Oxygen conductance to the tissues determines aerobic metabolic performance in most eukaryotes but has cost/benefit tradeoffs. Here we examine in lowland populations of a butterfly a genetic polymorphism affecting oxygen conductance via the hypoxia-inducible factor (HIF) pathway, which senses intracellular oxygen and controls the development of oxygen delivery networks. Genetically distinct clades of Glanville fritillary (Melitaea cinxia) across a continental scale maintain, at intermediate frequencies, alleles in a metabolic enzyme (succinate dehydrogenase, SDH) that regulates HIF-1α. One Sdhd allele was associated with reduced SDH activity rate, twofold greater cross-sectional area of tracheoles in flight muscle, and better flight performance. Butterflies with less tracheal development had greater post-flight hypoxia signaling, swollen & disrupted mitochondria, and accelerated aging of flight metabolic performance. Allelic associations with metabolic and aging phenotypes were replicated in samples from different clades. Experimentally elevated succinate in pupae increased the abundance of HIF-1α and expression of genes responsive to HIF activation, including tracheal morphogenesis genes. These results indicate that the hypoxia inducible pathway, even in lowland populations, can be an important axis for genetic variation underlying intraspecific differences in oxygen delivery, physiological performance, and life history.

How fast an organism consumes energy and its ability to perform various activities depend on the rate of oxygen delivery to the tissues, but too much oxygen causes toxic effects (Hochachka and Somero 2002; Lane 2002). Oxygen delivery rate in most macroscopic organisms is determined by the size and function of branching networks of tubes whose morphogenesis is controlled primarily by the evolutionarily conserved hypoxia-inducible factor (HIF) pathway (Fraisl et al. 2009). Size of oxygen delivery networks has been explored in relation to ontogenetic growth (Greenlee et al. 2009) and interspecific scaling of metabolic rates (West et al. 1997), but no attention has been paid to the possibility that tradeoffs involving oxygen might commonly maintain HIF pathway polymorphisms within species. In animals, such polymorphisms could contribute strongly to variation in features such as cardiovascular physiology, athletic ability and longevity, and affect life-history traits relevant to fields spanning from medicine to ecology and conservation biology.

By training at high altitude, human athletes activate the HIF pathway and trigger physiological responses resulting in enhanced oxygen delivery (Hoppeler and Vogt 2001), effects paralleled by certain doping methods that mimic HIF pathway activation (i.e., elevation of erythropoietin; Jelkmann 2003). Long-term exposure of populations to high altitude causes evolutionary responses in HIF pathway genes, as revealed by genomic scans for selection in both humans (Simonson et al. 2010; Yi et al. 2010) and yaks (Qiu et al. 2012) living in Tibet. Human habitation at high altitude is fairly recent and so the evolutionary response in HIF genes occurred extremely fast (Yi et al. 2010). Rapid evolutionary responses to hypoxia have occurred also in laboratory selection experiments with fruit flies (Ali et al. 2012), suggesting widespread presence in ancestral populations of allelic variation in HIF pathway genes. However, the presence of such variation and its physiological and ecological function in lowland populations has not been examined.

Here we test the hypothesis that variation in oxygen delivery network elaboration, aerobic metabolic performance and mitochondrial aging are associated with polymorphism in a HIF pathway gene in lowland populations of the Glanville fritillary butterfly (Melitaea cinxia, Nymphalidae). This species is a well-established model for large-scale ecology and metapopulation biology (Ehrlich and Hanski 2004), and more recently for relating allelic variation affecting flight, dispersal and life-history phenotypes to ecological and evolutionary dynamics (Haag et al. 2005; Hanski and Saccheri 2006; Zheng et al. 2009; Wheat et al. 2011; Hanski 2011). Two loci, phosphoglucose isomerase (Pgi) and succinate dehydrogenase d (Sdhd), have emerged as principal foci in these studies. Both have strong fitness effects, including context-dependent allelic associations with population growth rates (Wheat et al. 2011). Sdhd encodes one of the four subunits of succinate dehydrogenase (SDH; Rutter et al. 2010), a tricarboxylic acid (TCA) cycle enzyme whose activity level affects the concentration of succinate (Hobert et al. 2012), which regulates the HIF pathway (Gimenez-Roqueplo et al. 2001; Pollard et al. 2005; Selak et al. 2005; Smith et al. 2007; Baysal 2008; Burnichon et al. 2010; Puissegur et al. 2011; Fig 1).

Figure 1.

Diagram of the role of succinate dehydrogenase (SDH) and succinate in the hypoxia-inducible factor (HIF) pathway. Insufficient oxygen availability or genetic variation reducing SDH activity in the TCA cycle causes accumulation of succinate, which passes out of the mitochondria and inhibits PHD, an enzyme that tags HIF-1α for proteosomal degradation by von Hippel-Lindau (VHL) and the E3 ubiquitin ligase complex. HIF-1α binds with HIF-1β to form a transcription factor that turns on hypoxia-response genes, including Phd (a feedback loop that may quickly turn off hypoxia signaling as soon as oxygen is again sufficient). In insects, hypoxia-responsive genes include those controlling morphogenesis of tracheal tubes that deliver oxygen directly to cells and mitochondria. References are contained in the main text.

In insects, oxygen is delivered directly to cells and mitochondria via a branching network of air-filled tracheal tubes that proliferate in response to HIF pathway activation (Fig 1; Lavista-Llanos et al. 2002; Arquier et al. 2006; Centanin et al. 2008; Centanin et al. 2010). Differences in insect tracheal network morphology and oxygen conductance have substantial physiological effects (Greenlee and Harrison 2004; Harrison et al. 2006; Callier and Nijhout 2011) and evolutionary consequences (Henry and Harrison 2004; Kaiser et al. 2007; Klok and Harrison 2009), yet as for lowland animals in general, the possibility and importance of naturally occurring, intraspecific variation in HIF pathway activity and oxygen delivery has not been explored.

Methods Summary

We used Glanville fritillary (M. cinxia) butterflies collected as eggs from mated females originating from many local populations in the Åland Islands of Finland and two large populations near Provence, France. An additional sample comprised wild-caught adults near the start of the flight season from a large population near Rupit, Spain. All of this material (summarized in Table S1) should be representative of the genetic variation in outcrossed wild populations at the three sites.

To genotype individuals for the Sdhd indel polymorphism, a portion of the 3′-UTR containing the indel was PCR amplified from genomic DNA or cDNA using a fluorescently labeled primer. Fragment analysis was used to associate size of labeled amplicons with the indel allele. We tested for an excess of intermediate-frequency Sdhd alleles using the Ewens–Watterson test, as implemented in Arlequin v.3.1 (Excoffier et al. 2005). The sum of Sdhd indel haplotype allele frequencies (fobs) at the three distant sample locations was compared to their expected homozygosity values (fexp) by generating 99,999 random neutral samples. To compare the frequency of Sdhd indel alleles to variable sites throughout the genome, we determined the minority allele frequency of well-supported single nucleotide polymorphisms (SNPs) in the large pool of butterflies from Finland represented in Assembly 2.0 of the M. cinxia transcriptome (Vera et al. 2008; Wheat et al. 2011).

We measured SDH enzyme activity rate (Vmax; in this case a putative index of enzyme concentration because the indel polymorphism does not affect amino acid sequence) in homogenates of dorsoventral flight muscles using methods adapted from those described in (Selak et al. 2005).

To quantify flight metabolic rate, butterflies were stimulated to fly as continuously as possible in a 1 L jar through which dry CO2-free air was passed at 0.95 L min−1 (Haag et al. 2005). From those data, peak performance and the total volume of CO2 emitted were determined. A 14% oxygen-over-nitrogen mix was used to determine flight metabolism under mild hypoxia. The first respriormetry experiment used 2- to 4-day lab-reared males from France (N= 28); each was flown first in normoxia, then again in hypoxia after a 30-min recovery period so that all measures of post-flight gene expression and mitochondrial integrity came from butterflies last exposed to the same treatment. The second respirometry experiment used a mixture of male and female butterflies from Spain (N= 30), with flights shortened to 3-min duration and the order of normoxia or hypoxia (14% O2) randomized. These two flight tests occurred on consecutive days, between which the butterflies were fed and maintained indoors in a cool shaded cage where they did not fly.

Portions of dorsal longitudinal flight muscles (DLM) from butterflies used in the first respirometry experiment were fixed and used for microscopic analyses of tracheoles and mitochondrial morphology. For imaging tracheoles, muscle samples were stained with Calcofluor-white, which labels chitin, a material present in insect tracheae and tracheoles (Zimoch et al. 2005). Using stacked confocal images of 1-μm-spaced planes within 10-μm-thick sections of myofibrils, we determined the mean cross-sectional area of tracheoles from three randomly selected fields from each of five different serial sections from the DLM of each butterfly. Portions of the same DLM were prepared separately for transmission electron microscopy (TEM) using standard fixation, embedding, sectioning, and imaging protocols. Images were obtained from three random fields of the DLM of each individual, from which we measured the size and optical density (to estimate crista membrane structural integrity) of each mitochondrion that was fully contained within the TEM image. Tissue dissection, fixation, sectioning and imaging of tracheoles was successful for 27 of 28 samples; for TEM imaging of mitochondria we succeeded with 21 of 28 samples. For both tracheal and mitochondrial morphology, the replicate fields were used to estimate the mean for each individual; that single value was used in statistical analyses.

To examine effects of SDH metabolites on HIF signaling and downstream gene expression, pupae (5-day age) were injected in the abdomen with saline (negative control) or cell-permeable diethyl forms of SDH metabolites (fumarate, succinate) known from previous studies to inhibit Hif prolyl-hydroxylase (PHD) and thereby increase Hif-1α protein. Other pupae were held without injection for 24 h in mild hypoxia (7% O2) or normoxia. As an additional positive control, we injected pupae with 500 μg of cobalt chloride (CoCl2), an inhibitor of Hif-1α degradation (Yuan et al. 2003). After 24 h (or 4 h in the case of CoCl2 treatment), pupae were flash frozen and the thorax was removed and homogenized. A constant amount of total protein (10 μg) was used for Western blotting to visualize Hif-1α protein.

Gene expression was quantified for Sdhd, Cytoplasmic actin (control gene for normalization), and homologs of Drosophila melanogaster Phd (using primers for a constitutively spliced region), and two tracheal development genes, No ocelli and Nervana2, using Taqman primers and probes designed from sequences in M. cinxia transcriptome 2.0 (Wheat et al. 2011).

Methods details are contained in a supplement.



Glanville fritillaries have an apparent single-copy Sdhd gene containing an indel polymorphism located 492 nucleotides from the stop codon in the 3′-UTR (Fig 2A). There are two large deletion alleles (D, E; lacking 53 and 62 bases) and two insertion alleles (I, M; differing primarily by a 4-nucleotide mini-deletion). The polymorphism occurs in an AU-rich region (77% AU) characteristic of 3′-UTR regulatory sites (von Roretz et al. 2011). In human Sdhd there is a functionally important miR target in the 3′-UTR (Puissegur et al. 2011) and so it is noteworthy that the indel in Glanville fritillary Sdhd also contains a putative target site for a microRNA (miR-71; Fig 2A) expressed in flight muscle and elsewhere (data not shown) and known to have potent effects on longevity and oxygen radical toxicity in a model invertebrate (Pincus et al. 2011). Only the Sdhd M allele has a predicted mRNA structure likely to allow full access (Svoboda and Di Cara 2006) for binding with miR-71 (Fig. S1).

Figure 2.

Sdhd gene polymorphism in M. cinxia (GenBank JX870003- JX870007). Deduced amino acid length of the protein is 186 with Blastp identity of 51% to the 182-amino acid homolog of Drososphila melanogaster (CG10219). (A) Gene structure, including the 3′-UTR indel polymorphism and complimentary alignment of M. cinxia miR-71 (below; JX878561). (B) Allele frequencies at three locations containing genetically distinct clades (N indicates chromosomes sampled). (C) Mean SDH enzyme activity (Table S3; 106× delta absorbance min−1μg protein−1) and normalized Sdhd transcript abundance. Note the lower SDH enzyme activity in genotypes containing the M allele (see also Tables S3, S4).

Three of the Sdhd alleles are common (M, D, I) and occur at intermediate frequencies in locations that span the latitudinal range of M. cinxia in western Europe (Finland, France, Spain; Fig 2B) and contain distinct mitochondrial haplotype clades (Wahlberg and Saccheri 2007). We tested for deviations from neutrality in Sdhd genotype frequencies (Table S2) and found an excess of intermediate-frequency alleles at all three locations (Ewens-Watterson test; fobs= 0.34, 0.32, 0.32; P= 0.003, 0.05, 0.02). To determine if this might be an artifact of demographic effects, we calculated the minor allele frequency distribution of predominately biallelic SNPs in Finland-derived sequences of the M. cinxia transcriptome (Fig. S2). In Finland, the two Sdhd alleles likely to represent a single mutational event (I, D; different by a single 53nt deletion) were used to estimate a minor allele frequency (I/D + I), and this falls in the upper tail (95th percentile) of the genome wide distribution (Fig. S2). Similar results (98th percentile) were obtained when we compared the three Sdhd indel alleles present in Finland to the frequency distribution of tri-allelic SNPs in Finland. From these analyses it is apparent that allele frequencies at Sdhd depart from expected patterns of neutral evolution.


To test the hypothesis that the indel polymorphism is associated with variation in gene regulation and/or translation, we measured Sdhd transcript abundance in three body regions (head, thorax, abdomen; total N= 113) and SDH activity in flight muscle (N= 50). Sdhd transcript abundance in newly emerged and mature adults varied significantly with indel genotype (P= 0.006; Table S3), most notably a 12% increase in transcript abundance in genotypes containing one or more copies of the Sdhd M allele (Fig 2C). Even so, SDH activity (Vmax) was significantly reduced (27%) in M allele butterflies (P= 0.02 for the entire sample, P= 0.04 for a more uniform subsample; Fig 2C, Tables S3, Table S4), consistent with the hypothesis of allele-specific inhibition of Sdhd translation by regulatory effects determined by the polymorphic 3′-UTR. A nonsynonymous polymorphism at amino acid 72 (threonine–alanine) assorted independently of the 3′-UTR indel polymorphism and had no significant association with SDH activity (Tables S3, S4).


Reductions in SDH activity cause accumulation of intracellular succinate, which stabilizes HIF-1α protein and in humans leads to diseases that involve hypertrophy of oxygen delivery networks (Gimenez-Roqueplo et al. 2001; Selak et al. 2005; Pollard et al. 2005; Smith et al. 2007; Baysal 2008; Burnichon et al. 2010; Puissegur et al. 2011). Those cases involve heterozygotes for rare loss-of-function mutations, but more subtle differences in SDH function could in principle be associated with quantitative differences in oxygen network development and function. Accordingly, we sought to determine if common Sdhd alleles in M. cinxia are associated with quantitative variation in tracheal morphology. To do this, we combined an examination of tracheae with measures of metabolic performance in butterflies stimulated to fly continuously for 10 min in normal (21% O2) and hypoxic air (14% O2). A decline in performance between normoxia and hypoxia would indicate that metabolism is limited by tracheal conductance (Harrison and Lighton 1998). As expected, the decline in performance between the initial flight in normoxia and the subsequent flight in hypoxia was greater in butterflies with less cross-sectional area of tracheoles in their myofibrils (Fig. 3A, one-tailed P= 0.03). Furthermore, more elaborate tracheae were associated with higher peak performance and endurance (total CO2 emitted) in both atmospheres (P < 0.004 in all cases, Table S5). Butterflies carrying the Sdhd M allele, associated with reduced SDH activity, had nearly twofold greater cross-sectional area of tracheoles (Fig. 4A; P = 0.003).

Figure 3.

(A, B) Confocal microscopy images of Calcofluor-stained tracheoles in flight muscle of butterflies with low (A) and high (B) tracheal elaboration. (C, D) Traces showing flight metabolic rate during consecutive 10 min flights in normoxia (red trace) and mild hypoxia (blue trace) of the same two butterflies represented in panels A and B. Note the greater endurance and lack of a decline in hypoxia for the individual with greater tracheal elaboration.

Figure 4.

Genetic variation associated with differences in tracheal elaboration and expression of the hypoxia-responsive gene Phd after consecutive flights in normoxia and hypoxia. (A) Mean cross-sectional area (± SE) of tracheoles in flight muscle of butterflies with and without the Sdhd M allele (sample sizes shown above bars). (B) Mean post-flight transcript abundance of Phd in flight muscles of butterflies with and without the Sdhd M allele. (C) Continuous variation in Phd expression in relation to the fractional cross-sectional area of tracheoles in flight muscle.

Active skeletal muscles quickly reduce intracellular oxygen to levels low enough to activate the HIF pathway (Ameln et al. 2005) and therefore individual butterflies with greater oxygen conductance should have reduced hypoxic stress and HIF signaling after prolonged bouts of flying. Indeed, after the two 10-min flights, a hypoxia-responsive (Lavista-Llanos et al. 2002; D’Angelo et al. 2003) gene, Phd, was expressed at lower levels in butterflies carrying the Sdhd M allele (Fig 4B; F= 15.9, P= 0.0005). Across all genotypes, expression of Phd showed a continuous inverse relationship with cross-sectional area of tracheoles (Fig 4C; P= 0.002).


During flight experiments, nearly all of the butterflies showed repeated episodes of complete fatigue. They repeatedly recovered and flew again, but this may mask cellular damage, particularly to mitochondria, which in response to hypoxic ATP depletion lose volume homeostasis and swell (Kaasik et al. 2007). Transmission electron micrographs of flight muscle sections showed broad variation in mitochondrial swelling (Fig. 5A, B). In a multivariate model, mean mitochondrial size (cross-sectional area) was related inversely to the cross-sectional area of tracheoles (Table S6A; P = 0.05) and positively to the total CO2 emitted during the two flights (P = 0.008), indicating that mitochondrial swelling was more pronounced in butterflies that used their flight muscles more intensely in relation to their tissue-level oxygen supply (Fig. 5C, D).

Figure 5.

Variation in mitochondrial swelling. (A,B) Range of variation among individuals. For illustrative purposes, contrast was adjusted to make the mitochondria appear darker than myofibrils. (C, D) Plots showing the relationship between flight metabolic rate, tracheal cross-sectional area, and mitochondria size. The independent variables are residuals to simulate a model (Table S6A) containing both metabolic rate and tracheal cross-sectional area.

It is possible that large deletions in the 3′-UTR of Sdhd disrupt regulation of SDH activity and aerobic metabolism, with consequent effects on both function and health. For that reason we performed additional analyses to look for possible dominant physiological effects associated with the Sdhd D allele. Our sample in the initial respirometry experiment contained only two DM heterozygotes and therefore analyses based on presence of M or D alleles used nearly nonoverlapping subsets. Butterflies with one or more copies of the D allele had higher flight metabolic rate in normoxia when we controlled statistically for different levels of tracheal cross-sectional area (Table S5A, B), and they had a significant reduction in mitochondrial integrity (Fig 6A; P= 0.04; Table S6B). These results suggest that the large deletion in the Sdhd D allele may cause a mismatch between the TCA cycle rate and the oxygen supply rate, with detrimental effects on mitochondrial health.

Figure 6.

Mitochondrial integrity difference and decline in metabolic performance associated with the Sdhd D allele. (A) Average optical density of mitochondrial cristae (0–255 scale; where white = 0) related to Sdhd D allele genotype. Shown are LS means (± SE) from a model that included body mass (Table S6B). (B, C) Interaction effect (P= 0.005; Table S8) between peak performance during a 3 min flight on day 1 and total CO2 emitted during a 3 min flight on day 2. The data in panel A versus B & C come from two different clades and show a consistent association between the Sdhd D allele and accelerated aging of mitochondrial structure and function.


We tested the robustness of these results by repeating the respirometry experiment with an independent cohort (N= 30 butterflies from Spain). This sample contained no DM heterozygotes and so again we tested for dominant effects of the D and M alleles. As in the original experiment, Sdhd M butterflies produced more CO2 during flights in normoxia and hypoxia (P= 0.02, 0.06), and had higher peak metabolic rate when flying in hypoxia (P= 0.04), thereby confirming a robust allelic association between the M allele and enhanced oxygen conductance (Table S7). Butterflies with the Sdhd D allele had a significant decline in peak metabolic rate during their second flight (one-tailed P= 0.05), as predicted by the previous results regarding mitochondrial damage. Most strikingly, butterflies with the D allele showed an inverse relationship between their peak metabolic rate on day 1 and their total CO2 emitted (i.e., endurance) on day 2. In contrast, butterflies without the D allele showed the expected strong positive relationship, wherein better fliers on day 1 also performed better on day 2 (Fig. 6B, C; P= 0.005 for the interaction effect; Table S8). Hence, for butterflies carrying the D allele, high peak performance was damaging for future flight endurance.


Thoraces from pupae injected with cell-permeable forms of succinate or fumarate (Koivunen et al. 2007) and a positive control (CoCl2, an inhibitor of Hif-1α degradation) contained more of a protein that hybridized specifically with insect Hif-1α antibody (Fig 7A). This band had an apparent molecular weight similar to the predicted size (Danaus plexippus; GenBank: EHJ70114.1, 87 kDa predicted molecular weight) and was faint or undetectable in negative controls.

Figure 7.

Activation of the HIF pathway and upregulation of tracheal morphogenesis genes. (A, top panel) Myosin heavy chain (MHC) bands from total protein staining of the membrane; this is a loading control; (lower panel) HIF-1α protein (lower band) on a Western blot. The more darkly labeled band is an artifact; it was labeled nonspecifically by the secondary antibody but covaried in abundance with HIF-1α independently of MHC and total protein loading. Lanes are pupae injected with cobalt chloride (positive control), fumarate, succinate, saline (negative control), and an uninjected pupa in normoxia (negative control). (B) Transcript abundance (normalized to cytoplasmic actin) of Phd (± SEM) in pupae injected with fumarate, succinate, or saline, or not injected and held for 24 h in mild hypoxia (7% O2) or normoxia (N in each bar). The lack of a difference between mild hypoxia and saline injection may result from moderate hypoxia signaling in response to wounding by the injection. (C) Transcript abundance of the tracheal morphogensis gene Noc in relation to Phd in the same pupae. (D) Transcript abundance of another tracheal morphogensis gene, Nrv2, in relation to Phd in the same pupae (closed circles) and in untreated newly emerged adults (open circles).

To compare quantitative differences in HIF pathway activation across a larger number of samples, we used qPCR to measure the normalized transcript abundance of Phd, which in insects (Lavista-Llanos et al. 2002) and mammals shows a strong transcriptional upregulation in response to Hif-1α(Fig 1). Hence, to determine in vivo effects of elevated SDH metabolites on HIF pathway activation in Glanville fritillary butterflies, we injected whole pupae with cell permeable forms of succinate, fumarate, or saline (Fig 7B), As expected, the metabolite-injected pupae (N= 9) had higher Phd transcript abundance (Fig 7B; F= 6.3, P= 0.009) compared to negative controls that received saline (N= 5). Pupae not injected but kept in mild hypoxia (7% O2; N= 3) had slightly higher Phd expression than uninjected pupae (N= 2) maintained in normoxia.

Activation of the HIF pathway in insects stimulates expression of genes promoting tracheal network elaboration (Lavista-Llanos et al. 2002; Arquier et al. 2006). To test in Glanville fritillary butterflies for downstream effects of HIF pathway activation on genes involved in tracheal elaboration, we used data from a microarray experiment (Wheat et al. 2011) to identify tracheal morphogenesis genes that differed in expression (P < 0.05) according to presence/absence of the Sdhd M allele (Table S9). This revealed orthologs of four genes known in Drosophila to affect tracheal tube size or branching (CG14779, CG9535, CG9261, CG4491). We selected two of those (No ocelli and Nervana2; Contigs 20223 and 27909 in M. cinxia transcriptome assembly 2.0) and compared their expression level in pupae and newly emerged adults with levels of the HIF-responsive gene, Phd. No ocelli (Noc), which participates in the regulation of tracheal branching (Dorfman et al. 2002), was positively related to Phd transcript abundance (Fig 7C; F= 7.9, P = 0.01), and was higher in pupae injected with SDH metabolites (F= 5.7, P = 0.01). Nervana 2 (nrv2), which regulates the size of tracheoles (Paul et al. 2007), was expressed at very low levels and did not respond to experimental treatments in pupae, but was more abundant and increased with Phd transcript abundance in a sample (Table S1) of newly emerged adults (1 h post-emergence; Fig 7D; F= 8.9, P = 0.006), the stage during which tracheoles attain their final form. Together, these in vivo data support the hypothesis that butterflies with higher levels of succinate have increased activation of the HIF pathway, which in turn is related to greater expression of tracheal morphogenesis genes.


We found that lowland populations of a weak-flying butterfly contain common alleles associated with the size and function of oxygen delivery networks, flight metabolic performance, and the aging of mitochondria. In our butterfly model, differences in tracheal elaboration appear to be determined by quantitative variation in SDH enzyme activity and hypoxia signaling. Previous biomedical studies have linked rare heterozygous SDH loss-of-function mutations in humans to pseudohypoxia signaling and development of vascularized tumors (Gimenez-Roqueplo et al. 2001; Pollard et al. 2005; Selak et al. 2005; Smith et al. 2007; Baysal 2008; Burnichon et al. 2010; Puissegur et al. 2011). Here we show that more subtle changes in SDH can give rise to variation in physiological traits other than the known SDH-related disease states. By experimentally increasing intracellular succinate in pupae, we addressed key assumptions of the mechanistic model (Fig. 1) by showing that excess succinate increases the level of Hif-1α protein and expression of hypoxia-responsive and tracheal morphogenesis genes. Together, these data provide a coherent explanation for why genetic variation in Sdhd affects a signaling pathway, tracheal development, and the physiological differences underlying ecologically important life-history variation.

Samples from three populations spanning the latitudinal range of M. cinxia in western Europe all showed an excess of intermediate frequency Sdhd alleles. Based on phylogenetic and phylogeographic studies of this species (Wahlberg and Saccheri 2007; Leneveu et al. 2009), the mitochondrial haplotype clades in our samples from Spain and Finland last shared a common ancestor at least 500,000 generations ago, a divergence older than that of humans and chimps (Langergraber et al. 2012). The maintenance of intermediate allele frequencies in the Åland archipelago in Finland is particularly interesting, as this genetically isolated metapopulation originated from a north-Asian clade (Wahlberg and Saccheri 2007) following the last glacial maximum (i.e., a bottleneck) and now comprises a network of small populations subject to frequent local extinction, inbreeding and genetic drift (Saccheri et al. 1998). These characteristics cause rapid and substantial changes in allele frequencies at neutral loci (Wheat et al. 2010). A large-scale ecological study in this metapopulation revealed strong context-dependent fitness effects, including an interaction between the Sdhd M allele frequency, patch size and the growth rates of local populations (Wheat et al. 2011). Together, these results showing intermediate allele frequencies in clades with different demographic histories, large phenotypic effects on life history traits, and context-dependent fitness of alleles in wild populations constitute strong evidence that the polymorphism has been maintained by long-term balancing selection.

A potent effect of SDH enzyme rate on oxygen-related physiological and life-history traits of insects is corroborated by previous laboratory studies. A screen in Drosophila for experimentally induced mutations causing poor longevity in hyperoxia (Walker et al. 2006) found a Sdhb mutation that reduced SDH enzyme activity, increased the rate of oxygen radical production, and caused severe damage to mitochondrial cristae when the flies were exposed to hyperoxia. Similarly, Drosophila selected over many generations to survive in hypoxia (4% atmospheric oxygen) had a reduction in SDH activity (but no reduction in the activity of other mitochondrial respiratory complexes), and a reduced rate of oxygen radical production in hypoxia (Ali et al. 2012). These studies reveal a consistent tendency for reduced SDH activity to protect against cellular damage in hypoxic atmospheres but cause greater damage in hyperoxic atmospheres, perhaps because of the mechanism shown here in butterflies: reduced SDH activity leads to increased tracheal development and oxygen conductance to the tissues.

A previous study of Glanville fritillaries in Finland (Wheat et al. 2011) that flew butterflies only once and did not measure tracheal phenotypes found better endurance of flight metabolic performance in Sdhd D butterflies, but only when they also carried a particular Pgi allele (i.e., epistasis). We have begun to explore the Pgi variation in Spain, finding much more recombination among the SNP sites compared to the high LD and distinct functionally important Pgi haplotypes in the Finnish population. Pgi alleles themselves have strong effects on flight performance, dispersal, and longevity in Glanville fritillaries (Haag et al. 2005; Orsini et al. 2009; Niitepold et al. 2009), but the physiological mechanisms have not been determined. Pgi variants in other invertebrates differ in flux through the pentose shunt (Zamer and Hoffmann 1989), which may change the balance of NADH/NADPH, another effector of HIF pathway activation (Osada-Oka et al. 2010). Hence, there is a possibility that Pgi and Sdhd polymorphisms both affect tracheal morphology. Both are certainly involved in supplying ATP for flight. In any case, it appears that the allelic associations we show here for tracheal conductance phenotypes in southwestern European populations may differ in other populations due to genetic background at other loci, including Pgi.

A curious finding given the strong evidence for balancing selection is that all of our physiological data for flight-related phenotypes suggest fitness advantages for the Sdhd M allele, with no apparent tradeoffs. The I allele showed no dominant phenotypes and remains enigmatic, along with the rare E allele. The D allele was associated with mitochondrial degeneration and reduced performance in subsequent flights after only minimal flight activity. Based on these phenotypes, it is not clear what fitness effects counterbalance the flight-related advantages of Sdhd M and maintain the polymorphism. One likely possibility is that there is a tradeoff between tracheal elaboration and dessication so that butterflies lacking the M allele have reduced rates of respiratory water loss. Insects lose water at elevated rates during flight (Lehmann and Schutzner 2010), oxygen-water tradeoffs constrain developmental rates (Zrubek and Woods 2006), and Drosophila evolve smaller tracheae when reared for multiple generations in hyperoxia (Henry and Harrison 2004), all of which indicate that unnecessarily large tracheae have a negative effect on fitness (Harrison et al. 2012). Possible selection for smaller tracheae and a reduced rate of water loss fits well with the larval ecology of Glanville fritillaries, as their host plants tend to dry out in late summer, with the driest summers causing population declines (Nieminen et al. 2004). A fitness advantage for reduced tracheal network size in larvae could potentially offset the flight-related advantages of Sdhd M in adults.

All signaling pathways present large genomic targets for alleles with quantitative effects. Considering that oxygen plays a central role in animal development, design, aerobic performance, and aging, with abundant potential tradeoffs, polymorphisms affecting HIF pathway activity, oxygen delivery, athleticism, mitochondrial health, respiratory water loss, and related effects on longevity and life history are likely to be common across aerobic species. Accordingly, HIF-pathway polymorphism in butterflies is likely to be the “tip of the iceberg” of an abundant class of genetic variation affecting the way animals living at low elevations sense and respond to cellular oxygen availability.

Associate Editor: A. Kopp


We thank D. Arthurs for supplying butterflies from France, I. Hanski and S. Ikonen for butterflies from Finland, and C. Stefanescu for hosting the research and providing access to field sites in Spain. Thanks also to the C. Carnie family, J. Portet, E. Miralles, and Montseny Natural Park for supplying housing and field laboratory space, M. Hazen and N. Zembower for expertise and assistance with tissue preparation and microscopy, J. Pekny for assistance with respirometry and genotyping, and P. Wappner for sharing an insect Hif-1α antibody. This work was supported by NSF grant IOS-0950416 to JHM, NSF REU supplements IOS-1132133 and 1214224, Academy of Finland Grant 131155 to CWW, and a Postdoctoral Fellowship in Physiological Genomics from the American Physiological Society to RJS.