 2) and B. bivittata (Figure 3), can overlap and make them hard to distinguish. The full extent of observed variation is easier to demonstrate in plates rather than words when describing a species. We suggest that descriptions of new species in the future should be accompanied by extensive plates showing variation, included in publications or posted as supplementary online material.

The OFF complex was defined by Hardy (1969) and the definition refined by Drew and Hancock (1994). The species and specimens examined in this study fit their definition in all respects, except for scutum color, said to be mostly black (Drew and Hancock 1994) or black (Drew and Romig 2013). Several species included in the complex consistently have extensive pale markings on the scutum (e.g. B. arecae (Hardy and Adachi), B. bivittata, B. cacuminata, B. osbeckiae). Bactrocera dorsalis has a broad range of variation, from entirely black to extensively or almost entirely pale (Schutze et al. 2014b, Leblanc et al. 2013), a form that was described as the now-synonymized, B. invadens (Drew et al. 2005), which was not included in the OFF complex by Drew and Romig (2013). It is likely that at least some of the 21 other species complexes (Drew 1989, Drew and Romig 2013) are also polyphyletic and their morphological diagnostic characters not robust. Nonetheless, the B. dorsalis complex is likely to remain entrenched for some time in future literature, as an informal group referred to as a “collective group” in the International Code of Zoological Nomenclature (http://iczn.org/iczn/index.jsp). Caution must be exercised in literature to not refer to the group as a biological or evolutionary unit.

Collecting trips in Asia were funded through USDA-ARS Specific Cooperative Agreements 58-5320-9-382 and 58-5320-4-018, managed by the University of Hawaii’s College of Tropical Agriculture and Human Resources (UH-CTAHR). A study trip in Australia was covered by Farm Bill funding (project 3.0251), through a Cooperative Agreement between USDA and UH-CTAHR. Additional support was provided by USDA-NIFA Hatch projects HAW00942-H and HAW00956-H, administered UH-CTAHR. We also thank Anthony R. Clarke for reviewing an earlier version of the manuscript and Richard A.I. Drew and David L. Hancock for their insight, mentoring and long hours discussing the relationships among species in the OFF complex and the higher classification of dacine fruit flies. Desley Tree and Justin Bartlett provided access to the impressive QDAF fruit fly collection in Brisbane. The use or mention of a trademark or proprietary product does not constitute an endorsement, guarantee, or warranty of the product and does not imply its approval to the exclusion of other suitable products by the U.S. Department of Agriculture, an equal opportunity employer.