14.3: Frequency-Dependent Selection - Biology

In frequency-dependent selection, phenotypes that are either common or rare are favored through natural selection.

Learning Objectives

  • Describe frequency-dependent selection

Key Points

  • Negative frequency -dependent selection selects for rare phenotypes in a population and increases a population’s genetic variance.
  • Positive frequency-dependent selection selects for common phenotypes in a population and decreases genetic variance.
  • In the example of male side-blotched lizards, populations of each color pattern increase or decrease at various stages depending on their frequency; this ensures that both common and rare phenotypes continue to be cyclically present.
  • Infectious agents such as microbes can exhibit negative frequency-dependent selection; as a host population becomes immune to a common strain of the microbe, less common strains of the microbe are automatically favored.
  • Variation in color pattern mimicry by the scarlet kingsnake is dependent on the prevalence of the eastern coral snake, the model for this mimicry, in a particular geographical region. The more prevalent the coral snake is in a region, the more common and variable the scarlet kingsnake’s color pattern will be, making this an example of positive frequency-dependent selection.

Key Terms

  • frequency-dependent selection: the term given to an evolutionary process where the fitness of a phenotype is dependent on its frequency relative to other phenotypes in a given population
  • polygynous: having more than one female as mate

Frequency-dependent Selection

Another type of selection, called frequency-dependent selection, favors phenotypes that are either common (positive frequency-dependent selection) or rare (negative frequency-dependent selection).

Negative Frequency-dependent Selection

An interesting example of this type of selection is seen in a unique group of lizards of the Pacific Northwest. Male common side-blotched lizards come in three throat-color patterns: orange, blue, and yellow. Each of these forms has a different reproductive strategy: orange males are the strongest and can fight other males for access to their females; blue males are medium-sized and form strong pair bonds with their mates; and yellow males are the smallest and look a bit like female, allowing them to sneak copulations. Like a game of rock-paper-scissors, orange beats blue, blue beats yellow, and yellow beats orange in the competition for females. The big, strong orange males can fight off the blue males to mate with the blue’s pair-bonded females; the blue males are successful at guarding their mates against yellow sneaker males; and the yellow males can sneak copulations from the potential mates of the large, polygynous orange males.

In this scenario, orange males will be favored by natural selection when the population is dominated by blue males, blue males will thrive when the population is mostly yellow males, and yellow males will be selected for when orange males are the most populous. As a result, populations of side-blotched lizards cycle in the distribution of these phenotypes. In one generation, orange might be predominant and then yellow males will begin to rise in frequency. Once yellow males make up a majority of the population, blue males will be selected for.Finally, when blue males become common, orange males will once again be favored.

An example of negative frequency-dependent selection can also be seen in the interaction between the human immune system and various infectious microbes such as pathogenic bacteria or viruses. As a particular human population is infected by a common strain of microbe, the majority of individuals in the population become immune to it. This then selects for rarer strains of the microbe which can still infect the population because of genome mutations; these strains have greater evolutionary fitness because they are less common.

Positive Frequency-dependent Selection

An example of positive frequency-dependent selection is the mimicry of the warning coloration of dangerous species of animals by other species that are harmless. The scarlet kingsnake, a harmless species, mimics the coloration of the eastern coral snake, a venomous species typically found in the same geographical region. Predators learn to avoid both species of snake due to the similar coloration, and as a result the scarlet kingsnake becomes more common, and its coloration phenotype becomes more variable due to relaxed selection. This phenotype is therefore more “fit” as the population of species that possess it (both dangerous and harmless) becomes more numerous. In geographic areas where the coral snake is less common, the pattern becomes less advantageous to the kingsnake, and much less variable in its expression, presumably because predators in these regions are not “educated” to avoid the pattern.

Negative frequency-dependent selection serves to increase the population’s genetic variance by selecting for rare phenotypes, whereas positive frequency-dependent selection usually decreases genetic variance by selecting for common phenotypes.

Brisson Lab

Russell, A, MA Prusinski, J Sommer, C O’Connor, J White, R Falco J Kokas, V Vinci, W Gall, K Tober, J Haight, J Oliver, L Meehan, LA Sporn, D Brisson, PB Backenson. Accepted. Epidemiology and spatial emergence of Anaplasmosis, New York State, USA, 2010-2018. Emerging Infectious Diseases.

O’Connor, C, MA Prusinski, S Jiang, A Russell, J White, R Falco, J Kokas, V Vinci, W Gall, K Tober, J Haight, J Oliver, L Meehan, LA Sporn, D Brisson, PB Backenson. Accepted. A comparative spatial and climate analysis of human Granulocytic Anaplasmosis and human Babesiosis in New York State (2013-2018). Journal of Medical Entomology.

Tran, T, MA Prusinski, JL White, RC Falco, V Vinci, WK Gall, K Tober, J Oliver, LA Sporn, L Meehan, E Banker, PB Backenson, ST Jensen, D Brisson. 2021. Spatio-temporal variation in environmental features predicts the distribution and abundance of Ixodes scapularis. International Journal of Parasitology 51(4) , 311–320 . doi: 10.1016/j.ijpara.2020.10.002 Tran2021.EnvTicks.pdf

MacDonald, H, E Akcay, D Brisson. 2021. The role of host phenology for parasite transmission. Theoretical Ecology 14: 123–143. doi: 10.1007/s12080-020-00484-5 MacDonald2020.HostPhenologyTransmission.pdf

Oppler, ZJ, KR O’Keeffe, KD McCoy, D Brisson. 2020. Evolutionary Genetics of Borrelia in DS Samuels and JD Radolf ( Eds.), Lyme Disease and Relapsing Fever Spirochetes , Norwich, UK: Caister Academic Press

Brisson, D. 2020. The ecology of evolutionary transitions to multicellularity. Peer Community in Evolutionary Biology, 100099. 10.24072/pci.evolbiol.100099 **recommendation of Rose et al. 2020. Meta-population structure and the evolutionary transition to multicellularity (2020), bioRxiv, 407163, ver. 5 peer-reviewed by Peer Community in Evolutionary Biology. 10.1101/407163 **

Berry ASF, RS Salazar-Sánchez, R Castillo-Neyra, K Borrini-Mayorí, C Arevalo-Nieto, C Chipana-Ramos, M Vargas-Maquera, J Ancca-Juarez, C Náquira-Velarde, MZ Levy, D Brisson. 2020. Dispersal patterns of Trypanosoma cruzi in Arequipa, Peru. PLoS Neglected Tropical Diseases 14(3): e0007910. Berry2020.pdf

Berry ASF, RS Salazar-Sánchez, R Castillo-Neyra, K Borrini-Mayorí, C Chipana-Ramos, M Vargas-Maquera, J Ancca-Juarez, C Náquira-Velarde, MZ Levy, D Brisson. 2019. Sexual reproduction in a natural Trypanosoma cruzi population. PLoS Neglected Tropical Diseases 13(5): e0007392. Berry2019a.Meiosis.pdf

Berry ASF, RS Salazar-Sánchez, R Castillo-Neyra, K Borrini-Mayorí, C Chipana-Ramos, M Vargas-Maquera, J Ancca-Juarez, C Náquira-Velarde, MZ Levy, D Brisson. 2019. Immigration and establishment of urban Trypanosoma cruzi populations in Arequipa, Peru. PLoS ONE 14(8): e0221678. Berry2019b.PloSONE

B />risson, D. 2018. Combining epidemiological models with statistical inference can detect parasite interactions . Peer Community in Ecology, 100006 10 Oct 2018 **recommendation of Alizon et al. 2018. Detecting within-host interactions using genotype combination prevalence data. Peer Community in Ecology, bioRxiv, 256586, ver. 3 **

Ostfeld, RS, D Brisson, K Oggenfuss, J Devine, MZ Levy, and F Keesing. 2018. Effects of a zoonotic pathogen, Borrelia burgdorferi, on the behavior of a key reservoir host. Ecology and Evolution 8:4074–4083 Ostfeld2018a.pdf

Khatchikian CE, Nadelman RB, Nowakowski J, Schwartz I, Wormser GP, and Brisson D. 2017 The impact of strain specific immunity on Lyme disease incidence is spatially heterogeneous. Diagnostic Microbiology and Infectious Disease 89: 288–293.Khatchikian2017a.pdf

Brisson, D. 2017. Negative frequency-dependent selection is frequently confounding. bioRxiv 113324, ver. 3 of 20 June 2017. **peer-reviewed and recommended by Peer Community In Evolutionary Biology (Bravo IG. 2017. Unmasking the delusive appearance of negative frequency-dependent selection. Peer Community in Evolutionary Biology, 100024.10.24072/pci.evolbiol.100024)**

Clarke, EL, SA Sundararaman, SN Seifert, FD Bushman, BH Hahn, and D Brisson. 2017. swga: A primer design toolkit for selective whole genome amplification. Bioinformatics 33 (14): 2071-2077. Clarke2017.pdf

Springer YP, et al. 2016. Continental scale surveillance of infectious agents: Tick-, mosquito-, and rodent-borne parasite sampling designs for NEON. Ecosphere. 7(5):e01271. springer_et_al-2016-ecosphere

Sundararaman SA, Plenderleith LJ, Liu W, Loy DE, Learn GH, Li Y, Shaw KS, Ayouba A, Peeters M, Speede S, Shaw GM, Bushman FD, Brisson D, Rayner JC, Sharp PM, Hahn BH. 2016. Genomes of cryptic chimpanzee Plasmodium species reveal key evolutionary events leading to human malaria. Nature Communications. 7:11078. sundararaman2016

The emergence, maintenance, and demise of diversity in a spatially variable antibiotic regime

Antimicrobial resistance (AMR) is a growing global threat that, in the absence of new antibiotics, requires effective management of existing drugs. Here, we use experimental evolution of the opportunistic human pathogen Pseudomonas aeruginosa to explore how changing patterns of drug delivery modulates the spread of resistance in a population. Resistance evolves readily under both temporal and spatial variation in drug delivery and fixes rapidly under temporal, but not spatial, variation. Resistant and sensitive genotypes coexist in spatially varying conditions due to a resistance-growth rate trade-off which, when coupled to dispersal, generates negative frequency-dependent selection and a quasi-protected polymorphism. Coexistence is ultimately lost, however, because resistant types with improved growth rates in the absence of drug spread through the population. These results suggest that spatially variable drug prescriptions can delay but not prevent the spread of resistance and provide a striking example of how the emergence and eventual demise of biodiversity is underpinned by evolving fitness trade-offs.

Table S1. Mixed linear analysis of covariance for maximum growth rate in LB.

Table S2. Mixed linear analysis of covariance for relative fitness (ω) of resistant types, with random effect of isolate pair nested in population.

Table S3. Mixed linear analysis of covariance for maximum growth rate in [0.3 μg/mL] ciprofloxacin.

Table S4. Mixed linear analysis of covariance for relative fitness (ω) of resistant types, with random effect of isolate pair nested in population.

Fig. S1. SPAT selection regime.

Fig. S2. Coexistence of susceptible and resistant types maintained in SPAT treatment.

Fig. S3. Negative frequency-dependent selection for select pairs of resistant and sensitive isolates.

Fig. S4. Productivity of drug-containing and drug-free patches become similar by days 20 and 40.

Fig. S5. Plot of (final frequency of resistance – initial frequency of resistance) versus initial frequency of resistance.

Fig. S6. Negative frequency-dependence using Chevin fitness (maximum growth rate of evolved isolate - ancestor) versus log10MIC.

Fig. S7. Negative frequency-dependence using assumed starting frequencies.

Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Plasticity-led evolution: evaluating the key prediction of frequency-dependent adaptation

Plasticity-led evolution occurs when a change in the environment triggers a change in phenotype via phenotypic plasticity, and this pre-existing plasticity is subsequently refined by selection into an adaptive phenotype. A critical, but largely untested prediction of plasticity-led evolution (and evolution by natural selection generally) is that the rate and magnitude of evolutionary change should be positively associated with a phenotype's frequency of expression in a population. Essentially, the more often a phenotype is expressed and exposed to selection, the greater its opportunity for adaptive refinement. We tested this prediction by competing against each other spadefoot toad tadpoles from different natural populations that vary in how frequently they express a novel, environmentally induced carnivore ecomorph. As expected, laboratory-reared tadpoles whose parents were derived from populations that express the carnivore ecomorph more frequently were superior competitors for the resource for which this ecomorph is specialized—fairy shrimp. These tadpoles were better at using this resource both because they were more efficient at capturing and consuming shrimp and because they produced more exaggerated carnivore traits. Moreover, they exhibited these more carnivore-like features even without experiencing the inducing cue, suggesting that this ecomorph has undergone an extreme form of plasticity-led evolution—genetic assimilation. Thus, our findings provide evidence that the frequency of trait expression drives the magnitude of adaptive refinement, thereby validating a key prediction of plasticity-led evolution specifically and adaptive evolution generally.

1. Introduction

Phenotypic plasticity is commonplace [1,2], but whether and how it impacts evolution is controversial [3–5]. An evolutionary process in which plasticity has long been implicated is the origins of novel, complex phenotypes (e.g. [2,5–10]).

According to the ‘plasticity-led evolution’ hypothesis (sometimes dubbed ‘plasticity-first evolution’ [11,12]), a novel complex phenotype first appears in a rudimentary form when the phenotype (or its components) is expressed via plasticity following a change in environment. Such environmental change is typically stressful, and organisms can mitigate this stress by using plasticity to facultatively produce a phenotype better matched to the new environment. If underlying genetic variation exists in either the tendency or manner in which individuals respond to this environmental change (as is nearly always the case [13]), then selection can act on these ‘reaction norms’ and improve the phenotype's functionality by altering the phenotype's form. Moreover, selection can also promote a change in the phenotype's regulation. Specifically, depending on whether or not plasticity is favoured [14,15], selection can favour either increased environmental sensitivity—which might ultimately maintain the new phenotype as part of a ‘polyphenism’ [1]—or decreased environmental sensitivity—which might ultimately cause the plasticity to be lost and the phenotype to become canalized through ‘genetic assimilation’ (sensu [16]). Essentially, plasticity-led evolution occurs when selection promotes an adaptive change in an initially environmentally induced phenotype's form and/or regulation. Thus, plasticity itself can evolve (as has been long recognized (e.g. [13,15,17–19])), and, consequently, this evolution can facilitate the origin of a novel, complex phenotype.

Although laboratory studies support these ideas [16,20], and there are suggestive field studies (reviewed in [9,12]), many researchers remain sceptical of whether plasticity can facilitate evolution [3,4]. Such scepticism arises, in part, because the key criteria and predictions of the plasticity-led evolution hypothesis have not been made clear and evaluated in natural populations [3,4]. To address this concern, we [12] recently outlined four key criteria for testing this hypothesis, one of which (criterion 4) is that the focal trait should exhibit evidence of having undergone adaptive refinement as it is induced and exposed to selection repeatedly. Although this criterion is seldom validated, doing so is essential to rule out alternative explanations [12].

Moreover, of the few studies that have tested criterion 4 (cited in [12]), none have tested its critical, underlying prediction: that the rate and magnitude of phenotypic change should be positively associated with a phenotype's frequency of expression or use in a population [2,12,21,22]. This prediction is, in turn, rooted in two assumptions: (i) that individuals in ancestral lineages (where a rudimentary version of the focal trait is produced through plasticity) should express the trait less frequently than individuals in derived lineages (where the trait may be canalized) and (ii) that a trait in a population in which it is expressed (and exposed to selection) more frequently should evolve greater and more rapid refinement [2]. Essentially, during plasticity-led evolution, as an environmentally induced phenotype is recurrently produced (e.g. by persistent selection pressure favouring that phenotype), it will be exposed to selection more frequently and therefore have greater opportunity for adaptive refinement.

This notion that the frequency of trait expression drives the magnitude of adaptive refinement is a critical prediction not only of plasticity-led evolution, but also of evolution by natural selection more generally. Yet, ‘frequency-dependent adaptation’ has rarely been demonstrated empirically (but see [23–25]). (Note that frequency-dependent adaptation is a separate, albeit related, process from frequency-dependent selection, which arises when the fitness of an individual phenotype depends on its frequency in the population. Unlike frequency-dependent adaptation, frequency-dependent selection has been thoroughly studied e.g. [26–28]). However, indirect support for frequency-dependent adaptation comes from studies: (i) using reciprocal transplants that demonstrate adaptation to local (i.e. frequently experienced) conditions and maladaptation to alternative conditions (e.g. [29–31]) (ii) of clinal variation in adaptation that have shown a pattern of changing phenotype ratios (including environmentally induced phenotypes) along the cline such that the greatest divergence occurs at the clinal extremes (e.g. [32–34]) and (iii) exploring adaptive radiation where generalist or plastic ancestors experience greater specialization over time (e.g. [35–38]).

Here, we perform an explicit empirical test of frequency-dependent adaptation. We do so by focusing on amphibian populations that have diverged in production of a novel, environmentally induced ecomorph. If the frequency of trait expression does indeed determine the degree to which that phenotype is refined by selection, then individuals from populations that produce this ecomorph more frequently should be superior competitors for the resource on which this ecomorph specializes. As we describe below, our findings are consistent with this expectation.

2. Material and methods

(a) Study subjects

We studied plains spadefoot toads, Spea bombifrons, from natural populations in the western USA. In many parts of its range, S. bombifrons has evolved a larval polyphenism in which it produces two, environmentally induced, resource-use ecomorphs (see the electronic supplementary material, figure S1): (i) omnivores, which are dietary generalists that feed mostly on detritus and small plankton, and which are normally produced by default and (ii) carnivores, which are dietary specialists that feed on, and are induced by the consumption of, anostracan fairy shrimp or other tadpoles [39–45]. In most populations, omnivores are the more frequently produced of the two ecomorphs (e.g. earlier studies that sampled diverse sites in allopatry found an average of 20% carnivores and 80% omnivores [41–43] see the electronic supplementary material, figure S2). However, in populations where S. bombifrons co-occurs with a congener (Spea multiplicata), these two species have undergone ecological character displacement, resulting in S. bombifrons producing nearly all carnivores (e.g. earlier studies found an average of 95% carnivores and 5% omnivores in sympatry [41–43] see the electronic supplementary material, figure S2). We refer to these S. bombifrons populations that occur with and without S. multiplicata as ‘sympatric’ and ‘allopatric’, respectively. Because the two species have come into secondary contact following range expansion by S. bombifrons [46,47], sympatric populations represent the ‘derived’ state, whereas allopatric populations represent the ‘ancestral’ state.

For the experiments below, we created 10 full sibships of S. bombifrons by breeding adults that were recently collected from diverse populations in allopatry (electronic supplementary material, table S1 and figure S2), all of which probably experience ongoing gene flow [46]: Colorado (four sibships), northern Nebraska (one sibship), southwestern Nebraska (two sibships), Oklahoma (two sibships) and Texas (one sibship). These 10 sibships constituted our allopatric animals. We also created eight full sibships of S. bombifrons by breeding adults that were recently collected from six populations in the San Simon valley of southeastern Arizona (where S. multiplicata is present electronic supplementary material, table S1 and figure S2), all of which probably experience ongoing gene flow [47]. These eight sibships constituted our sympatric animals. Breeding was induced by injecting adults with 0.04 ml luteinizing hormone-releasing hormone (Sigma L-7134) at a concentration of 0.01 µg µl −1 and leaving pairs overnight in separate nursery tanks. The next day, adults were removed, and the eggs from each sibship were kept in these tanks until they hatched.

(b) Testing whether frequency of trait expression predicts its adaptive refinement

We predicted that sympatric tadpoles would be superior competitors for shrimp—and therefore grow more on shrimp—compared to allopatric tadpoles. This is because S. bombifrons produces carnivores (the ecomorph that specializes on shrimp) more frequently in sympatry than in allopatry (see §2a). Conversely, we predicted that allopatric tadpoles would be superior competitors for detritus compared to sympatric tadpoles, because S. bombifrons produces omnivores (the ecomorph that uses detritus) more frequently in allopatry than in sympatry (see §2a).

For these tests, we had to give each tadpole a population-specific mark (to differentiate it from its tankmate e.g. see [48]). We therefore needed to grow tadpoles to a sufficient size to receive these marks. To do so, we randomly selected 95 tadpoles from each of 15 sibships (eight allopatric, seven sympatric) and placed them in an outdoor wading pool (1.5 m diameter) for 4 days (water temperatures approx. 30°C). At the start, each pool received 50 ml of plant-based fish food. After returning the tadpoles indoors, we placed each sibship in clean water and fed them 400 mg of fish food. Twenty-four hours later, we measured the snout–vent length (SVL) of these tadpoles and created our experimental units.

Each experimental unit consisted of three tanks (18 × 13 × 8.5 cm, filled with 1.2 l of dechlorinated water) containing: (i) a single allopatric tadpole, (ii) a single sympatric tadpole, and (iii) two tadpoles, one of which was a sibling of the allopatric tadpole, and the other of which was a sibling of the sympatric tadpole. The first two were dubbed ‘singleton tanks’, whereas the third was dubbed a ‘competition tank’. All tadpoles in each experimental unit were similar in SVL at the start (individuals varied by less than 2.5%). To distinguish between tankmates in the competition tanks, we injected pink elastomer [48] into the dorsal tail of one individual (equal numbers of allopatric and sympatric tadpoles were injected). All three tanks in each experimental unit were placed adjacent to each other.

Half of the competition tanks received daily 40 mg of crushed fish food (hereafter, ‘detritus’), which simulates in form and nutrition the detritus on which Spea omnivores feed in natural ponds [44]. The other half received twice daily 100 live brine shrimp (Artemia), which simulate the fairy shrimp (Thamnocephalus or Steptocephalus) on which Spea carnivores feed in natural ponds. Preliminary tests indicated that these amounts of detritus and shrimp induced competition i.e. food was completely eaten between feedings. Singleton tanks received half of these amounts thus, the per capita amounts of food provided to singleton and competition tanks were identical. All tanks experienced 50% water changes every other day. We had 51 replicate units per diet. After 10 days, we ended the experiment by euthanizing tadpoles in a 0.8% aqueous solution of tricaine methanesulfonate (MS-222) and preserving them in 95% ethanol.

We evaluated the predictions outlined at the start of this section in three ways. First, we used likelihood ratio tests to compare a series of mixed models. ‘Diet’ (i.e. detritus or shrimp) and ‘selective environment’ (i.e. allopatry or sympatry) were fixed categorical variables and ‘sibship’, ‘competitor sibship’ and ‘replicate’ were random effects. We compared a null model that contained only the random effects to single-factor models that retained the random effects and included either diet or selective environment as a fixed effect, and to two-factor models (with and without an interaction term). The ‘best model’ was determined if it was significantly better than all other models according to likelihood ratio tests (performed using ‘anova’ in R). The biological interpretation of each model is described in the electronic supplementary material.

Second, we performed a type III sum of square analysis of variance (ANOVA) on the interaction model to corroborate our observations from the above test. We also calculated the effect size (Cohen's d) between diets for each selective environment to determine if tadpoles from sympatry have experienced greater divergence in growth between diets (i.e. greater growth on shrimp and/or reduced growth on detritus) than tadpoles derived from allopatry. If there was a significant interaction, we performed post hoc multiple comparisons tests by grouping selective environment with diet (i.e. allopatry.shrimp, allopatry.detritus, etc.) and using the ‘pairwise.t.test’ function with ‘fdr’ correction in R.

Finally, for each competition tank, we categorized each tadpole as the ‘winner’ of competition if it grew more than its tankmate. We then performed one-tailed Fisher's exact tests to determine if the number of winners differed between selective environments. One-tailed tests were used because we had the a priori prediction that there would be more allopatric winners on detritus and more sympatric winners on shrimp (this a priori prediction was based on patterns of trait expression in nature see §2a). We also used Levene's tests to evaluate differences in the amount of variation in growth among selective environments and diets. If competitive differences between selective environments happen to be diet-dependent, then we would expect greater variation on one diet than on the other.

(c) Evaluating mechanisms of adaptive refinement

The results of the previous experiment revealed that (see §3a): (i) tadpoles from sympatry grew more than tadpoles from allopatry on shrimp, and (ii) tadpoles from allopatry grew more than tadpoles from sympatry on detritus. Based on previous work [41,42,44,48–52], we evaluated five, non-mutually exclusive mechanisms that could explain these differences between selective environments in competitive ability. We specifically tested whether tadpoles from the two selective environments have diverged in: (i) intrinsic growth rate, (ii) time budgets, (iii) trait integration, (iv) shrimp capture ability, or (v) trophic morphology. Each test is described in the electronic supplementary material.

(d) Testing for genetic assimilation of trophic morphology

Finally, we tested if tadpoles from sympatry developed more carnivore-like features, even in the absence of the cue that normally induces the carnivore morphology (ingestion of live shrimp or tadpoles). Finding such a pattern would suggest that trophic morphology has been genetically assimilated in sympatry. To perform this test, we randomly selected 10, two-day old tadpoles from each sibship in §2b (we selected these tadpoles before they had been fed). We euthanized and preserved tadpoles as in §2b. We then measured SVL and the width of the jaw muscle (orbitohyoideus muscle OH), which is diagnostic of ecomorphology [39]. We standardized OH for body size (SVL) by regressing log OH on log SVL [44,53]. We then compared these size-corrected OH and SVL values between allopatry and sympatry using a likelihood ratio test and linear mixed effects models (fitted with maximum-likelihood in the R package ‘lme4’). Specifically, we used a likelihood ratio test (through the ‘anova’ function in R [54]) to compare a null model only containing the random effect ‘sibship’ with a full model that retained this random effect and also included ‘selective environment’ as a fixed effect.

3. Results

(a) Testing whether frequency of trait expression predicts its adaptive refinement

At the start of the experiment, tadpoles from the two selective environments did not differ in body size (likelihood ratio test between null model and selective environment model: χ 2 = 0.81, p = 0.3681 type III sum of squares ANOVA: χ 2 = 1.1548, p = 0.2825). At the end of the experiment, however, tadpole growth showed a significant diet by selective environment interaction (table 1a,b). That is, the magnitude of diet-dependent growth differed across selective environments. A multiple comparisons test revealed that tadpoles from the two selective environments had comparable growth on a detritus diet, but that sympatric tadpoles grew more than allopatric tadpoles on a shrimp diet (table 1c). This difference on a shrimp diet created a significantly greater slope between diets for sympatry than for allopatry (figure 1a). Consistent with this difference in slope, we found that effect size between diets (Cohen's d) was greater for sympatry (Cohen's d = 1.206) than for allopatry (Cohen's d = 0.722). This pattern matches our prediction for sympatric tadpoles: they exhibited greater adaptive refinement (improved growth achieved through superior competitive ability) than allopatric tadpoles on the diet that is frequently consumed in sympatry (shrimp).

Figure 1. Evidence of frequency-dependent adaptation. Tadpoles from sympatric populations (where carnivores are produced frequently): (a) grew more on and (b) won more contests over, the resource for which carnivores are adapted—shrimp—than did tadpoles from allopatric populations (where carnivores are produced infrequently). By contrast, tadpoles from allopatry (b) won more contests over detritus, a resource for which omnivores are adapted. (Online version in colour.)

Table 1. Results from competition experiment, including (a) summary statistics from our model selection procedure (b) results from our ANOVA on the interaction model and (c) distance between group means in growth and their associated p-value following false discovery rate correction (in parentheses). ((a) and (b) indicate that the interaction between diet and selective environment was significant (c) shows that this interaction was driven primarily by shrimp-fed sympatric tadpoles (Sym.shr) growing more than shrimp-fed allopatric tadpoles (Allo.shr), while detritus-fed tadpoles had comparable growth across both selective environments (Allo.det and Sym.det).)

Furthermore, when we categorized each tadpole as ‘winner’ or ‘loser’ (depending on whether or not it grew more than its competitor), sympatric tadpoles were more often the winner on shrimp (35 sympatric winners versus 16 allopatric winners p = 0.0002), whereas allopatric tadpoles were more often the winner on detritus (31 allopatric versus 20 sympatric winners p = 0.0236 figure 1b). This result is consistent with the observation that the slopes of two selective environments intersect near the detritus category (figure 1a). These results also support our prediction (see §2b): sympatric tadpoles were superior competitors on shrimp, and allopatric tadpoles were superior competitors on detritus.

Finally, a shrimp diet yielded greater variation in growth than a detritus diet (σ 2 = 4.64 versus 2.58, respectively p = 0.0067), and there was greater variation in growth for sympatric tadpoles than allopatric tadpoles (σ 2 = 5.14 versus 3.67, respectively p = 0.0397). However, when tadpoles were grouped by diet and selective environment simultaneously, the differences in variation only approached significance (p = 0.0724). Generally, these results, again, suggest that there is a greater effect of competition on shrimp (i.e. greater growth variance), and that sympatric tadpoles had a greater difference in growth between diets than allopatric tadpoles (i.e. greater variance for sympatry).

(b) Evaluating mechanisms of adaptive refinement

Sympatric and allopatric tadpoles did not differ in: (i) intrinsic growth rate on alternative diets (electronic supplementary material, table S2) (ii) time spent resting, swimming, eating or active (electronic supplementary material, table S3) or (iii) trait integration (electronic supplementary material, table S4). These two groups did differ in: (iv) time to eat shrimp and (v) certain trophic traits. Regarding time to eat shrimp, sympatric tadpoles captured and consumed shrimp faster than allopatric tadpoles (χ 2 = 5.11, p = 0.0238 figure 2). Also, as expected, there was significantly lower variance in shrimp capture time for sympatric tadpoles than for allopatric tadpoles (σ 2 = 3340 versus 20824, respectively p < 0.0001).

Figure 2. A mechanism of frequency-dependent adaptation. Tadpoles from sympatric populations (where carnivores are produced frequently) ate shrimp faster than tadpoles from allopatric populations (where carnivores are expressed relatively infrequently). Diamonds, group means. (Online version in colour.)

Regarding trophic traits, as predicted, sympatric tadpoles had significantly more carnivore-like mouthparts than allopatric tadpoles (mean ± s.e.m. mouthparts scores = 2.8 ± 0.2 versus 1.8 ± 0.1 for sympatric and allopatric tadpoles, respectively electronic supplementary material, table S5). For jaw muscle (OH) width, there was a significant diet by selective environment by treatment interaction. Delving into this interaction revealed that, for tadpoles reared in competition, sympatric tadpoles did not differ between diets, but allopatric tadpoles did (electronic supplementary material, table S6). Specifically, sympatric tadpoles showed consistently large OH widths across diets (thereby providing evidence of canalization in this trait see also §3c), but allopatric tadpoles showed plasticity (larger OH on shrimp than on detritus). A multiple comparison test confirmed this pattern: detritus-fed allopatric tadpoles had significantly smaller OH widths than all other groups (electronic supplementary material, table S6B). When we focused on singletons, sympatric tadpoles had significantly larger OH widths than allopatric tadpoles, but there was no diet by selective environment interaction (electronic supplementary material, table S7). In contrast with the patterns for mouthparts and jaw muscles, sympatric tadpoles had significantly more omnivore-like denticle rows than allopatric tadpoles (8.3 ± 0.6 versus 4.5 ± 0.3 for sympatric and allopatric tadpoles, respectively electronic supplementary material, table S5). Gut length did not differ between diets, selective environments or treatments.

(c) Testing for genetic assimilation of trophic morphology

On average, sympatric tadpoles had significantly larger OH widths (0.041 ± 0.012) than allopatric tadpoles (−0.029 ± 0.010 χ 2 = 6.58, p = 0.0103 figure 3). Because this difference was already apparent in tadpoles that had not experienced the dietary cue(s) that normally induce carnivores, and because these sympatric tadpoles represent the derived state (see §2a), this finding suggests that sympatric tadpoles have undergone genetic assimilation in trophic morphology.

Figure 3. Evidence of genetic assimilation of trophic morphology. Even in the absence of a dietary cue that normally induces carnivores, tadpoles from sympatric populations (where carnivores are produced frequently) developed more carnivore-like jaw muscles (OH) than tadpoles from allopatric populations (where carnivores are expressed relatively infrequently). (Online version in colour.)

4. Discussion

A key prediction of plasticity-led evolution, and of evolution by natural selection generally, is that the frequency of a trait's expression will determine the degree to which its functionality is improved by selection [2,12,21,22]. In particular, compared to individuals from populations that express a particular phenotype infrequently, those from populations that express this phenotype more frequently should produce a superior version of the phenotype [12,18,25,32,55–58]. We tested this expectation of frequency-dependent adaptation experimentally by using S. bombifrons tadpoles from natural populations that have diverged in the frequency with which they produce an environmentally induced carnivore ecomorph.

Our results were consistent with frequency-dependent adaptation. Specifically, compared to tadpoles from allopatric populations (which express the carnivore ecomorph relatively infrequently see §2a), those from sympatric populations (which express the carnivore ecomorph frequently): (i) were superior competitors for shrimp, a resource for which carnivores are specialized [59] (figure 1 and table 1) (ii) were more efficient at capturing and consuming shrimp (figure 2) (iii) showed less variation in shrimp-capturing ability (iv) had more exaggerated carnivore features (electronic supplementary material, table S5) and (v) were more carnivore-like prior to experiencing an environmental cue—shrimp ingestion—that normally induces production of the carnivore ecomorph [39,45,60] (figure 3). We also found evidence that the omnivore ecomorph has undergone adaptive refinement in populations where this phenotype is expressed more frequently. Compared to tadpoles from sympatric populations (which seldom express the omnivore ecomorph see §2a), those from allopatric populations (which express the omnivore ecomorph frequently) were superior competitors for detritus, a primary resource of the omnivore ecomorph [59] (figure 1b). These results therefore suggest that neither selective environment produces tadpoles that are intrinsically superior across both diets. Instead, the selective environment that produces a given ecomorph more frequently (carnivores in sympatry, omnivores in allopatry) appears to produce a competitively superior version of that ecomorph. Thus, our data provide empirical support from natural populations for frequency-dependent adaptation.

Regarding possible mechanisms of this frequency-dependent adaptation, we found no evidence that tadpoles from the two selective environments differed in: (i) intrinsic growth rate (electronic supplementary material, table S2) (ii) time spent resting, swimming, eating or active (electronic supplementary material, table S3) or (iii) trait integration (electronic supplementary material, table S4). Sympatric tadpoles did, however, eat shrimp faster (figure 2) and exhibited less variation in shrimp-eating time than allopatric tadpoles. Thus, the competitive advantage of sympatric tadpoles in using shrimp (table 1 and figure 1) could be explained, in part, by sympatric tadpoles being better at capturing and consuming shrimp.

Differences between selective environments in tadpole trophic morphology probably also contributed to the sympatric tadpoles' competitive advantage on shrimp. Sympatric tadpoles had larger jaw (OH) muscles and mouthparts than allopatric tadpoles (electronic supplementary material, table S5). Both traits aid in the capture of large, mobile prey, such as fairy shrimp and tadpoles [50,61,62]. Indeed, previous work found a similar pattern. One such study [63] compared the morphology of experimentally reared tadpoles from sympatry versus allopatry and found that the former were more likely to express the carnivore morphology the former also had significantly different jaw muscle (OH) allometry (the slope of the relationship between OH width and body length was steeper for tadpoles derived from sympatry than for those from allopatry). Another study [25] found that wild-caught tadpoles from sympatric populations were more carnivore-like in their morphology than wild-caught tadpoles from allopatric populations. Together with the present study, these studies suggest that populations which express the carnivore morph more frequently produce more exaggerated carnivore features and that those exaggerated features improve fitness. Thus, more frequent trait expression predicts greater magnitude of adaptive refinement.

As noted above, we found that sympatric tadpoles produced larger (more carnivore-like) jaw muscles than allopatric tadpoles, even prior to cue exposure (figure 3). This result implies that sympatric tadpoles: (i) may be primed to eat shrimp from early development (larger jaw muscles are needed to eat shrimp) and (ii) do not need an environmental cue to develop the carnivore morphology. This result further suggests adaptive refinement of the carnivore ecomorph in sympatry relative to allopatry. At a mechanistic level, because tadpoles from sympatry start out more carnivore-like, they may have greater difficulty overcoming a potential trade-off in the ability to switch between morphs [64], and thus have greater difficulty developing as omnivores. Regardless of the exact mechanistic cause, this early phenotypic bias may be adaptive, given that selection favours carnivore production in sympatric populations of S. bombifrons [53].

Our finding that sympatric tadpoles do not need an environmental cue to produce carnivore-like jaw muscles suggests that the ancestors of these tadpoles might have undergone genetic assimilation of trophic morphology. Although genetic assimilation has previously been demonstrated in laboratory experiments [16,65,66], and theory supports its role in enabling populations to adapt to rapidly changing environments [6,10,67–72], its relevance to natural populations has been questioned (e.g. [3,4]). Interestingly, we also found evidence of genetic assimilation in jaw musculature from our competing tadpoles (electronic supplementary material, table S6). In this case, sympatric tadpoles had larger (more carnivore-like) jaw muscles in both diet treatments and exhibited the flat reaction norm characteristic of genetic assimilation [12,73]. Similarly, Levis et al. [74] found evidence of genetic assimilation in patterns of gene expression. Whereas gene expression profiles of allopatric tadpoles differed between detritus and shrimp diets, those of sympatric tadpoles did not. Furthermore, a transcription factor (btf3) exhibited loss of diet-dependent expression plasticity, and a peptidase gene (pm20d2) showed an overall decrease in expression in sympatry relative to allopatry, suggesting possible improved efficiency [74]. These studies, combined with those of other natural systems (e.g. [36,37,55,75–79]), point to the generalizability—and possible importance—of genetic assimilation.

Additional studies are needed, however, to identify the mechanisms underlying any such genetic assimilation [72]. For instance, the gene expression differences mentioned above are consistent with genetic assimilation, but they could also be caused by persistent epigenetic changes (e.g. see [80]). To distinguish between genetic assimilation and ‘epigenetic assimilation’ as mechanisms underlying constitutive expression of a phenotype will require investigating whether constitutively expressed phenotypes are associated with DNA sequence changes versus epigenetic ‘tags’ (e.g. methylation).

Returning to frequency-dependent adaptation, why should the frequency of trait expression drive the magnitude of its adaptive refinement? Frequency-dependent adaption is expected to occur for at least two, non-mutually exclusive, evolutionary reasons. First, differences in the size of subpopulations that express the phenotype should lead to differences in both: (i) the strength of selection relative to that of genetic drift and (ii) the number of variants exposed to selection per generation [2,81]. A rough analysis suggests that the subpopulation of S. bombifrons carnivores in sympatry is at least twice as large as that in allopatry (electronic supplementary material, appendix S1). All else being equal, this difference across selective environments in numbers of carnivores suggests that selection should be at least twice as effective at acting on sympatric carnivore subpopulations than on allopatric carnivore subpopulations. Although the exact selection coefficients and effective populations sizes are unknown, selection favouring extreme carnivores in sympatry is probably stronger than selection favouring carnivores in allopatry: sympatric populations are under strong directional selection, whereas allopatric populations are under weak disruptive (i.e. quintic rather than quadratic) selection [53]. Thus, the recurrent exposure of a relatively larger population size of carnivores in sympatry may have played a causal role in the adaptive evolution of this phenotype. Both factors—recurrence of phenotype expression and large population producing the phenotype—are probably needed for rapid adaptation, and the relative importance of each factor warrants further study.

A second reason frequency-dependent adaptation should occur is that as a trait's frequency of expression increases, so should the bias in the direction of selection on non-specific modifiers of that trait [2,82]). That is, selection on loci that show antagonistic pleiotropy among alternative phenotypes should favour those alleles that are best suited to the most frequently expressed phenotype [2]. This bias in modifier accumulation can alter the fitness consequences associated with different phenotypes in a population, and it can even cause such fitness effects to diverge among populations that diverge in the frequency at which these phenotypes are expressed. Indeed, a re-assessment of data from previous studies of this system [43,44] suggests that the covariance between carnivore morphology and body size (a proxy for fitness [48,52,53]) is nearly twice as large for sympatric tadpoles than for allopatric tadpoles (0.04013 versus 0.02103). Using the Price equation [83], this suggests that sympatric populations might accumulate modifications to the carnivore phenotype nearly twice as fast as allopatric populations. Interestingly, for a difference in carnivore morphology of the magnitude observed between allopatric and sympatric populations to arise, it would take approximately 70 spadefoot generations (electronic supplementary material, appendix S2), which corresponds to the estimated time (approx. 150 years) that these two selective environments have been separated and diverged in ecomorph production [46,47]. Thus, differences in modifier accumulation, as a result of biases in phenotype production, may drive patterns of genetic and phenotypic divergence within (and potentially between) species.

In conclusion, our findings provide evidence that the frequency of trait expression drives the magnitude of adaptive refinement. Thus, our results thereby support a key prediction of both plasticity-led evolution and adaptive evolution.


The University of North Carolina's IACUC approved all the procedures. Field collections were conducted under Scientific/Education Collecting Permits AZ SP594327 and SP615759, CO 15HP995, NE 574, OK 6650 and TX SPR-0316-094.

14.3: Frequency-Dependent Selection - Biology

Natural selection can also be influenced by the frequency of different phenotypes within the population, a process known as frequency-dependent selection.

In positive frequency-dependent selection, as a phenotype becomes more common, the fitness of that phenotype also increases.

For example, there are many toxic-colored morphs. Heliconius butterfly species. When one more is common, birds have already learned that it is poisonous and will avoid it. However, when birds encounter a rare morph and learn that it is also poisonous, the fitness and frequency of that morph will also increase.

On the other hand, in negative frequency-dependent selection, the fitness of a phenotype decreases as it becomes more frequent.

For example, to avoid predation, viceroy butterflies mimic the coloration and pattern of toxic monarch butterflies. When the mimics are rare, birds will avoid them, since they have likely encountered more of the poisonous monarchs. However, when mimics are common, encounters between the birds and butterflies are more likely positive, decreasing the fitness of viceroys.

31.3: Frequency-dependent Selection

When the fitness of a trait is influenced by how common it is (i.e., its frequency) relative to different traits within a population, this is referred to as frequency-dependent selection. Frequency-dependent selection may occur between species or within a single species. This type of selection can either be positive&mdashwith more common phenotypes having higher fitness&mdashor negative, with rarer phenotypes conferring increased fitness.

Positive Frequency-Dependent Selection

In positive frequency-dependent selection, common phenotypes have a fitness advantage. This scenario is often seen in interactions where mimicry is involved. In the Neotropical region of Central America, the butterfly species Heliconius cydno and Heliconius sapho are involved in a Müllerian mimicry partnership. Both butterflies are black and white, a common aposematic signal in the animal kingdom that warns of toxicity, venom, bad taste, or other predator deterrents.

Interestingly, H. cydno can hybridize with a closely related sister species, H. melpomene, and produce offspring. H. melpomene is predominantly black and red. The resulting mixed white-red-black hybrid offspring are significantly less fit. In addition to the female hybrids being sterile, predators do not recognize the colors as deterrent warnings, and butterflies of either parent species do not recognize the hybrids as potential mates. Therefore, the most common phenotype&mdashblack and white&mdashis selected for. However, the more frequent the white-red-black hybrids become, the more relatively fit the phenotype becomes because predators are more likely to have learned about the warning pattern through a previous encounter with another hybrid individual.

Negative Frequency-Dependent Selection

Negative frequency-dependent selection is a form of selection in which common phenotypes are selected against. One type of negative frequency-dependent selection occurs when rare phenotypes of a prey species confer higher fitness because predators do not recognize the organisms as prey. This is known as apostatic selection.

A classic example of apostatic selection is found in the grove snail and one of its predators, the thrush. The grove snail displays polymorphic shell patterning, but the predatory thrushes tend to focus on one or two common forms of shell patterning when searching for prey. These common phenotypes, therefore, experience stronger negative selection pressure.

Another example of negative frequency-dependent selection is found in plant self-incompatibility systems. In angiosperms, homomorphic self-incompatibility is crucial to prevent self-fertilization that typically involves genetic mechanisms that prevent pollen germination or tube growth if the pollen and pistil express identical alleles. This is controlled by a multiallelic genomic region called the S-locus. Because of this, plants expressing common forms of the S-locus will often encounter false &ldquoselves&rdquo&mdashwhere a potential reproductive event, and therefore gene flow, is blocked due to the self-incompatibility genes. This means that rarer forms of the S-locus are under positive selection, while common forms are selected against.

Naisbit, R. E., C. D. Jiggins, and J. Mallet. &ldquoDisruptive Sexual Selection against Hybrids Contributes to Speciation between Heliconius Cydno and Heliconius Melpomene.&rdquo Proceedings of the Royal Society B: Biological Sciences 268, no. 1478 (September 7, 2001): 1849&ndash54. [Source]

Brisson, Dustin. &ldquoNegative Frequency-Dependent Selection Is Frequently Confounding.&rdquo Frontiers in Ecology and Evolution 6 (2018). [Source]

Perlin Lab Research

The research in my lab focuses on the evolution of interactions between pathogens and the hosts on which they cause disease. At present, this work has two main areas of emphasis: fungal/plant interactions and population dynamics of bacteria resistant to antibiotics.

Fungal/Plant Interactions leading to Pathogenesis

Our work with fungal/plant interactions has examined primarily the “smut” fungi.

Plant Diseases Caused by Smut Fungi:

These are basidiomycetous fungi that have several different morphological stages in the course of their development. The final stage is obligately parasitic, in that production of the diploid spores requires successful infection of a suitable host plant. The general lifecycle is illustrated below.

Microbotryum violaceum

Ustilago maydis

Representative life cycle of a smut fungus

Questions of Host Specificity and Evolution of Sexual Systems

We began our work with smut fungi over 20 years ago with the anther smut, Microbotryum violaceum, a fungus that infects members of the carnation family (Caryophyllaceae) and replaces the pollen of anthers with its own teliospores. Our early work attempted to establish such molecular methods as genetic transformation and provided DNA “fingerprints” for individual isolates from different host species. Since then it has become clear that what has been called Microbotryum violaceum (or Ustilagoviolacea, Pers. Roussel ) for hundreds of years is actually a species complex, comprised of many different species or incipient species, each of which is adapted to infection of its own host species. What has emerged is an exciting system for examination of the ecology and evolution of host/pathogen interactions in “wild” non-agricultural environments, a model for studying the evolution of sexual systems and sex chromosomes, and gene expression during infection that may give clues to important genes involved in pathogenicity.

Fungal Dimorphism and the Link to Pathogenicity

A major aspect of our work with fungi involves genes that control morphogenesis and pathogenicity in fungi. One link that ties these two processes together is the dimorphic switch between yeast-like and filamentous forms. In model organisms such as the ascomycete yeast, Saccharomyces cerevisiae, and the basidiomycete smut, Ustilago maydis, membrane proteins that sense and transport ammonium have been implicated in this switch. The MEP proteins from yeast have been intensively studied, while those in other fungi, particularly pathogenic organisms, have been much less scrutinized.

The yeast S. cerevisiae can adopt several alternative developmental fates depending on the availability of specific nutrient sources. When nitrogen and fermentable carbon sources are both plentiful, haploid and diploid yeast cells reproduce by budding. When nitrogen is limiting but an abundant fermentable carbon source is present, diploid yeast cells undergo pseudohyphal differentiation (i.e., a dimorphic switch) to form filamentous colonies that forage for nutrients. These alternative fates allow this nonmotile organism to appropriately adapt to its surroundings. The signaling cascades by which yeast cells sense and respond to nutrients serve as models to understand how all cells sense and respond to the environment. Ammonium is a preferred nitrogen source for many fungal and bacterial species and for plants, as well. Moreover, the ability of microbes to sense and transport nitrogen is critical, not only for survival, but also as a prelude to a variety of developmental processes. These responses to nutrient availability are governed by both a mitogen-activated protein kinase (MAPK) cascade and the cAMP-dependent protein kinase pathway, with cross-talk between these parallel pathways at several important points these cascades also control several aspects of pathogenic development discussed below. How nutrient signals in the extracellular environment control the activation of these globally conserved signaling cascades is a central question.

1. Nutrient sensing controls pathogenicity

Many fungal pathogens utilize a similar switch between a yeast-like form and a filamentous form as an integral part of their overall strategy of disease production. For example, the dimorphic transition from budding yeast to filamentous forms is crucial for infection for Candida albicans, and it is important for Cryptococcus neoformans. Similarly, in some plant pathogens, such as the maize pathogen Ustilago maydis, the dimorphic switch plays a critical role in both morphogenesis and pathogenicity. In fact, there is a high degree of conservation in the signaling cascades controlling development and virulence of these divergent ascomycete (S. cerevisiae, C. albicans) and basidiomycete (C. neoformans, U. maydis) yeasts. Based on these conserved pathways, studies in the model yeast S. cerevisiae have provided insight into fungal pathogenesis in humans and plants. In turn, studies in the pathogens will not only validate studies from yeast but also are expected to begin to identify universal and species specific components of ammonium sensing networks.

Several factors can cause cell differentiation in Dimorphic Fungi

Nutrient sensing in fungal plant pathogens

As mentioned above, many of the conserved features of signaling are also found in fungal plant pathogens. The smut fungus U. maydis exists as haploid yeast-like sporidia, and mating of cells of opposite mating-type, accompanied by signals from host plants, leads to formation of infectious dikaryotic hyphae. U. maydis infection causes tumor formation on all maize shoot tissues, including tumor formation, with particular severity on the ears. Such damage can result in annual losses exceeding $200 million. For U. maydis mating occurs readily on rich media, but haploid or diploid sporidia grow filamentously on nitrogen starvation medium. The first genes encoding proteins involved in ammonium transport were cloned from yeast and Arabidopsis. As mentioned above, for S.cerevisiae, low nitrogen in the presence of an abundant fermentable carbon source leads diploid yeast cells to grow as pseudohyphae. The genes primarily responsible for both transport and sensing of available nitrogen include MEP1, MEP2, and MEP3, each of which encodes an integral membrane protein that transports ammonium ions or the toxic analog methylammonium. Of the three,Mep2 is a high-affinity permease, that acts as both an ammonium transporter and sensor. The yeast Mep2 protein controls the transition to pseudohyphal growth in response to ammonium limitation. mep2/mep2 mutant strains have no vegetative growth defect on nitrogen limiting medium, but in contrast to wild type cells are unable to undergo pseudohyphal differentiation. This defect can be suppressed by exogenous cAMP or by dominant activation of the PKA signaling pathway, supporting models in which Mep2 either functions upstream of PKA signaling, or the two pathways function in parallel. In U. maydis, we have identified two homologs of Mep2, called Ump1 and Ump2, and importantly Ump2 is required for filamentous growth on low nitrogen medium and can restore both growth and filamentous growth when heterologously expressed in S. cerevisiaemep1,2,3 mutant strains. Ump2 was identified as being more highly expressed under conditions of low ammonium. Thus, the S. cerevisiae Mep and U. maydis Ump proteins are physical and functional homologs of each other and function in both ammonium transport and ammonium sensing. The yeast Mep and fungal Ump proteins are part of a larger family of proteins that are conserved between bacteria, yeast, fungi, and humans.

2. Regulation of transporter function by phosphorylation

Several examples exist where proteins involved in transport are regulated by phosphorylation/dephosphorylation cycles. Our preliminary work examining the phosphorylation of the yeast Mep proteins indicates that a putative PKA site is present. Furthermore, the target residue is necessary for appropriate Mep function to control filamentation in both S. cerevisiae and U. maydis. Our ongoing studies aim to define the function of this phosphorylation, and will provide insight into conserved regulatory mechanisms that operate in these and other sensory receptors. We are also using split-ubiquitin and protein association assays to identify interactions between the Ump proteins and possible signaling components of U. maydis.

3. Additional Roles for Signaling Pathway Components

Proteins of the 14-3-3 and Rho-GTPase families are functionally conserved eukaryotic proteins that participate in many important cellular processes such as signal transduction, cell cycle regulation, malignant transformation, stress response, and apoptosis. However, the exact role(s) of these proteins in these processes is not entirely understood. Using the fungal maize pathogen, Ustilago maydis, we were able to demonstrate a functional connection between Pdc1 and Rho1, the U. maydis homologues of 14-3-3epsilon and Rho1, respectively. Our experiments suggest that Pdc1 regulates viability, cytokinesis, chromosome condensation, and vacuole formation. Similarly, U. maydis Rho1 is also involved in these three essential processes and exerts an additional function during mating and filamentation. Intriguingly, yeast two-hybrid and epistasis experiments suggest that both Pdc1 and Rho1 could be constituents of the same regulatory cascade(s) controlling cell growth and filamentation in U. maydis. Overexpression of rho1 ameliorated the defects of cells depleted for Pdc1. Furthermore, we found that another small G protein, Rac1, was a suppressor of lethality for both Pdc1 and Rho1. In addition, deletion of cla4, encoding a Rac1 effector kinase, could also rescue cells with Pdc1 depleted. Inferring from these data, we propose a model for Rho1 and Pdc1 functions in U. maydis, as follows. Rac1 sequesters Cla4 to the growing tip. Thus, Rho1 could interfere with Rac1 activities by preventing Rac1 from localizing at the growing pole. Other possibilities for how Rho1 could indirectly act as negative regulator of Rac1 are by sequestering Cdc24 (the RhoGEF) or Cla4 away from Rac1. Yeast two-hybrid analyses suggest possible interaction between Cdc24 and Rho1. As with Rho1, Cdc24-null cells are nonviable. In vivo, Rho1 and Rac1 could compete for Cdc24. On the other hand, Cla4 has been shown to localize at the bud-neck in S. cerevisiae. According to our GFP-Rho1 data, Rho1 also localizes (possibly with the help of Pdc1) at the bud-neck or in the septation area. Thus, it is a possibility that Rho1 is somehow involved with sequestration of Cla4 to the septation area and away from the growing tip and Rac1. In such cases, cells would then undergo budding instead of filamentation.

4. Control of Mitochondrial Inheritance

The mitochondrion is a critical organelle, whose importance far exceeds its common function as the energy-producing “powerhouse” of the cell, as it has been found to be involved in fundamental processes such as apoptosis, aging and metabolic homeostasis (Seo et al., 2010). Thus, appropriate inheritance of mitochondria is essential for growth and development of progeny. More importantly, such controls and the negative consequences of their disruption, have important ramifications for human health. Sexually reproducing eukaryotes have evolved similar mechanisms that allow the mitochondria of a single parent to be passed on to the offspring (uniparental inheritance or UPI). This pattern of UPI is nearly universal across eukaryotes, from isogamous protists with equal-sized gametes, to anisogamy, as seen in animals and plants with extreme gamete-size asymmetry. UPI may facilitate purifying selection against deleterious mutations, restrict intergenomic conflicts and limit mitochondrial heteroplasmy, the case where both parental types occur and are maintained in cells. On the other hand, cytoplasmic mixing reduces variation, impeding the efficacy of selection against defective organelles or selfish genetic elements (Razvilavicius et al. 2017).

The mating type contributing the majority of mitochondria to the next generation is “maternal”, while the other is “paternal”. Control of mitochondrial inheritance can similarly be maternal or paternal. In such cases, nuclear genes in one mating type control destruction of the mitochondria in the gametes of one particular mating type. Maternal control involves destruction of the partner’s mitochondria after fertilization, whereas in paternal control, nuclear genes in one mating type control destruction of its own mitochondria during gamete formation. Such controls may be found in isogamous organisms, but for multicellular organisms where there is gamete asymmetry, maternal control would amount to the targeting and elimination of mitochondria from sperm post-fertilization. In contrast, under paternal control, the exclusion or disabling of mitochondria occurs during spermatogenesis (before entering the oocyte). It is expected that paternally controlled organelle destruction is precluded because of the lack of long-term linkage between the paternal nuclear genotype and its own mitochondria, as the cytoplasm is exclusively maternally inherited (Razvilavicius et al. 2017). Such theoretical predictions typically explore isogamous systems, but should apply to multicellular organisms, since a paternal nuclear gene that causes the exclusion of sperm mitochondria cannot build up linkage with maternally inherited organelles, as this relationship is re-set every generation (Razvilavicius et al., 2017).

The pathogenic smut fungus Ustilago maydis (and its close relative Sporisorium reilianum) serves as an excellent model to study inheritance of mitochondria in higher eukaryotes (Fedler et al., 2009). The life cycle of this pathogen has been extensively characterized and its common host is Zea mays (maize or corn). The growth form of wild type U. maydis cells and pathogenicity on maize are inextricably linked because the filamentous, dikaryotic stage is the natural pathogenic cell type. The haploid strains of the fungus are saprophytic, budding cells that are easily cultured and manipulated in the laboratory. When two compatible U. maydis haploid cells are in close proximity, they produce fusion hyphae that grow towards each other and eventually fuse to form the infectious dikaryon this cell type is an obligate biotroph that requires maize tissue for proliferation and development. Compatible haploid cells fuse and form a stable dikaryon only if they carry genes with different specificities at both the a and b mating-type loci. Cell fusion is controlled by the a mating-type locus, which has two alternative non-allelic forms (termed, a1 and a2 idiomorphs). The a locus encodes pheromones and pheromone receptors similar to those of yeast. Notably, U. maydis contains genes for the appropriate segregation of parental mitochondria during mating. The lga2/rga2 system controls a degradation-mediated UPI mechanism, in which all offspring will contain mitochondrial material from a single parent (a2). This mechanism is consistent with the several mechanisms of maternal mitochondrial inheritance seen in higher eukaryotes. Previous studies have elucidated the relationship between the lga2/rga2 system and the mating program of U. maydis. Pheromone stimulation upon the encounter between compatible partners confers low-level expression of the lga2/rga2 system in the a2 partner (Urban 1996). The protein product of rga2 has been linked to protection of a2 mtDNA, while the protein product of lga2 degrades unprotected mtDNA by a currently unknown mechanism that ultimately leads to selective mitophagy (Fedler et al., 2009). Heterozygosity at the multiallelic b locus is required for the production and maintenance of a stable filamentous dikaryon, and for pathogenicity (Brachmann et al., 2004). The b mating-type locus encodes two homeodomain containing proteins, bE and bW, that interact when produced from different alleles. The b heterodimer controls events after cell fusion required for the production of a stable filamentous dikaryon and regulates the transcription of a set of target genes that directly or indirectly control morphogenetic transitions and pathogenicity (Brachmann et al., 2001, 2004). After conjugation and subsequent cellular fusion of partners, the organism enters the pathogenic stage of its life cycle. The heterodimeric transcription factor bE/bW is upregulated during pathogenic development and is responsible for the upregulation of genes involved in host infection (Heimel et al., 2010). Interestingly, the bE/bW complex upregulates lga2 expression, leading to the degradation of unprotected a1 mtDNA (Fedler et al., 2009).

Sporisorium reilianum, another smut fungus, is a close relative of U. maydis, as evidenced by phylogenies based on ribosomal DNA sequencing as well as comparative genomics. They share a similar life cycle nonetheless, mating in S. reilianum can occur between three parental strain types, a1, a2 and a3 . Interestingly, the a2 form in S. reilianum also houses orthologues of the lga2/rga2 mitochondrial inheritance system. As a result, offspring resulting from a cross between the a2 parent and either a1 or a3 are expected to be of the a2-mitotype. However, what occurs in a cross between the a1 and a3 strains? We propose that a cross between a1 and a3 partners will produce heteroplasmic offspring (a1/a3-mitotype). We predict that, given the importance of mitochondria in an organism’s development and the genetic conflict that heteroplasmy imposes, offspring of the a1/a3-mitotype will be less fit for survival. We are currently exploring these questions in collaboration with Dr. Michael Menze in the Biology Department here and with Dr. Jan Schirawski at the Matthias-Schleiden-Institut, in Jena, Germany.

Evolution of Antibiotic Resistance in Bacterial Populations

Microbial diversity is important in a variety of contexts, in particular having implications for infectious diseases, bioremediation, and environmental engineering. Examination of the mechanisms underlying such diversity and its evolution are important for providing a roadmap that can lead to better understanding of the above-mentioned and other, related, areas. Several examples are available where mutualism in microbial communities can lead to the maintenance of microbial diversity. Fewer examples have been presented where some individuals provide protection to others in a population without a concomitant benefit being returned to the protector. Protection of sensitive members of a population against antimicrobials by resistant genotypes has been observed in biofilms. In these cases, spatial proximity to the protector/producer was a prerequisite, or at least an important component, for survival by otherwise sensitive individuals. We have extended these studies to shaking liquid or planktonic populations in order to examine the dynamics of such unrewarded protection or altruism. We have developed a family of mathematical models that have examined antibiotic resistance in such systems (Dugatkin et al., 2003, 2005) and have generated preliminary data which support some aspects of these models (Dugatkin et al., 2004). Our results so far show that such altruists can provide frequency-dependent antibiotic resistance to other members in the population. Furthermore, frequency-dependent selection of the traits of these altruists promotes microbial diversity. We are currently conducting more detailed experiments designed to test our model's predictions, and will then use the results obtained to further refine the models. Specifically, we will use competitions between near-isogenic Escherichia coli strains that are either sensitive to ampicillin or are resistant, by virtue of a plasmid-encoded b-lactamase. Among those strains that are resistant, we compare the outcome for those strains that only protect themselves from ampicillin versus those strains that can also provide protection to others in their vicinity, i.e., by destroying ampicillin nearby in the medium. These analyses have also been extended to competition experiments with greater relevance for natural clinical or human settings. Specifically, we have examined whether ampicillin-resistant E. coli can protect sensitive Salmonella in their vicinity. Our results (Perlin et al., 2009) demonstrate clearly that non-pathogenic antibiotic-resistant bacteria may protect otherwise susceptible pathogens by a mechanism that does not involve gene transfer. Evolution of such transient survival mechanisms may thereby hinder therapeutic use of antibiotics, and should be considered in devising effective treatment strategies.

The Brandenberg Gate in Berlin. The program provides scientific and cultural opportunities.

National Science Foundation International Research Experiences for Students (IRES)

Track I IRES Sites: Training with Smut Fungi in Germany

Through a three-year grant from the US National Science Foundation, our lab has developed an exchange program whereby graduate students and undergraduates from my lab visit the labs of research collaborators in Germany from 8-12 weeks each summer. Students accepted for these exchanges are trained and prepared during the normal academic year and then undergraduates are paired with a US graduate student to work on independent projects of mutual interest, to the students and to the labs involved in the collaborations. We endeavor to recruit members of groups traditionally under-represented in the sciences, both to expand opportunities for such student populations and to also provide all students chosen to expand their cultural horizons by being able to experience different ways to approach science and to view the world.

New landscapes and experiences await.

Year 1 (May-September, 2019)

In Fall 2018, we recruited 3 Doctoral (1 Hispanic/Latino and 2 White non-Hispanic students) and 3 undergraduate students (1 African-American female, 1 White non-Hispanic female, and 1 White non-Hispanic male) as candidates for this project. This group was trained in basic molecular biological techniques and in some aspects of species-specific techniques prior to travel. In general, undergrads were paired with their graduate student mentors, both for training purposes and so that they would become familiar with each other on a working basis prior to their visit to Germany. In addition, all students received formal or informal training in elementary German language.

Joining the crowd in Koln Cathedral.

On May 15, 2019, students began their research visits to their respective labs: 2 graduate student/undergrad pairs went to Jena, Germany to the lab of Prof. Dr. Jan Schirawski, Matthias-Schleiden-Institut. These students are working on questions relating to Sporisorium relianum specifically, they are investigating fungal proteins that manipulate host plants or are investigating mitochondrial inheritance and mitochondrial function in this organism. One graduate student/undergraduate pair joined the lab of Dr. Dominik Begerow at the RWTH University of Bochum. There they are investigating ways of genetically manipulating M. violaceum strains, conducting large-scale greenhouse experiments to investigate relative pathogenicity of strains, and learning microscopy techniques to follow the infection process.

In addition to research trips to dark places, participating students can engage in research brainstorming sessions and, possibly, other traditional forms of inspiration.

Major Research Goals:

Bochum Group:

Graduate Student (William Beckerson): Develop and compare methods for implementing CRISPR Cas to disrupt genes in the fungus, Microbotryum lychnidis-dioicae. This involves alternative methods for introduction of the Cas9 enzyme into cells and improved methods of fungal transformation.

Phillip Sullivan in the greenhouse.

Undergraduate student (Phillip Sullivan): Analyze the relative virulence of wild type infection by Microbotryum species on their natural hosts compared with those of M. lychnidis-dioicae modified to express a putative effector from the sister species, M. silenes-dioicae.

Graduate Student (J. Ham): Express an effector from M. lychnidis-dioicae in Sporisorium relianum. Test such recombinant strains on the maize host to see if they elicit symptoms already observed in the Microbotryum hosts, e.g., early flowering.

Undergraduate Student (Kiara Smith): Express an effector from S. reilianum in M. lychnidis-dioicae and examine whether these strains lead to suppression of apical dominance, as observed in the natural interaction.

Graduate Student (Hector Mendoza): Conduct infection studies of S. reilianum on maize to compare relative virulence of different combinations of mated strains, in order to interrogate the roles of mating type on this process. Also, examine mitochondrial inheritance after such infections.

Undergraduate Student (Caroline Culver): Generate a pair of S. reilianum strains with their alternative oxidase (AOX) genes disrupted. Test these for susceptibility to stresses and relative virulence.

The Major Training Goals of this project were to:

1) Recruit both Doctoral student and Undergraduate student participants, preferably from URM groups.

2) Prepare them ahead of their travel for an optimal experience in the German labs participating in this project.

3) Give them training specifically in work with smut fungi (i.e., Sporisorium reilianum, Ustilago maydis, Microbotryum violaceum spp.)

4) Provide them with specialized training in microscopy.

5) Provide them with intellectual and cultural experiences in the host country.

Questions about this program and possibilities for joining this group may be directed to Dr. Perlin at [email protected]

Phillip Sullivan at the bench

Local cultural experience: Community mine restoration project

A display of mustards

Current Students in Dr. Perlin’s lab

William Beckerson at the bench.

William Beckerson , recent PhD. Examined fungal effectors across the Microbotryum violaceum species complex developed CRISPR/Cas9 technology to disrupt genes for putative effector genes so as to functionally test their roles in manipulation of host plants. Moving on to a Post-doctoral position exploring "Zombie ant" fungi at the University of Central Florida, with Dr. Charissa de Bekker.

Sunita Khanal , recent PhD. Characterized genes involved in nitrogen assimilation in U. maydis . Moving on to a Post-doctoral position at the University of Maryland Cardiology Division, developing CRISPR/Cas9 to study myosin phosphatase in hypertension and other vascular disorders with Dr. Steven Fisher.

Hector Mendoza , graduate student. Examining mechanisms for control of mitochondrial inheritance and fitness costs associated with its evolution in S. reilianum and U. maydis .

Ming-Chang "Nelson" Tsai , graduate student. Studying roles of gene expression for various target genes in M. violaceum .

Joseph P. Ham , graduate student. Studying the roles of fungal effectors and whether they can alter host preference or host manipulation across species boundaries.

Roxanne Leiter , graduate student. Investigating the roles of fungal effectors in establishing whether a fungus will be a “specialist” and only be able to infect one or a few host species or a “generalist” able to infect a broad range of host species.

Shikhi Baruri, graduate student.

Otniel “Alex” Nava-Mercado , undergraduate. Working on nitrogen assimilation in U. maydis .

Caroline Culver , undergraduate. Investigated role of alternative oxidase function in U. maydis and S. reilianum . Currently enrolled in University of Louisville medical school.

Grace Long , undergraduate. Examining possible host interactors with fungal effectors.

Phillip Sullivan , undergraduate. Exploring the role of host-specific effectors on host range.

Luke Schroeder , undergraduate. Working on nitrogen assimilation in U. maydis .


Lalu Villa Krishna Pillai , PhD. Identified a role for a PTEN orthologue in U. maydis in teliospore development and germination, as well as in pathogenicity. Currently Research Regulatory Specialist Norton Healthcare.

Su San Toh , PhD (deceased). Examined gene expression in plants infected with M. violaceum . Participated in the first genome and transcriptome work for these fungal species. Helped develop the first reproducible genetic transformation system for these fungi. Author, co-author, or collaborator on over 11 publications about this system.

Michael Cooper , PhD. Exploring nitrogen metabolism in U. maydis . Currently working at Murray State University.

R. Margaret Wallen , PhD. Examined the interaction between ammonium transport and mating in U. maydis . Currently, conducting research on endosomes in communication between different cell types also, teaching biology at the college level.

Swathi Kuppireddy , PhD. Characterized fungal effectors for M. violaceum and examined the role of genes in controlling their host plants.

Hannah Mianzo , currently, Genetic Counselor.

Ben Lovely , PhD. Examined signaling via Hsl proteins in U. maydis and their connections to other signaling pathways. Currently, faculty, Department of Biochemistry and Molecular Biology, University of Louisville Medical School.

Charu Agarwal, PhD. Co-mentored with Dr. David Schultz, Department of Biology. Charu analyzed cAMP turnover in U. maydis via a genetic and biochemical analysis of cyclic phosphodiesterases of the fungus.

Jinny Paul, PhD. Characterization of ammonium transporter homologues of several fungi, with emphasis on those from U. maydis and M. violaceum.

Cau Pham, PhD. Exploring the roles of 14-3-3 and Rho proteins in cytokinesis, cell polarity, morphology, and pathogenesis for U. maydis. Currently a Research Associate at the Centers for Disease Control.

Zhanyang Yu, PhD. Isolation and characterization of Rho1 homologue in U. maydis and identification of interacting proteins and pathways.

Kiara Smith , undergraduate, Honors student. Investigated role of nitrite reductase in U. maydis . Currently enrolled in University of Louisville medical school.

Jack Desmarais , undergraduate, Honors student. Investigated role of PTEN orthologue in U. maydis . Currently enrolled in University of Louisville medical school.

Madison Furnish , undergraduate. Investigated roles of mating-type on gene-specific function in U. maydis . Currently in PhD program on Pharmacology at the University of Colorado, Denver.

Jared Andrews , undergraduate, Honors student. Characterizing the roles of sugar transporters in cell morphology and mating in M. violaceum . Currently in PhD program at Washington University, St. Louis.

Dominique Razeeq , undergraduate. Characterizing the roles of secretory lipases in cell morphology and mating in M. violaceum . Currently in medical residency, after graduating University of Louisville medical school in 2019.

Trisha H. Patel, undergraduate. Member of the team identifying interacting proteins with signaling components from U. maydis that may affect nitrogen metabolism.

Ann-Claude Rakotoniaina, undergraduate. Member of the team examining gene expression of M. violaceum during plant infection.

Tia Alton, undergraduate, Honors Student. Member of the team working on competition experiments with bacteria. Currently medical physician.

Evan Raff, undergraduate, Honors Student. Member of the team identifying interacting proteins with signaling components from U. maydis. Currently medical physician.

Alexander Bajorek , undergraduate, Honors Student. Member of the team working on competition experiments with bacteria. Currently medical physician.

Courtney McKenzie , undergraduate, Honors Student. Member of the team working on competition experiments with bacteria. Currently medical physician.

Cayse Powell , undergraduate, Honors Student. Member of the team working on competition experiments with bacteria. Currently medical physician.

Himati Patel, undergraduate, Honors Student. Member of the team working on competition experiments with bacteria. Currently pharmacist.

Lacey Hazel, undergraduate. Work on developing a gene disruption system for M. violaceum. Currently applying to graduate programs in neurobiology.

Greg Shaw, undergraduate, Honors Student. Member of the team identifying interacting proteins with signaling components from U. maydis. Currently medical physician.

Trisha Patel, undergraduate, Honors Student. Member of the team identifying interacting proteins with signaling components from U. maydis. Currently medical physician.

Anna Hellman, undergraduate, Honors Student. Member of the team identifying interacting proteins with signaling components from U. maydis. Currently in Pediatric Neurology residency, University of Louisville.


We would like to thank two anonymous reviewers for helpful comments on earlier versions of this manuscript. We would like to thank for help with fieldwork I༚ki Mezquita, Tomás Latasa, Mario Garc໚-París, Bernat Garrigós, Pere Luque, Xoaquín Baixeras, Francisco Cano, Jean Pierre Boudot, Jürgen Ott, Cedrick Vanappelghem, Phillip Lambert and Phill Watts. This research was funded by the Spanish Ministry of Science and Innovation, grant CGL2008-02799 and CGL2008-03197-E. RSG is supported by a grant (Formación de Personal Investigador) from Spanish Ministry of Science and Innovation. BH and EIS are supported by grants from the Swedish Research Council, and MW by a Marie Curie Intra European fellowship. Permits to capture damselflies in Spain were issued by each Regional Government to RSG.

Detecting balancing selection in genomes

One challenging question in evolutionary genetics is how the multiple changes required to generate a functionally new variant can accumulate despite of drift and/or negative selection exerted on intermediate stages. This question is particularly challenging for traits under balancing selection as new variants arise and co-occur with alternative variants, with ample opportunity for recombination between them. Three main mechanisms have been recurrently proposed to promote the emergence of differentiated phenotypes: first, the recruitment of major-effect mutations, second, gene duplication, and finally introgression of ‘ready-made’ alleles from hybridizing species.

Recruitment of large-effect mutations

Classical models propose that new adaptive alleles could be formed through a first mutation of large beneficial effect, followed by successive mutations of smaller effect which further hone the phenotype (‘two step theory’, (Poulton 1912 )). For example, in species exhibiting polymorphic mimicry, an initial mutation might result in imperfect mimicry of a new model, with sufficient resemblance to the new model to confer some degree of protection from predators, and this mutation could be followed by successive mutations of smaller effect size, refining the mimetic phenotype (Nicholson 1927 ). However, for traits showing balanced polymorphisms, recombination might prevent successive mutations from accumulating, which might restrict the emergence of new functional variants to cases where one or few mutational events in ‘key’ genes are sufficient to produce differentiated phenotypes.

Genes with many pleiotropic effects or with effects in multiple tissues during development (‘developmental genes’) are sometimes proposed as good candidates where point mutations could induce large phenotypic changes. Key genes identified as underlying polymorphic mimicry in butterflies indeed tend to have a large phenotypic effect and are involved in developmental pathways known to influence spatial identity in wing cells. For instance, the gene engrailed, a major transcription factor controlling morphogenesis in different tissues and stages, is associated with female-limited polymorphic mimicry in the Papilio dardanus (Timmermans et al. 2014 ). Similarly, the gene doublesex, well-known for its role in the sex-determination cascade as well as in sexually dimorphic morphologies in Drosophila and other insects, controls female-limited polymorphic mimicry in Papilio polytes (Kunte et al. 2014 Nishikawa et al. 2015 ) as well as in Papilio memnon (Komata et al. 2016 ). The recruitment of doublesex for sex-specific variations in wing colour pattern might be explained by its pre-existing sex-specific role. The repeated use of the same genes involved in different wing phenotypes in different species, as found in the Lepidoptera, may indicate that tweaking the regulatory regions of genes expressed in the target tissue could facilitate the rapid emergence of new adaptive variants (Martin et al. 2014 van't Hof et al. 2016 Nadeau et al. 2016 Wallbank et al. 2016 ). Similarly, the transcription factor gene pdm3 has been repeatedly recruited in several species of the Drosophila montium subgroup in the control of female-limited pigmentation polymorphism, probably involving several independently evolved variants of its cis-regulatory region (Yassin et al. 2016b ). By contrast, this evolution of similar genetic architectures contrasts with that found in Drosophila erecta where female-limited pigmentation polymorphism is controlled by an enhancer of the well-known melanization gene tan (Yassin et al. 2016a ). In summary, it is unclear whether pleiotropic genes are recruited in the control of differentiated forms because their pleiotropy allows them to control multiple differences simultaneously. An alternative is that pleiotropic genes are especially likely to be expressed at a key developmental time and/or in relevant developing tissues where phenotypic differences may be controlled, making them repeated targets of balancing selection.

More generally, it is still unclear to what extent genes expressed in multiple different tissues or positioned at certain stages in development are more likely to evolve into polymorphic loci switching between differentiated phenotypic forms. One of the reasons may be that we still ignore much of the genetic underpinnings of modularity itself, that is how easily the recruitment of a gene positioned upstream in a developmental cascade may result in the co-option of its entire downstream effects when expressed in a new tissue. The expectations regarding the distribution of mutational effects at a given locus are also still poorly defined and we may find that the multiple elements participating to a composite locus or a supergene may be expected to repeatedly hit the same gene, or target genes with large mutational possibilities, for example those with long and complex cis-regulatory regions (e.g. genes optix and doublesex in butterflies, Martin et al. 2014 ).

Gene duplication

The emergence of divergent alleles characterized by the accumulation of several mutations affecting traits under balancing selection might be facilitated by gene duplications. Gene duplications frequently arise as a result of errors in DNA replication and are thought to be an important source of genetic variation. In the case of heterostyly in Primula vulgaris, the switch between pin (long style/short anthers) and thrum (short style/long anthers) morphs is now known to be derived from the duplication of a floral homeotic gene, followed by a neo-functionalization of one copy allowing the emergence of the thrum phenotype (Li et al. 2016 ). Several other documented traits under balancing selection, in particular those involved in pathogen recognition, display frequent gene duplications (R-genes in plants (McDowell & Simon 2006 ), MHC in Vertebrates (Piertney & Oliver 2006 ), resulting in large copy number variations: for instance, the number of MHC genes is highly variable across species (Mehta et al. 2009 ), but also within species (Bonhomme et al. 2008 ) and populations (Eimes et al. 2011 ). Copy number variations can result from birth and death processes well described in multigene families (Nei & Rooney 2005 ), where duplicated copies can be promoted by natural selection and subsequently lost because of drift or changes in adaptive value. In the evolutionary arms race between hosts and pathogens, new MHC gene copies are indeed regularly promoted when they recognize frequently encountered pathogens, but some copies are also frequently lost because they are no longer adapted to current pathogen communities. The presence of several copies creates a range of variation which can be neutral most of the time but may sometimes allow a rapid response to the invasion of new pathogens. However, gene conversion, by allowing genetic exchange among the different copies, might favour rapid adaptations to new pathogens but may also homogenize variations among copies, and therefore limit divergence. Furthermore, in some cases, duplication can erase heterozygote advantage by creating a superior haplotype with both alleles in tandem, as observed in the insecticide resistance gene ace-1 in the mosquito Culex pipiens (Labbe et al. 2007a ). Duplications can therefore facilitate the emergence of differentiated variants, but may also limit within-locus polymorphism.


Another source of highly divergent adaptive variants highlighted in balanced polymorphisms is the introgression of new haplotypes favoured by selection. This is illustrated by the recent adaptive introgression of a MHC allele from domestic goat to Alpine Ibex (Grossen et al. 2014 ), or the multiple introgressions of alleles from the self-incompatibility locus in the closely related species Arabidopsis halleri and A. lyrata (Castric et al. 2008 ). In the white-throated sparrow, the white haplotype, an inversion-based supergene allele of >100 Mb which confers differences in pigmentation and components of social behaviour compared with the alternate white haplotype, was also recently suggested to be introgressed from a closely related species (Tuttle et al. 2016 ). The strength of balancing selection acting on alternative variants probably favours the capture of new haplotypes already shaped by selection in isolation in a related lineage. Introgression thus appears as a potentially important driver of adaptation, allowing the recruitment of variants carrying multiple co-adapted mutations. Adaptive introgression is also particularly likely when NFDS operates on new alleles which benefit from their rarity upon entering a new population.


Accurately fitting rational functions to the frequency response of modal impedances is crucial for including frequency dependency in lumped parameter models of transmission lines. Vector fitting is widely used for fitting rational functions to the frequency response of modal impedances and then an R-L equivalent circuit is obtained from the rational functions. A single-step method based on the properties of Foster equivalent circuit is proposed to directly fit an R-L equivalent circuit to the frequency response of modal impedances. The positive peaks, negative peaks and the positive zero crossings of the slope change plot of the frequency response are found to provide a good approximation of zeros, poles and the flat region locations of the frequency response. An enhanced fitting algorithm is proposed based on these observations. A close enough fitting is achieved using the proposed method with less number of passive elements. Using the proposed model in EMTP-RV, 400, 765, 1200 kV transmission lines and a 11-bus 500 kV network are simulated for switching transients. The results are compared with the constant parameter cascaded π -model and the Frequency-dependent line model in EMTP-RV. The switching transient results of the proposed model are found to be comparable to the Marti's model in EMTP-RV.


We are grateful to Renata Pardini, Jean-Paul Metzger and Fabiana Umetsu (University of São Paulo) as well as Christoph Knogge and Klaus Henle (UFZ Leipzig) for unfailing scientific and logistic support as well as very constructive discussions. We would like to thank Jörg U. Ganzhorn for constant backing. Furthermore, we thank our numerous field assistants and Brigit Bieber and Anke Schmidt for assistance in the genetic laboratory. We are grateful to Jürg B. Logue and Tomer Czaczkes for linguistic advice, as well as to three anonymous reviewers for their constructive scientific comments on an earlier version of the manuscript. We appreciate the financial support provided by the BMBF Germany (German Federal Ministry of Education and Research, project ID: 01 LB 0202). This study is part of the BIOCAPSP project (Biodiversity conservation in fragmented landscapes on the Atlantic Plateau of São Paulo, Brazil).

Watch the video: Forms of Selection (January 2022).