Microbial pathogens have evolved a number of strategies for protecting themselves from their hosts. One of these is the so-called heat shock response, which is elicited by a sudden increase in ambient temperature (13) and induces the synthesis of a limited set of proteins (called heat shock proteins [HSPs] or molecular chaperones). Homologues of known HSPs have been identified and partially characterized in E. histolytica (1, 11). With the aim of determining gene expression changes during E. histolytica's adaptive response during infection, we developed an oligonucleotide-based microarray with transcript information randomly obtained from a cultured virulent strain of the pathogen. Array analysis revealed that gene transcription in E. histolytica exposed to heat shock is dramatically reduced, since 471 of 1,131 unique genes were down regulated, whereas specific HSP-encoding genes were up regulated. In conjunction with real-time PCR results, these genetic information data reveal for the first time a very interesting differential allelic expression of key genes participating in virulence, such as the immunodominant antigen Gal/GalNAc lectin, certain cysteine proteinases, and the so-called 20-kDa antigen.

The aims of this study were (i) to establish a highly discriminating method for monitoring gene expression changes in E. histolytica and (ii) to determine the mRNA expression profile of E. histolytica cells growing in a drastically modified environment. We decided to construct an oligonucleotide-based microarray, using information obtained directly from sequence analysis of E. histolytica transcripts, a strategy that is generally thought to overcome problems due to gene redundancy and the presence of introns. A cDNA library of the virulent E. histolytica strain HM-1:IMSS and a liver-specific cDNA subtraction library were prepared and sequenced. The bioinformatic analysis of sequenced clones enabled us to define 1,300 bona fide transcripts, all from parasites growing in vitro and enriched with randomly chosen transcripts from parasites growing in the liver. A comparison of our file with the recent publication of the E. histolytica genome sequence (10) allowed us to detect and reference 96 transcripts that have not yet been annotated in the genome (referenced as gDNA in our public database at http://genoscript.pasteur.fr). Additionally, of the 1,131 individual genes in the list, 342 corresponded to coding sequences of unknown function. Finally, E. histolytica genes known to be involved in virulence or in cytoskeleton activities were also added to the file. Due to the high AT base content of the E. histolytica genome and in order to achieve a high specificity and limit background noise in the hybridization procedure, we constrained the design of our 70-mer oligonucleotides with the following parameters: a 32% GC content, AT stretches of no longer than 30 nucleotides, and the absence of degenerate bases. All experimental procedures for microarray setting and statistical analysis of data are compiled at http://genoscript.pasteur.fr and summarized in the supplemental material.

We exposed E. histolytica trophozoites growing exponentially at 37°C to a heat shock by moving them to a temperature of 42°C for 4 hours. Cell viability was monitored by using the trypan blue exclusion test, which indicated that 96% of trophozoites were still alive after the treatment. The 3.4-fold increase of Hsp70 indicated that a heat shock response was obtained under our experimental conditions (data not shown). Given that the array represented a relatively small number of genes and that major changes in gene expression are likely during changes in the parasite's milieu, we chose to compare two normalization methods. The first takes into account the overall mean value of E. histolytica gene expression. In the second method, normalization incorporates the use of external controls by spotting oligonucleotides (spikes) prepared from an unrelated source (Arabidopsis thaliana). In both cases, statistical analysis was performed using a paired Student t test. After drawing locally weighted regression normalization curves, we noted that normalization with spikes fitted the asymmetric data set well, with linear correlation coefficients between biological replicates and between technical replicates being >0.9 and >0.95, respectively; the analysis allowed us to conclude that the method using spikes showed a greater specificity for changes in gene expression observed after heat shock treatment, together with a more asymmetric data set (few induced genes and numerous repressed genes).

The most up and down-regulated genes under heat shock conditions were grouped together according to function (Tables ​(Tables11 and ​and2).2). Some of these sequences are discussed further below, and the complete list of clones is presented in our public, MIAME-compliant database (http://genoscript.pasteur.fr). The functional annotation of genes modulated by heat shock was organized according to (i) the gene's role in the heat shock response, (ii) potential regulatory factors, and (iii) virulence-related genes.

Genes encoding heat shock proteins were among the 40 most highly expressed genes (Table ​(Table1).1). We identified three isoforms of Hsp90, two isoforms of Hsp70, and several well-known chaperone cofactors, such as DNAJ, a protein containing a BAG domain, a homologue of cochaperone p23, a candidate presenting a TPR domain, and a protein with structural homologies with CHIP. In E. histolytica, heat shock activates ubiquitin and a component of the 26S proteasome regulatory complex, as well as several other ubiquitination components involved in the removal of damaged proteins. A copy of the Ehssp1 stress-inducible gene (16), which has a possible role in cellular adaptation to stress conditions, was up regulated.

One of the most striking findings generated by the microarray analysis was the observation of differential expression of alleles for the heavy-chain subunit of the Gal/GalNAc lectin, which is an abundant surface protein complex involved in contact-dependent killing and phagocytosis of target cells by E. histolytica (14). The Gal/GalNAc lectin is composed of two subunits, a light (Lgl), glycosylphosphatidylinositol-anchored, 31/35-kDa subunit and a heavy (Hgl), 170-kDa subunit. The complex is associated with an intermediate, 120-kDa subunit (Igl). Each lectin subunit is represented in the E. histolytica genome by multiple alleles; however, the functional distinctions between these alleles have not yet been established. Six genes with 53 to 85% identity encode Lgl isoforms, and at least two genes encoding specific Hgl isoforms (Hgl-2 and Hgl-3) have been isolated and well studied (15, 19). Furthermore, another Hgl isoform and an incomplete sequence have been described from the genome (10). We found that the Hgl-2 allele and one Lgl allele (which we suggest calling Lgl-6) were involved in the heat shock response, whereas all other lectin family components were not. In contrast, the Hgl-3 allele and all other Lgl and Igl subunits underwent gene repression during heat shock. This was also true for other factors that have been associated with the Gal/GalNAc lectin, such as peroxiredoxin (the principal thiol-containing surface antigen in E. histolytica). Peroxiredoxin is important during the transition from the anaerobic environment of the large intestine to the aerobic environment in human tissue (5) and has been found to associate with the Gal/GalNAc lectin in two-hybrid and cellular experiments (8). Our findings suggest that there is a need to assess Hgl-3's specific activity during oxygen peaks. Quantitative reverse transcription-PCR confirmed that transcription of the Hgl-2 allele (16.t0014) and the Lgl-6 allele (168.t0024) increased roughly sixfold during heat shock, whereas all other Hgl, Igl, and Lgl alleles were down regulated (see the supplemental material). Finally, one of the previously described 20-kDa antigen (202.t00007) (21)-encoding genes was found to be up regulated, whereas the second (132.t00016) was down regulated.