Phenotype is determined partly by the interaction of genotype and environment and partly by random internal noise during development. The phenotype is also generally robust to the combined effects of mutation, environmental change, and internal noise 1. This robustness is determined by the interplay of canalization and developmental stability, two types of developmental buffering that probably share underlying regulatory mechanisms but are functionally distinct 2, and also phenotypic plasticity, an adaptive change of phenotype in response to different environments 3. Debat and David 4 define developmental stability as "a set of mechanisms historically selected to keep the phenotype constant in spite of small, random developmental irregularities potentially inducing slight differences among homologous parts within individuals." While the assumption that developmental stability is largely the result of selection may be debated, instability during development generally is thought to indicate stress. Primary interest in the subject of developmental instability has been fueled by its potential utility as a general environmental bioindicator of environmental or genetic stress 567 or as an indicator of good genes (i.e. stable development) within the context of mate choice 89. Traditionally, the term "developmental instability" has been equated loosely with "developmental noise" 10, however, because developmental instability is often responsive to the environment and, yet, always exists to some degree in the absence of environmental changes, it must have both internally- and externally-driven components. In this paper, I restrict the definition of developmental noise to the internal component of developmental instability and investigate the internal accuracy of the developmental process in a relatively unstable morphological trait, the aphid wing.

Developmental instability most often is estimated by fluctuating asymmetry (FA), the right and left side difference in size or shape in a single trait across the population 6111213. Generally, it is assumed that FA is the result of both measurement error and some level of genetic-based buffering of environmentally-linked noise during development. Therefore, the average level and variability of FA observed in any population potentially depends on four influences: 1) some normally distributed measurement error, 2) an environmental sensitivity in the development of the phenotype, 3) a genetically based and, presumably, variable capacity to buffer this sensitivity, and 4) some level of stochastic internal noise that always is potentially present during the developmental process, even in the absence of gene by environment interaction. In more modern studies of FA, measurement error is controlled carefully. However most estimates of FA reported in the literature cannot speak to the relative contribution of the remaining factors to overall FA within a sample because they usually are collected from genetically heterogeneous populations.

Biologists have often assumed either that FA is driven mostly by uniform individual responses to a variety of stessors encountered by a natural population in a heterogeneous environment or that FA is driven mostly by variability among individual capacities to buffer against a relatively uniform level of stress presented by a homogenous environment. This difference of opinion as to whether FA is mostly environmentally-based or has a significant genetic component is debated 1415, and often seems to rest upon whether an author's primary aims are in demonstrating the utility of FA as a bioindicator of environmental stress or investigating whether organisms can use FA as an indicator of stable (i.e. "good") genes during mate choice. Both views implicity assume that the level of FA is highly reflective of the combination of gene and environment that occurs with each individual in a population, thereby also assuming the development of most traits is relatively robust to internal noise (Figure 1) 16. This assumption that FA is not influenced heavily by internal stochastic processes during development is necessary if one is to hypothesize that FA could act as any kind of indicator (i.e. either environmental or genetic quality). Furthermore, this assumption largely remains unexplored due to the simple fact that the "individual" phenotype develops only once. However, in the absence of genetic variation, developmental noise could hypothetically be estimated through the comparison of intra-individual to inter-individual variation that represent, respectively, the influences of the internal and external environments on phenotype. Therefore, the response of FA in isogenic or monoclonal populations reared in a graded environment could be used to quantify the level of developmental noise within phenotypic traits and compare it to the response of developmental instability to the environment. Only a few studies have observed FA in isogenic or clonal organisms 2717181920, and none for purpose of quantifying the role of developmental noise in comparison to phenotypic response to the environment.

The body of a typically sized adult human is estimated to contain between 10 and 100 trillion cells. Building this body, therefore, requires roughly 50 doublings of the initial cell population (by comparison, an insect still may require ≈ 45 cell doublings). Milan et al. 21 has demonstrated that cell cycle synchrony is maintained in the Drosophila imaginal wing disc over an average distance of only 2–10 cells (2–4 doublings) regardless of the stage of development. This large difference between adult body size and the comparatively small extent of synchronized cell behavior implies a large potential for the accumulation of errors caused by random differences in the timing of cell cycles within expanding cell populations during growth and development. This source of developmental noise is amplified strongly by expanding populations of growing cells during the exponential growth phase, causing the multiplicative accumulation of developmental error. Without mechanisms to regulate this kind of noise, the ascertainment of an accurate and symmetric phenotype would be nearly impossible. However, empirical observations of the symmetry of various morphologies demonstrates that most organisms actually exhibit low levels of developmental instability, as estimated by FA. Clearly, while the potential for error that leads to developmental noise must be regulated dynamically during the growth process (see 22), it is also probably never eliminated completely because of the lack of cell cycle synchrony between neighboring clusters of cells.

Despite the likely contribution of multiplicative error to FA, it has traditionally been assumed, that developmental instability somehow originates at a subcellular molecular level and that these effects are independent and additive in their contribution to overall body asymmetry [1112 and 23 but also see 242526]. Simply because the difference in scale between the size of molecules and the size of cells is so large, it would be unlikely that additive and independent effects of molecular noise (in a traditional Brownian sense) would comprise an important source of variation in the functioning of growing cells. This is because accuracy in the "laws" that describe physical-chemical systems has a relative probable error of 1n MathType@MTEF@5@5@+=feaafiart1ev1aaatCvAUfKttLearuWrP9MDH5MBPbIqV92AaeXatLxBI9gBaebbnrfifHhDYfgasaacPC6xNi=xH8viVGI8Gi=hEeeu0xXdbba9frFj0xb9qqpG0dXdb9aspeI8k8fiI+fsY=rqGqVepae9pg0db9vqaiVgFr0xfr=xfr=xc9adbaqaaeGacaGaaiaabeqaaeqabiWaaaGcbaqcfa4aaSaaaeaacqaIXaqmaeaadaGcaaqaaiabd6gaUbqabaaaaaaa@2ED6@ with n equal to the number of molecules cooperating to bring about the "law" 27. Hence, the traditional explanation, that FA results from additive effects of chance events at the molecular level that accumulate to produce a normal distribution of FA would seem to lack logical grounding. However, Leamy and Klingenberg 28 rationalize that molecular noise could scale to the level of tissue when developmentally important molecules exist in very small quantities (e.g., regulated transcript or protein), and therefore FA actually may represent a stochastic component of developmental gene expression. The temporal variation in cell cycle that this could cause is also implicated by the observation of high kurtosis (or, more specifically, in this case a non-normal distribution due to power law scaling in the tails) often observed in FA data 2529. Babbitt 30 has also linked increased scaling phenomena (and kurtosis) to increased genetic heterogeneity in a population.

Despite the potential for individual differences in gene and environment to influence the developmental stability of the phenotype, almost nothing is known about the overall level of internal developmental noise in a typical phenotypic trait. More importantly, how does this level of noise compare to the response of fluctuating asymmetry observed when an organism's environment changes? This is the central question in this investigation, which reports the percentage of variation in FA due to noise compared to the environmental response of FA along a temperature gradient in a genetically homogeneous population of organisms with complex morphology.

In this study, both the noise component of FA and its response to environment (temperature response) in the cotton aphid, Aphis gossipyii, is characterized. This species can reproduce parthenogenetically (apomictic) and often produces wings that are easily measured using multiple landmarks. Cotton aphids demonstrate large visible variation in body size, wing size, and wing FA, even within monoclonal cultures and are one of the few insects that demonstrate quite visible wing asymmetry within many individuals. Cotton aphids are also phenotypically plastic in response to temperature, producing smaller lighter morphs at high temperatures and larger, darker morphs at low temperatures. This unique feature allows observation of two genetically homogeneous groups in which differences in gene expression exist (causing the two color/size morphs). Therefore, this model system can allow for partitioning of the effect of variation in gene expression from variation in genotype on developmental stability.

Under the conditions of this experiment, the aphid wing was found to be a highly developmentally unstable trait, by comparison to levels of FA usually observed in the wings of other insects (see Figure 3), exhibiting nearly four times the level of FA that is typical of other insects 29. This was clearly not the result of directional asymmetry possibly inflating the overall FA measured in this study. The surprisingly high noise-to-response ratios discovered here, combined with relatively low measurement error, also indicated that nearly half of the response of FA to environmental temperature could be also generated by internal stochastic developmental noise. Therefore, this level of developmental noise could impose a significant lack of sensitivity of FA, to either environment changes and/or genetic differences, in the ability to buffer development between clonal populations. If developmental noise varies across species and traits, this may explain why FA often has been observed to be highly species and trait specific 14. High percentages of developmental noise in aphid wing symmetry are also interesting because they may imply that the high FA in the aphid wing may also be associated with a lack of evolvability on this trait, due to the interference that a noisy phenotype would have on any selection pressures imposed on the population. This may be rather unusual in the evolution of insect wings, as aphids are notoriously weak flyers and use their wings primarily for passive dispersal on wind currents. Therefore, there may be no real need for accurate development of wings in this group. This further suggests that, in general, stable development may come at some unknown physiological cost, perhaps a cost that some organisms, like aphids, refuse to pay when selection is relaxed. The observation of winged individuals in other aphid species that lack flight muscles and thus are incapable of powered flight 39 would also seem to support the hypothesis of relaxed selection on wing symmetry in aphids. This type of elevated response of wing FA to relaxed selection has also been observed in flightless beetles 40.

The comparison of developmental noise between the two phenotypically plastic color/size morphs in this species indicates no differences in developmental noise regarding wing size but a large difference regarding wing shape, with the development of a specific shape being much noisier in the light (high temperature) morphs. Temperature trends in mean FA, which are sloped differently in the two color/size morphs (Figure 4), were also found to be different regarding whether a size or shape-based approach was used. Additionally, while the low levels of kurtosis in the distribution of FA are consistent with those of a genetically homogeous population (see 30), it is interesting that the average kurtosis differs significantly between the light and dark aphid morphotypes (Figure 5). In combination, all of these observed differences in the response of wing symmetry to temperature in these two isogenic, but differentially expressed genotypes may indicate the importance of gene regulation and expression in determining many aspects of phenotype. More directly, these results demonstrate that differences in the pattern of gene expression can also lead to fundamental differences in how developmetal noise is propogated within individuals. This lends further support to the hypothesis of Leamy and Klingenberg 28 that the basis of developmental instability results from a stochastic component of developmental gene expression.

Both environmental temperature and FA potentially interact with growth rate and size in ectothermic species and, therefore, they should indirectly influence each other. There are several ways that FA might respond to temperature. First, increased temperature may increase molecular perturbation, which may further act to increase overall levels of developmental noise during development. This should predictably increase FA with temperature. Second, increased temperature may act as a behavioral cue to shorten development time 41, thereby reducing the total time in which developmental errors may occur. This should predictably have the opposite effect, reducing FA as temperature increases. A third possibility is that a species specific optimal temperature exists. If so, FA should increase while approaching both the upper and lower thermal tolerance limits of organisms.

Only a few studies have directly investigated the relationship between FA and temperature. The results are conflicting. FA is found to increase on both sides of an "optimal" temperature 424344, to be highest at low temperature 45, to simply increase with increasing temperature 4647, or not to respond at all 48. In none of these studies were genetic differences between individuals in populations controlled for, therefore, the relative contribution of internal developmental noise versus external environmental effect could not be assessed.

In this study, there is a clear overall decrease in FA with increasing temperature. This strongly supports that developmental instability is a function of developmental time rather than growth rate. Thus, the longer time spent in development and therefore, the more interrupted the growth process becomes, the larger the accumulation of multiplicative developmental errors. This was especially evident in the curve of centroid size-based FA, which closely follows the curve of development time. The slope of the temperature trend in mean FA is much steeper in dark morphotypes, which also have significantly higher FA, further indicating that environmental influences on FA are primarily related to individual developmental time. This result is consistent with the explanation of the basis of FA by Emlen et al. 24 and counters the assumption of Moller and Pomiankowski 9 that rapid growth is stressful and should, therefore, impart higher FA.