Indeed, it is always possible to hypothesize that the centromeres of C. albicans exhibit GC and TA replication-related skews with unique characteristics in terms of direction and magnitude compared to the other origin loci.
Let's consider GC skews, which are the more significant. The evidence for a genome-wide replication-associated GC skew corresponding to an enrichment in G for the lagging strand is strong. Global analysis of interorigin intervals clearly shows an increase in GC skew with interorigin position (P values equal to 4x10-6 and 2x10-7 for third codon positions and for intergenes, respectively). It is true that the positions of replication origins determined through analysis of replication timing profiles are not precisely defined. This lack of precision makes it all the more remarkable to find a clear increase in GC skew along interorigin intervals. More precise localization of replication origins should increase the significance of this result.
If the GC skew variations observed at centromeres are linked to replication, then it means that in these parts of the genome, replication-related mutation rates are much higher than elsewhere and lead to different values of GC skews (with the lagging strand enriched in C, not in G). Since we don't currently know what determines the different mutation rates of the leading and lagging strands and since centromeres have special characteristics like the presence of CENP-A, this hypothesis cannot be formally ruled out.
However, it has to be noted that, in this case, the pattern of GC skew variations (Figure 3D in (Koren et al., 2010. http://www.plosgenetics.o...)) indicates that only the centromeric core sequences (on average about 5 to 6 kb on each chromosome) would be submitted to these centromere-specific substitution rates. It could be objected that Figure 3D in (Koren et al., 2010) and Figure 6A in (Marsolier-Kergoat, 2012) represent GC skew variations that take into account all sequences (genes and intergenes), and that the coding sequences bordering the centromeres could be submitted to selective constraints preventing GC skew formation. However, the same pattern of GC skew variation (with GC skew values going back to zero at a distance of ~ 3 kb from the zero-intersection point) is also observed when only intergenic sequences are taken into account (MCMK, unpublished results). If adjacent intergenic sequences located farther than 3 kb away from the centromere GC skew zero-intersection point were submitted to these centromere-specific substitution rates, they should also exhibit high skew values (either positive or negative), which is not the case.
To sum up, one of the conclusions of the article (Marsolier-Kergoat, 2012) could be qualified as follows: "Under the assumption that the amplitude and direction of replication-related skews are similar in all parts of C. albicans genome, our results indicate that the skew variations at centromeres are not associated with replication. If the skew variations at centromeres were linked to replication, then the amplitude and the direction of these skews would be specific to the core centromeric sequences."

However, even if the association of skew variations at centromeres with replication cannot be theoretically excluded, alternative hypotheses cannot be ruled out either. One possibility is that C. albicans centromeres are under selective pressure to form a bipartite structure with a 5' CA-rich sequence followed by a 3' GT-rich sequence on a single strand. It is said in the above comment that the sequences of C. albicans centromeres are "free of selective constraint", but this assumption is not obvious. In an elegant article (Ketel et al., 2009. http://www.plosgenetics.o...), Berman and collaborators showed that neocentromeres can form at multiple positions on C. albicans chromosome 5, which demonstrates indeed that there are no strict sequence requirements for centromeres on this chromosome. However, this article also showed that the loss rate of chromosomes harboring neocentromeres are higher than the loss rate of chromosomes with wild-type centromeres, and that neocentromeres are less stable and tend to associate with less CENP-A that wild-type centromeres. Moreover, another article (Sanyal et al., 2004. http://www.pnas.org/conte...) reported that the deletion of the centromere on C. albicans chromosome 7 resulted in high levels of chromosome 7 instability, suggesting that neocentromeres do not form efficiency on this chromosome. The facts that neocentromeres do not form efficiently on all chromosomes and are not as functional as wild-type centromeres suggest that some selective pressure could lead to the fixation in the centromeric sequences of mutations that improve their efficiency. It is therefore plausible that the specific composition of C. albicans centromeres could stem from functional constraints.

Marie-Claude Marsolier-Kergoat.