00 [16] and ARLEQUIN version 3.5.1.3 [17]. Deviations from Hardy-Weinberg sellckchem equilibrium (HWE) were tested for each locus and over all loci in ARLEQUIN using a Markov chain approximation [18]. All estimates were assessed for significance using a test analogous to Fisher’s exact test, with 100,000 steps in the Markov chain and 5000 dememorization steps. Significance for all estimates was placed at the 0.05 level. Other details on the microsatellite protocol are given elsewhere [19]. Calculations of Kimura’s [20] 2-parameter genetic distances (d) were carried out in MEGA version 5.0.5 [21]. Genetic diversity indices were calculated in DnaSP version 5.00.07 [22]. Neutrality tests (Tajima’s D [23] and Fu’s FS [24]) were carried out in ARLEQUIN.

Fu’s FS test is also useful for detecting signatures of population expansions, which lead to large negative values in the test statistic [24, 25]. The significance of FS at the 0.05 level is indicated when P values are <0.02 [17]. Networks for COI haplotypes were constructed using statistical parsimony implemented in TCS version 1.21 [26]. The connection limit among haplotypes was set to the default value of 95%, unless indicated otherwise.2.3. Phylogenetic AnalysesRelationships among COI haplotypes in Sonoran Desert Culex were examined using maximum parsimony (MP) and Bayesian inference. For all phylogenetic analyses, sequences for Cx. tarsalis were trimmed from 624 to 611bp to correspond to the sequence length of the other samples (Table 1). We also incorporated GenBank sequences for several different species of Culex into the data matrix, including Cx.

(Neoculex) territans Walker and Cx. (Culiciomyia) nigropunctatus Edwards. All other Culex species treated here are presently assigned to the subgenus Culex [4]. Culiseta inornata (Williston) from the tribe Culisetini was used as the outgroup based on results of previous molecular studies of Culicidae [11, 27]. Maximum parsimony analyses were carried out in MEGA using the CNI heuristic search option and 100 random additions of sequences. Relative support for tree topology was obtained by bootstrapping [28] using 1000 pseudoreplicates. Bayesian analyses were implemented in MrBayes version 3.1 [29]. The model of nucleotide substitution that best fitted the data set, determined with jModelTest 0.1.1 [30] using the Akaike Information Criterion was, TVM + G.

The substitution model was set to nst = ��2�� and rates = ��gamma��, and the analysis was Cilengitide run for 1,000,000 generations, sampled every 250th generation (4,000 trees sampled), using the default random tree option to begin the analysis. We also conducted an analysis with nst = ��6,�� used for the more highly parameterized GTR substitution model, and obtained the same tree topology and similar clade support values.