SEARCH

SEARCH BY CITATION

Keywords:

  • coevolution;
  • compensatory evolution;
  • Drosophila melanogaster;
  • fertility;
  • fitness;
  • mutation;
  • sexual selection;
  • viability

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

A wild-type population of Drosophila melanogaster was used to assess the impact of a known deleterious mutation, nub1, when it had (1) evolved for up to 180 generations with the mutation or (2) recently had the same mutant allele introgressed into it. Relative to this benchmark, we observed much stronger initial fitness depression in males (−74%) than in females (−38%) and also relatively greater fitness recovery by evolved males (+55%) than females (+17%). Experimental assays revealed amelioration in both juvenile and adult fitness and suggested that the greater relative recovery of male fitness was from gains through sexual selection. These evolutionary changes in male fertility depended on pairing with their coevolved mates for both mate choice and post-copulatory components of sexual selection. Without replication at the population level, these results are used to motivate a general hypothesis rather than definitively test it: Differences in reproductive optima may generally skew mutational effects towards the more strongly sexually-selected sex due to genic capture and condition dependence.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Despite the action of natural selection, harmful mutations can accumulate in populations through three genetic processes: genetic drift, hitchhiking and transient fitness advantages. The first two processes are driven by chance, finite population size and linkage. The third results when a mutation has a selective advantage under specific environmental conditions but is costly when conditions change. For example, antibiotic resistance mutations are favoured in microbes during drug treatment, but often result in fitness costs under permissive conditions. The attenuation of such costs through the evolution of modifiers is known as compensatory evolution. This process has been widely cited in the origin of ‘superbugs’, organisms that exhibit resistance to a biocidal agent without the negative direct or pleiotropic side-effects that would lead to their remission in the absence of the selective agent (e.g. Schrag, 1996).

Amelioration of fitness costs through compensatory evolution has been investigated in several bacterial species possessing drug resistance mutations (Lenski, 1988; Cohan et al., 1994; Schrag, 1996; Levin et al., 2000; Moore et al., 2000; Reynolds, 2000). For example, Bouma & Lenski (1988) demonstrated that a plasmid conferring antibiotic resistance to susceptible Escherichia coli initially had a fitness cost to the host when antibiotics were absent, but the host genome evolved rapidly to compensate for this cost. After 500 generations, the plasmid had become beneficial to the host, even under permissive conditions. Bouma & Lenski (1988) likened this process of co-adaptation to the host genome crossing a valley of low fitness on an adaptive landscape (Wright, 1932) and winding up on a new, higher fitness peak.

To date, the majority of compensatory evolution studies have involved asexual organisms, such as bacteria and viruses. The advantages of these systems are considerable, as they have rapid generation times, large population sizes and fitness is relatively easy to measure. Sexual species present a set of unique considerations and challenges. Because the two sexes often have different fitness-maximizing strategies, the impact of new mutations may be asymmetrical. We suggest that, in general, the effects of a visible mutation should be more severe for the more sexually selected sex (usually the male). This sex will experience the negative impact of the mutation on general performance, any genic capture of the novel variation by sexual selection (reviewed in Tomkins et al., 2004), and further reduced success in competition for mates if the trait is a target of mate choice. Moreover, exaggerated male secondary sexual characters may exhibit negatively biased mutational effects because they are often intricate, involve multiple developmental or physiological pathways, or are near the limit of energetic expenditure (Pomiankowski et al., 1991). The prediction that males will express mutational variance more strongly than females in terms of fitness is a genetic extension of Bateman's (1948) principle.

For sexual species, compensatory evolution has only been well investigated in diazinon-resistant populations of the Australian sheep blowfly, Lucilia cuprina (e.g. McKenzie et al., 1982; McKenzie, 1993). Following the origin of the mutation, resistant populations demonstrated the dramatic recovery of population-level fitness. However, these studies did not explicitly consider sex differences in fitness depression or modes of recovery.

Here we develop the hypothesis that mutational impacts on fitness should be sexually asymmetric and dependent on the degree and nature of sexual selection operating in a population. We are motivated by observations of fitness in outbred laboratory marker stocks of Drosophila melanogaster and the observation of evolutionary recovery of fitness over time. The specific population was artificially fixed for a highly deleterious wing mutation (nub1) and allowed to evolve for up to 180 generations under identical conditions to its ancestor base stock. We were able to recreate the origin of the mutation in the ancestor population by introgressing the mutation into the same base population later on. By contrasting the evolved population with its naïve, simulated ancestor and the wild-type benchmark, we provide preliminary data suggesting a sexual bias in mutational effects and modes of compensatory evolution.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

Stocks

The wild-type base population (LHM) has been described elsewhere (Chippindale et al., 2001). Briefly, LHM is a large outbred population that has been adapting to the lab since 1988 with juvenile census sizes of at least 9000–12 000 and adult census sizes of 1800–2000 per generation. For the deleterious mutation, we used the autosomal recessive nub1 mutation, located on the left arm of chromosome II (Lindsley & Zimm, 1992; Ng et al., 1995). This mutation dramatically minimizes the halteres, partially deletes the wing hinge and reduces the wing blades to tiny stumps. The mutation is pleiotropic, with expression in at least 22 body tissues (http://flybase.bio.indiana.edu/.bin/fbidq.html?FBal0013178).

Creation of mutant populations: backcrossing protocol

We replaced the genetic background of the original inbred nub stock from the Bloomington Stock Center with that of the wild-type LHM population through serial backcrosses beginning in 1995. Each cycle of backcrossing consisted of mating mutant males to virgin wild-type females to create F1 hybrids, F1 random mating and recovery of recombinant F2 nub homozygotes. At least one backcrossing cycle was performed using F2 virgin mutant females and wild-type males to replace Y chromosomes. This process involved 1000–1300 adults per generation. In January 1997, after 20 cycles, the mutant population was considered effectively genetically identical to the wild-type population (i.e. less than 1 × 10−4% different for the autosomes, less than 1 × 10−7% different for the X chromosomes) with the exception of the nub1 mutation and perhaps very closely linked regions of chromosome II. This population was then allowed to evolve with the nub1 allele fixed and is referred to as the ‘E-nub’ (evolved nub) population. E-nub was maintained under identical culture conditions in parallel with the wild-type population at a census population size of 1280 adults per generation.

A second nub1 backcross population, ‘X-nub’, was created in July 2002 from a copy of the E-nub population. This population underwent six cycles of backcrossing to the LHM population before the first fitness experiment. X-nub was therefore 98.4% identical to LHM for the autosomes, 99.6% for X chromosomes and 100% identical for Y chromosomes, on average. The population was continuously backcrossed throughout experiments so that it never evolved more than one generation before assay. This X-nub population therefore acted as a simulated, unevolved ancestor to the evolved E-nub population.

Fitness assays

Assays were designed to measure net fitness under conditions that closely replicated normal culture conditions. The first fitness assay was conducted after the E-nub population had evolved for 150 generations and the X-nub individuals were taken from the sixth F2 generation of backcrossing. Subsequent assays were performed at E-nub generations 160, 170 and 180.

For each assay, females from each of the three populations (LHM, E-nub and X-nub) were allowed to oviposit on 35 mm plates of food medium for 12–14 h. Standard (moderate) density vials were set up with exactly 90 wild-type eggs and 90 nub (E- or X-) eggs. Eggs were counted using a fine brush lubricated with an isotonic saline solution and were transferred to each vial. Juvenile fitness (viability) was measured as the percentage of eggs that survived to adult eclosion censused 11 days following oviposition. 40 vials were initiated for each contrast in each experiment.

After scoring juvenile fitness, half of the vials from each contrast were used to create adult male competition chambers. Adults were transferred into half-pint bottles containing standard medium and 40 mg of yeast, which simulated the reduction in density imposed on populations during normal culture. After 2.5 days, 30 nub females were isolated from each bottle and individually transferred into test tubes containing food medium, where they were allowed to oviposit for 16 h before being removed. Progeny were scored and counted after all adults had eclosed (11–13 days later). Male fitness for each population was measured as the percent of offspring fathered across the 30 females (paternity success). The proportion of females mated and offspring numbers per brood were also recorded. The use of nub females allowed paternity to be easily assigned, as the nub1 mutation is recessive. For each contrast, 20 adult competition chambers were established, with 600 broods (approximately 15 000 offspring) scored to estimate male mating success.

The remaining juvenile fitness vials were used to test female fecundity. After scoring juvenile survival, three vials containing food medium and 10 mg of yeast per vial were created using eight pairs of wild-type flies and eight pairs of nub flies for each contrast. Two days later, females were isolated to create four unyeasted vials containing four nub females or four wild-type females from each vial. The females were allowed to oviposit for 15 h and female fecundity was estimated as the mean number of eggs laid per female. Thus, for each fitness contrast, the fertility of 480 females of each population was assayed in 120 groups of four females.

Reciprocal mating assay

This assay was conducted to differentiate between the effects of male competition and female choice on male fertility. The experiment began by creating separate vials with 180 ± 20 eggs from the three populations (wild-type, E-nub, X-nub), as described above. Nine to ten days later, virgin males and females from the two nub populations were collected, along with virgin wild-type males. Males were collected at eight per vial and females at 12 per vial. By combining vials we created the following orthogonal contrasts: (i) wild-type and E-nub males competing for E-nub females, (ii) wild-type and E-nub males competing for X-nub females, (iii) wild-type and X-nub males competing for E-nub females and (iv) wild-type and X-nub males competing for X-nub females. Across four independent replicates, 80 vials were made for each contrast. Adults were allowed to interact for 2.5 days under standard conditioning with 7.5 mg of yeast, at which point 10 nub females were isolated from each vial and placed into test tubes. Females were allowed to oviposit for 16 h, and progeny were scored 11–13 days later, with male fertility estimated as in the fitness assays.

Statistical analysis

All data were initially tested for normality using the Shapiro–Wilk test. When necessary, proportional data underwent arcsine transformations and count or relative fitness data underwent logarithmic transformations to normalize their distribution. The basal level of replication in these assays was the fitness vial, with fitness measured on these subpopulations rather than individuals. The next level of replication was the four independent assays. Data pooled across multiple assays used a two-factor, fixed-effect anova model, with assay, population and their interaction as factors. The Tukey–Kramer HSD post-hoc test was used to identify the differences in means. Analyses were carried out using JMP 5.0.1 (SAS Institute, Cary, NC, USA) on Macintosh computers.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

For all comparisons reported, we found significant results whether tested in each individual assay or over all four assays via factorial anova. Here we report the latter because for no comparison of relative fitness was there a significant effect of experimental generation or interaction between experiment and selection treatment (population).

Initial effects of the nub1 mutation on fitness

We found that the nub1 mutation significantly reduced juvenile and adult fitness when it was first placed in the wild-type genetic background. Juvenile viability of the X-nub population averaged 79.2%, 6% lower than its wild-type control (Fig. 1a; F1,240 = 49.81, P < 0.0001). Female fecundity was significantly reduced by the mutation with a mean decline of 13.1 eggs/female in the 15 h period (Fig. 1b; F1,146 = 402.65, P < 0.0001), and X-nub male paternity success was 74% lower than the control population (Fig. 1c; F1,130 = 1839.23, P < 0.0001). Thus, while both sexes experienced marked reductions in fertility, the nub1 mutation had sex-specific effects: adult X-nub males were 36% less fit than females relative to their respective wild-type LHM competitors (F1,138 = 272.50, P < 0.0001).

image

Figure 1. Mean (±1 SE) juvenile and adult fitness measures for evolved E-nub (across 150, 160, 170 and 180 generations) and unevolved X-nub populations. Fitness measures are relative to the wild-type (LHM) competitor (shown with a dashed line). Numbers above the bars indicate the actual values obtained for (a) juvenile viability (percent survival from egg 11 days following oviposition), (b) female fecundity (mean number of eggs laid per female in 15 h) and (c) male fertility (percent of offspring fathered across 30 nub females).

Download figure to PowerPoint

Evidence for fitness recovery

Relative to the simulated ancestor (X-nub), the E-nub population had significantly higher juvenile viability (Fig. 1a; F1,242 = 65.65, P < 0.0001), female fecundity (Fig. 1b; F1,146 = 49.73, P < 0.0001), and male paternity success (Fig. 1c; F1,133 = 96.46, P < 0.0001). The juvenile viability of the E-nub population was slightly but significantly greater than LHM (Fig. 1a; F1,244 = 21.34, P < 0.0001). In contrast, the nub1 mutation continued to have deleterious fitness effects on adults, as both females and males were approximately 20% less fit than the wild-type LHM population (females: F1,146 = 116.84, P < 0.0001; males: F1,136 = 54.04, P < 0.0001). The fact that fitness recovered to approximately the same level in both sexes (F1,141 = 0.79, P = 0.39) indicates that males recovered fitness to a substantially greater degree than females did. The E-nub population may have reached a plateau in its selection response, as there was no detectable increase in the relative fitness of either sex from the generation 150–180 assays (both regressions of relative fitness on time N.S.).

Components of male fitness

The two components of male paternity success assayed, the percent of females mated and number of offspring per brood, are illustrated in Fig. 2. The nub1 mutation was found to reduce the courtship success (Fig. 2a) of X-nub males, as they mated with 65% fewer females than their LHM competitors (F1,130 = 1031.00, P < 0.0001). The evolved E-nub males only partially recovered from this courtship hindrance, as they mated with 28% more females than the X-nub males (F1,133 = 56.88, P < 0.0001), still 37% fewer females than wild-type competitors (F1,136 = 179.76, P < 0.0001).

image

Figure 2. Means (±1 SE) for two components of male fitness: (a) percent of females mated and (b) number of offspring per brood, measured for evolved E-nub (across 150, 160, 170 and 180 generations) and unevolved X-nub populations. Fitness measures are relative to those of wild-type (LHM) competitors for each trait, as indicated with a dashed horizontal line; actual values for each trait are indicated above the bars.

Download figure to PowerPoint

The nub1 mutation also reduced the number of offspring per brood (Fig. 2b), as those fathered by X-nub males contained 16% fewer individuals than those fathered by wild-type LHM males (F1,120 = 21.88, P < 0.0001). In contrast, the E-nub population had substantially increased numbers of offspring per brood, with those fathered by E-nub males containing over 38% more offspring than those fathered by LHM males (F1,136 = 55.24, P < 0.0001). E-nub broods also contained more offspring than X-nub broods (F1,128 = 88.86, P < 0.0001).

Reciprocal mating tests

The results of the reciprocal mating assay are shown in Fig. 3. Significant differences in the male paternity success of the four contrasts were recorded (Fig. 3a; F3,315 = 3.32, P < 0.05). There was no difference in X-nub male fertility when they were housed with either X- or E-nub females, suggesting E-nub females continued to discriminate against wingless males. When E-nub males competed for mating with X-nub females, they had significantly higher fertility than X-nub males did, although the increase was only about 3%. This suggests E-nub males only slightly increased their success in sexual selection independent of female choice. When E-nub males competed for mating with their coevolved E-nub females, their fertility was significantly greater than that of X-nub males. However, E-nub male fertility with coevolved mates was also significantly greater (21%) than with X-nub females, suggesting female co-evolution played a major role in the recovery of E-nub male fitness.

image

Figure 3. The results of the reciprocal mating assay, which combined males and females from the evolved E-nub and unevolved X-nub populations. Means (±1 SE) for (a) male fertility and (b) number of offspring per brood are measured relative to wild-type (LHM) male competitors. Male/female combinations are indicated on the x-axis in the format ‘male × female’ (E: E-nub; X: X-nub).

Download figure to PowerPoint

The two major components of male fertility, the percent of females mated and number of offspring per brood, were also compared across the four treatment groups. This assay found no differences in the proportion of females mated based on population (F3,315 = 0.17, P = 0.91). When the numbers of offspring per brood were compared, the combination of E-nub male and E-nub female produced significantly more offspring per brood than the others (Fig. 3b; F3,286 = 10.08, P < 0.0001). These results suggest that the E-nub males evolved substantially increased fertility through post-copulatory effects specific to their coevolving mates.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

The present work is an opportunistic experiment in evolution. The original mutant E-nub population was created for other purposes (as a competitor stock) and was not replicated at the population level. As a result, our adaptive inferences are limited. Nonetheless, we suggest that the main outcome of this experiment is consistent with compensatory evolution: we measured dramatic improvements in fitness when E-nub was allowed to evolve with a debilitating mutation. Here we discuss the implications of these findings, their specific manifestations and the utility of our approach for the study of mutation and compensatory evolution.

Utility of the backcross approach

Strong reductions in fitness are often associated with visible mutations in Drosophila (e.g. Bateman, 1948; Ludwin, 1951; Geer & Green, 1962) but such observations are often confounded experimentally with inbreeding depression because of small population sizes in typical ‘tester’ stocks. By serially backcrossing the mutation into a large outbred population and preventing co-adaptation, we were able to estimate additive mutational effects across many genetic backgrounds. In our experiments, fitness loss occurred for all measured traits in both juvenile and adult life-stages after the nub1 mutation was experimentally introduced. At the same time, installing the mutation in a variable sexual population may have accelerated the rate of evolutionary compensation compared to what would occur in a small isolated population, or an asexual population that must wait for mutations at modifier loci. The backcrossing approach we applied therefore has the potential to capture, in a controlled experiment, some of the long-term dynamics of small natural populations that would be prone to fixation of deleterious alleles. These findings also have implications for the common practice of using tester stocks as benchmarks for Drosophila fitness assays; the experimenter must be prepared for the fact that these stocks are unlikely to be static, potentially evolving in ways that complicate experimental interpretations.

Evidence for asymmetric fitness costs and recovery

We found that the introduction of the nub1 mutation sharply reduced juvenile viability under competitive conditions, relative to the wild-type population (LHM) or the evolved mutant population (E-nub). In fact, viability of the E-nub population recovered and slightly surpassed the wild-type population from which it was derived. Similarly, the mutation had strong initial effects on adult fitness (fecundity, fertility and mating success), and both sexes exhibited substantial recovery of fitness. However these effects on the two sexes were highly asymmetrical. In females, fecundity was initially reduced to approximately 62% that of the wild-type population, while male paternity success was reduced to only 26% of wild-type. Since both sexes recovered to approximately the same level relative to the wild-type control (80%), both the initial effect and degree of recovery were stronger in males.

We suggest that this may be a general result for promiscuous species like D. melanogaster. If males are subjected to strong sexual selection in addition to the general impact of the mutation on naturally selected traits, then they will (1) be more strongly affected by mutations and (2) have more potential avenues of fitness recovery. For example, mutant males in this species are known to have difficulty in obtaining mates when competing with wild-type males, whereas mutant females are courted relatively indiscriminately (Spieth, 1974). These differences will be magnified when the target of mutation is an instrument of sexual selection. Drosophila wings are integral to courtship displays and songs (Ewing, 1964; Spieth, 1974; Greenspan & Ferveur, 2000), so it is not surprising that nub1, which dramatically reduces wing area, hinders males more than females. If we assume that the physiological effects of the mutation are similar for the two sexes, we can estimate that E-nub males increased their fitness by approximately 38% through sexually selected traits alone, suggesting that the two modes of selection in combination resulted in greater fitness amelioration.

Sexual selection and intersexual co-evolution

Despite the experimental removal of the main organs of male courtship, we estimated that E-nub males experienced a 38% gain in net fitness through improved success in sexual selection. However, nub male net fitness was measured in competition for females from their own population. As a result, we could not distinguish between changes in male competitive abilities and changes in female preference functions in response to mutant males. We used a sexual selection experiment with a reciprocal cross design to tease apart these alternatives in the E-nub and X-nub populations.

In this assay, both evolved and unevolved mutant males competed with wild-type males for mating with females from each mutant population. The power of this experiment lies in the fact that backcrossed (X-nub) females are over 99% genetically identical to wild-type females, and are therefore expected to have the same set of preference functions. In contrast, evolved (E-nub) females may have developed preferences for the males they coevolved with. If E-nub males continued to display high fitness when competing for unevolved X-nub females, this would suggest male fitness increased independently of female choice. If X-nub males performed better when mated with E-nub females than with their own females, this would indicate a decline in E-nub female discrimination against wingless males. The latter result would be comparable to Cook (1973), who found increased female receptivity to wingless males after 30 generations of exposure.

The results suggest both male- and female-driven increases in E-nub male fitness. First, there were no significant differences in unevolved (X-nub) male fitness regardless of which female was present, suggesting E-nub females do not have a preference for (or decreased discrimination against) the nub1 male phenotype itself. Second, E-nub males displayed evidence of generalized improvement in sexually selected traits, as they were significantly more successful than X-nub males with wingless females that possess wild-type preferences. This increase was modest, however, as E-nub males only experienced a 3% increase in paternity success over X-nub males. It appears that changes in female choice played a role in the fitness increase of E-nub males, as their fitness was strongly enhanced only when they competed for coevolved females. Therefore both novel male signals and novel female preferences appear to have contributed to the increased fitness of the E-nub males relative to the unevolved X-nub males.

To our surprise, the improvement in the behavioural component of E-nub male fertility accounted for only 28% of the selection response, while the total improvement in male fitness was 55%. Since male seminal fluid effects on female oviposition rate and survival are well known in Drosophila (Chapman et al., 1995; Rice, 1996; Wolfner, 1997), we further examined the data for evidence of post-copulatory male effects on female fertility by measuring the number of offspring per brood fathered by the different types of males. When mated with their own females, unevolved (X-nub) males sired 16% fewer offspring per brood than wild-type males mated to females from the same population. On the other hand, when we examined offspring counts of E-nub females mated to E-nub males, we found that they contained approximately 38% more offspring than those fathered by wild-type males. These effects are only slightly influenced by differences in viability, and therefore appear to relate to the induction of female egg production by mating. Interestingly, the induction of higher fertility by the evolved males disappeared when they were combined with unevolved (X-nub) females. It therefore appears that, as for mating success, substantial intersexual co-evolution for post-mating effects occurred in the E-nub population.

General conclusions

We introgressed a debilitating mutation into an outbred genetic background through extensive backcrossing and allowed the population to evolve (E-nub). We then contrasted the evolved population with two other treatments: the wild-type stock and a newly created unevolved population fixed for the same mutation (X-nub). The results of these comparisons allowed us to evaluate the impact of the mutation when fixed in a diverse genetic background (i.e. its additive genetic effects) and the apparent compensatory response for those effects in light of the wild-type benchmark. While we assert that this is a powerful experimental approach, we also acknowledge the shortcomings of our particular design. These data represent point estimates, limiting our ability to argue that adaptation has occurred. However, most of the results presented here cannot be easily explained by other evolutionary mechanisms, such as genetic drift or inbreeding depression. The rapid and extensive fitness recovery from the harming effects of the nub1 mutation rather suggests the operation of selection.

Because we lack replication at the population level (i.e. populations within treatments), we wish to motivate a general hypothesis rather than definitively test it. We suggest that mutations should have asymmetrical effects on the sexes whenever they experience different selective forces underlying fitness. In support of this hypothesis, the nub1 mutation was more costly to males than to females, presumably because male D. melanogaster are more strongly sexually selected and wings are still essential to their fitness under laboratory conditions. This asymmetry appears to have produced different avenues of compensatory change, with males being more influenced by sexual selection and co-adaptation than females were. Furthermore, it is possible that the apparent compensation resulted from evolutionary tradeoffs, with the energy that would have previously been devoted to flight muscle and wing production being rerouted to other areas that would be of most benefit. Whatever the particular mechanism, mutational effects on fitness and the dynamics of compensatory evolution appear to be strongly affected by gender. Because most foregoing work in this area has been with asexual microbes, these findings show the potential for future studies examining the sexual specificity of mutational effects and compensatory evolution using the approach described here.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References

We thank T. Long for lab assistance and comments on the manuscript, R. Ferrier for technical help early on, Tufty the squirrel and the rest of the Superfly lab for help later on. This research was supported by NSERC Discovery Grant funding to AC.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References