A comparison of the Giardia lamblia trophozoite and cyst transcriptome using microarrays

Giardia lamblia (G. duodenalis, G. intestinalis) is a diplomonad parasite which causes over 20,000 reported cases of giardiasis a year in the United States [1]. In addition to its importance as a widespread human and animal pathogen, the long evolutionary history of the diplomonad lineage makes G. lamblia an interesting system for studying eukaryotic evolution and the evolution of parasitism. Research on G. lamblia is aided by the fact that the entire life cycle can be studied outside the host, and that the differentiation from cyst to trophozoite and the reverse process of encystation can be reproduced in vitro. Recently, the availability of the complete genome sequence [2–5] have facilitated genome-wide analyses.

Although many Giardia proteins and organelles have been studied in detail, genome-wide studies of the transcriptome and proteome have been few [6–11]. No microarray analyses of the transcriptome of cysts obtained from infected animals have to our knowledge been performed. Serial Analysis of Gene Expression (SAGE) was used to survey changes in the G. lamblia transcriptome during encystation and excystation [9]. This study grouped about 10% of predicted G. lamblia genes into six clusters with related transcriptional profile. SAGE was also used to analyze the relative abundance of transcripts encoding cytoskeleton proteins [8]. This analysis found that the level of mRNA transcripts encoding proteins localized in the adhesive decreases as the parasite encysts, and also found a lack of association between mRNA and protein level. Morf and co-workers focused on transcriptional changes associated with encystation [12]. This study used micoarrays to identify genes which are induced during encystation and found evidence of transcriptional co-regulation mediated by a shared transcription factor binding motif in the promoter region of such genes.

The extensive morphological changes which take place during the parasite's life cycle have for years motivated the study of transcriptional regulation of selected genes during differentiation. Reverse-transcription PCR has been frequently used to monitor changes in the level of specific mRNA transcripts, such as those encoding enzymes involved in energy metabolism [13], recombination [14], structural functions [15] or regulatory functions [16].

We wished to compare on a global level the transcriptional landscape of trophozoites and cysts. We found that in cysts many genes are either not transcribed, or that the transcripts they encode are too rare to be detected with microarrays.

The fact that the entire life cycle of G. lamblia can be reproduced in vitro makes this species an attractive model to study the differentiation of cyst into trophozoite and the reverse process of encystation. Recently, genome-wide studies of G. lamblia transcriptional regulation have been undertaken [9, 12] but no global comparison of the cyst and trophozoite transcriptome has to our knowledge been published. The study of the trophozoite and cyst transcriptome is relevant to understanding the G. lamblia life cycle and the evolution of encysted forms which are essential to the survival of many enteric organisms. Given that cysts don't divide and are assumed to have little metabolic activity, it is likely that for many proteins in cysts no mRNA is present. Combined transcriptome and proteome analyses [7] will generate a more comprehensive view of the composition and metabolic activity of cysts.

Microarray and RT PCR data clearly show that the cyst transcriptome is much reduced in terms of abundance and complexity as compared to that of trophozoites. DAVID analysis of over-represented GO terms [19] suggests an overall resemblance in the composition of the transcriptome throughout the life cycle, but the analysis of highly expressed genes highlights significant differences.

As in most quantitative analyses, the comparison of microarray data required calibration against a benchmark. As described in Methods below, we used RNA quantity of as benchmark by using an equal amount of amplified RNA for preparing Cy3 labelled probes. The differences in transcript levels are thus to be interpreted as relative to total RNA extracted from cysts and trophozoites.

To what extent rRNA and tRNA which constitutes the bulk of cellular RNA varies is unknown. An alternative calibration would have been to normalize the data against the number of cysts, trophozoites or nuclei. This approach was discarded because of the possibility that extraction of RNA from cysts is less efficient than extraction from trophozoites. Had we chosen to normalize against cell number, it would have been difficult to assess whether differences between cyst and trophozoite were genuine or a result of cyst nucleic acid being more difficult to extract.

The experiments were constrained by the fact that G. lamblia microarrays are designed from the assemblage A genome and that the only source of cysts we could identify uses assemblage B. Because DNA sequence identity between assemblage A and B genome averages 77% [3], the possibility that analyzing assemblage B cyst cDNA with assemblage A microarrays could artificially reduce the hybridization signal was considered. Replicate microarray hybridizations were performed with cDNA originating from assemblage A and B trophozoites (Additional file 1). These controls showed no evidence of differential hybridization of cDNA originating from different assemblages under the hybridization conditions we used. This does not exclude that highly polymorphic transcripts were missed, but indicates that for the vast majority of genes annealing to the 70 mer microarray oligonucleotides was sufficiently stable to tolerate mismatches. Moreover, the vast majority of fluorescent signal from Arabidopsis control spots and empty spots present on the array were well below background (mean Cy3 fluorescence = 1552, n = 3860), confirming the specificity of the hybridization signal and demonstrating adequate stringency of the hybridization protocol.

Because we expected significant differences in the magnitude and diversity of cyst and trophozoite mRNA transcriptome we did not directly compare trophozoite and cyst transcriptome using a conventional 2-color microarray protocol. Two-color microarrays require normalization to eliminate the effect of differential labelling of dyes, which is typically accomplished with microarray analysis programs [20]. These programs normalize Cy3 and Cy5 fluorescence based on the assumption that the samples being compared contain similar amounts of mRNA, as would be the cases with, say, healthy and diseased cells. Since we did not expect this assumption to hold, we chose to use only background-subtracted single-channel Cy3 fluorescence values. Since these data originated from calibrated amounts of Cy3 labelled probe, the resulting data are directly comparable. In the context of this study, an additional advantage of the single-dye design over a more conventional Cy3/Cy5 ratio is the feasibility to include fluorescence values below background, i.e., values equal zero. Since a large proportion of transcripts were not detected in cysts, the exclusion of ratios with a numerator or denominator equal zero would have excluded biologically relevant information.

The elevated expression of some genes observed in the microarray dataset confirms previous observations. For instance, we found high levels of ubiquitin mRNA in trophozoites and cysts, which is consistent with previous RT PCR analyses [21]. The expression of ubiquitin in trophozoites is not unexpected, but the abundance of ubiquitin mRNA in cysts suggests extensive protein turn-over. Other top-ranking genes in cysts and trophozoites include histone. This observation is consistent with the constitutive expression of various histones during the trophozoite mitotic cycle [22], but had not been observed previously in cysts. The absence of mRNA encoding histone modifying enzymes suggests that histone modification does not occur in cysts, and is consistent with many genes not being transcribed in this phase of the life cycle. This interpretation is in agreement with the previously observed decrease of histone acetylation during trophozoite encystation and the predicted importance of epigenetic regulation of transcription in the life cycle of G. lamblia [23]. Finally, we notice the unexpected expression in cysts of several genes encoding variant surface protein.

The comparison of SAGE and microarray data raises interesting questions regarding the properties of cysts produced in culture. Cysts encysted in vitro have been extensively characterized with respect to morphology, antigenic property [24], and cyst wall biosynthesis [25], as have many processes occurring during encystation. A direct comparison of the transcriptome and proteome of native cysts and cyst produced in vitro has to our knowledge not been performed. In light of the results presented here, such an analysis is warranted to assess to what extent cysts produced in vitro can serve as surrogates for native cysts. As RNA-Seq has become a more widely available technique for transcriptome profiling, an accurate comparison of the cyst transcriptome is now feasible.

The transcriptome of G. lamblia cysts and trophozoites was investigated using oligonucleotide microarrays. Although in both life cycle stages transcripts related to ribosomal function are overrepresented, clear quantitative differences were observed. This global comparison of the cyst and trophozoite transcriptome indicates that, in comparison to trophozoites, in cysts only about 5% of mRNA species are expressed at level detectable with microarrays.

The comments and views detailed herein may not necessarily reflect the views of the WateReuse Research Foundation, its officers, directors, employees, affiliates or agents.

Microarray data were deposited in the GEO database [GPL:11228].