- Top of page
Weeks et al. (2006) have reported their inability to find a cline in the frequencies of the major Thr-Gly encoding length variant alleles of the period gene in Drosophila melanogaster in Eastern Australia. This is in contrast to a study by Sawyer et al. (2006), who found a cline on this continent from samples collected in 1993. Weeks et al. then cast doubt on the validity of a robust cline found for these variants in Europe by Costa et al. (1992), criticizing their molecular techniques and sampling methods. We show how these claims are unjustified, and reveal a number of potential problems in their own methodology. Finally by reanalysing the subset of their data which they state is more reliable, we suggest that their results from Australia may be reasonably consistent with our own.
Weeks et al. (2006) have described a study where they examined natural length polymorphism in the Thr-Gly repeat of the period gene and in the poly-glutamine encoding repeat of the Clock gene in Drosophila melanogaster populations collected from Eastern Australia in 2000, 2002 and 2004. In both cases, they fail to find any evidence for robust clines. In the case of the Thr-Gly repeat, they contrast their results with those from Costa et al. (1992), who found a cline in Thr-Gly length in European populations whereby the (Thr-Gly)20 allele predominated in the higher and the (Thr-Gly)17 in the lower latitudes. They conclude that because no cline could be found in Australia, this casts doubt on the validity of the European cline.
We would like to point out that in an extensive study of Thr-Gly variation in Australia and Africa, we find a significant cline for one of the major Thr-Gly length variants in Australia that is consistent with the one in Europe (Sawyer et al., 2006). Although this cline is weaker than that found in Europe, we concluded that this may be because D. melanogaster had been introduced to Australia only 80–90 years before our collections were made (compared with c. 10–15 000 years in Europe, David & Capy, 1988).
Weeks et al. further question the resolving power of the agarose gels used by Costa et al. (1992) to define the different length Thr-Gly variants. Their logic is that when they moved to acrylamide, they found a greater number of variants in their Australian samples, namely 8, than did Costa et al. (1992) in their European samples. Their implicit assumption is that European samples must have the same number of alleles as Australia, and that Costa et al. (1992) misclassified these. However, from a similar collection of flies sampled from eastern Australia in 1993 (James et al., 1995, 1997), which we used to examine Thr-Gly length polymorphism (Sawyer et al., 2006), we resolved all eight of these Thr-Gly length alleles, and at similar frequencies to Weeks et al. (2006). In addition, the frequencies of the rarer Australian alleles that fell outside of the 14–17–20–23 Thr-Gly repeat periodicity, were also similar to those of Weeks et al. (2006). Consequently, the power of agarose in our hands where differences in length of 6 bp (encoding a Thr-Gly dipeptide) can be easily resolved (see for example Costa et al., 1991; Rosato et al. 1994), is as good as that of Weeks et al. (2006) with acrylamide. Thus their criticism that Costa et al. (1992) may have misclassified Thr-Gly length variation, is demonstrated to be incorrect.
Secondly, in Sawyer et al. (2006), the 1993 Australian sample that we analysed does show a significant latitudinal cline for the (Thr-Gly)20 [not the (Thr-Gly)17 allele as stated in error in the discussion of Sawyer et al. (1997)], that is consistent with the European cline of Costa et al. (1992). Given the population history of D. melanogaster in Australia vs. Europe (David & Capy, 1988), it is hardly surprising that the Thr-Gly cline is not as robust as that in Europe. In addition, those very rare variants that we find in Europe, but we find at significantly higher frequencies in Australia, and which fall outside the (Thr-Gly)3 periodicity, tend to be found in the lower latitudes of Australia, just as in the data of Weeks et al. (2006). As these variants have poorer temperature compensation of their circadian clock compared with the major variants (Sawyer et al., 1997), this may explain why they are tolerated rather more in the Australian tropics, than in Europe. In this regard, Weeks et al. (2006) overlook the study of Zamorzaeva et al. (2005), who studied Thr-Gly length variation in ‘Evolution Canyon’ in Haifa, Israel. Not only did they find that the (Thr-Gly)17 variant predominated in that general area of southern Europe/Middle east (consistent with Costa et al., 1992), but they also found that on the cooler northern slopes, the ratio of (Thr-Gly)17 to (Thr-Gly)20 frequencies was significantly reversed compared with that on the hotter southern slopes, again totally consistent with the thermal hypothesis (Costa et al., 1992; Sawyer et al., 1997; Peixoto et al., 1998; Sawyer et al. 2006).
Thirdly, in their 2000 and 2002 collections, Weeks et al. took fertilized females from the wild, then in the F1, took another female, froze it and subsequently performed the genotyping. Distinguishing females of the sibling sympatric species D. simulans from D. melanogaster is difficult (Ashburner, 1989 and refs therein), so we presume that they checked the brothers of these females prior to genotyping, to make sure that they were D. melanogaster. In their 2004 collection however, field females and males were stored immediately and then genotyped. Thus in this sample we are confident that the males were D. melanogaster (they are very easy to distinguish from D. simulans), but not the females (Ashburner, 1989). As D. simulans have a different profile of Thr-Gly repeats lengths (Wheeler et al., 1991; Rosato et al., 1994; Rogers et al., 2004), this may have introduced simulans‘contamination’ into the 2004 sample. DNA sequencing of their PCR products would have clarified this potential problem, but they did not appear to sequence their alleles.
In addition, there are ambiguities associated with incompletely denatured PCR products that have secondary structures and give anomalous mobilities (as occurs with the use of denaturing gels by Weeks et al. (2006), but not with agarose as used by Costa et al., 1992). To complicate things further, as per is sex-linked, females should not be used to avoid problems with heteroduplex identification. We use males in our studies because of these potential problems and always sequence routinely not only any potential new length variant, but also iso-length variants at random because we are also interested in the sequences of iso-length repeats (Costa et al., 1991; Sawyer et al., 2006). Thus there are a number of technical questions arising from Weeks et al.’s (2006) study that require attention.
Weeks et al. (2006) also suggest that the methods of Costa et al. (1992) were not clear. In that study, we state that one male per isofemale line was sampled in that study, and all individual flies analysed that were from isofemale lines established immediately after collection or from mass cultures containing thousands of flies within five generations (Costa et al., 1992). The fact that flies were collected in different years, 1989, 1990, 1991 should potentially add noise to the results, yet the cline was statistically robust. Altitude was also controlled, as described by Peixoto (2002) in a review of all of our geographical studies on European populations from 1989 to the mid-1990s (Costa et al., 1992; Rosato et al., 1996, 1997), which incidentally, also shows that between years, the cline appears to be quite stable. The one ‘outlier’ we found, came from high altitude, and is therefore also consistent with a thermal hypothesis (Peixoto, 2002).
Finally, as an academic exercise, we took the acrylamide-based data from the Weeks et al. (2006) collections from 2002 and 2004, in which they state they are more confident, and averaged the frequencies of the (Thr-Gly)17 and (Thr-Gly)20 samples when they came from the same latitude (to the nearest degree), irrespective of the year. We did this because some of the changes in gene frequencies for a single location between the two sampling years were rather large, suggesting that they were sampling ephemeral and nonstable populations. We took out the ‘Bowen 2004’ sample that was based on the smallest number of alleles scored (n = 18), and the Tasmanian samples, because it is an island and more prone to drift. We also found similar high variance in Tasmanian samples in our study, although it did not affect our identification of the cline (Sawyer et al., 2006). For the (Thr-Gly)17 frequencies we obtained a negative correlation with latitude of −0.271 (regression slope of −0.0054, SE 0.0044), and a positive correlation of 0.29 (slope +0.0064 ± 0.0048) for (Thr-Gly)20 (both N’s = 21). The results are in the direction predicted by Costa et al.’s (1992) analysis of the European polymorphism and the difference between the two correlations is marginally significant (z = 1.69, P = 0.0462, one-tailed).
The study of Weeks et al. (2006) is not a replication of the European one, and their critiques of the methodology of Costa et al. (1992) do not stand up to any serious examination. If their study does stand similar scrutiny, then their conclusion may in fact be similar to that of Sawyer et al (2006), that the Thr-Gly cline is Australia is not as robust as in Europe, probably due to the population history of the recently invading species. If we take Weeks et al. (2006) at face value, we might add that the younger Australian Thr-Gly cline is also more sensitive to inter-yearly changes in environmental variables, thereby modulating its expression.