All chemicals were purchased from VWR (UK) unless otherwise specified. DTT, CHCA, iodoacetamide, and EDTA were obtained from Sigma Aldrich (UK). Modified porcine trypsin was a product of Promega (UK). HPLC grade acetonitrile, HPLC grade methanol and glacial acetic acid was purchased from Fisher Scientific (UK). Oocysts of C. parvum passaged in lambs (IOWA strain) were purchased from Moredun Research Institute (MRI, Scotland). This strain was continually passaged in sheep by MRI. Oocysts were concentrated by sucrose density centrifugation, washed and resuspended in phosphate-buffered saline (PBS; pH7.2). The parasite suspension was stored at 4℃ in the presence of 1,000 U per mL penicillin and 1,000 µg per mL streptomycin.

For one dimension electrophoresis, frozen sporozoite pellets were disrupted in 40 µL of gel loading buffer containing 50 mM Tris Hydrochloride (pH 6.8), 100 mM DTT, 2% (w/v) SDS, 0.1% (w/v) Bromophenol blue and 10% glycerol. The mixture was boiled at 100℃ for 10 min and chilled on ice before loading into the SDS-PAGE gel lane. A standard broad-range protein molecular weight marker (RPN 5800; Amersham Biosciences, UK) was used as the ladder in a separate lane. Polyacrylamide gels (12%) were made using a mini gel apparatus (BioRad, UK). The resolving gel consisted of 30% acrylamide in 1.5 M Tris-HCl (pH 8.8), 10% (w/v) SDS, 10% (w/v) ammonium persulphate (APS) and 10 µL N,N,N',N'- tetramethylethylenediamine (TEMED). The stacking gel consisting of 30% acrylamide in 1.5 M Tris-HCl (pH 6.8), 10% (w/v) SDS, 10% (w/v) APS and 5 µL TEMED was used for the quantification of protein extracts. The SDS electrophoresis buffer was prepared by dissolving 25 mM Tris-base, 192 mM glycine and 0.1% (w/v) SDS in 400 mL of double distilled deionised water. Separation was performed by electrophoresis at 120 V for 2 h and then the gels were stained with Coomassie Brilliant blue or by Colloidal coomassie staining technique [15].

Two-dimensional-nLC-MSMS analysis was performed using an Ultimate 2D nLC system (Ultimate Famos Switchos; Dionex, USA) in the standard configuration, interfaced via a 20 µm i.d 8 µm orifice Picotip (New Objective, USA) mounted on a Protana nanospray interface (Protana, Denmark) to a QStar Pulsar i mass spectrometer running the AnalystQS software (Applied Biosystems, USA). A 1 × 15 mm BioSCX trap, 0.3 × 5 mm PepMap trap and 75 µm × 15 cm PepMap column were used in the analysis (Dionex, USA). Flow rates were 30 µL min-1 on the high flow side and approximately 200 nL min-1 on 93 the low flow side. 10 salt cuts at 0, 20, 40, 60, 80, 100, 150, 200, 300 and 500 mM KCl were used and a gradient of 2~50% acetonitrile in water with 0.5% formic acid for the reversed phase separation. Data was collected using an IDA protocol with a 2s survey scan 400~2,000 Da, and the four most intense ions above a threshold of 20 counts not on the exclusion list chosen for analysis using 3s MSMS scans in the 50~2,000 Da range. Masses were then added to an exclusion list for 360s.

The CryptoDB proteome database (release 3.1) was used as a source to download the genome, EST and GSS datasets into a local server connected to the mass spectrometer.

The MASCOT searching of MS data was performed either against the non-redundant National Centre for Biotechnology Information (NCBI, USA) database or locally downloaded CryptoDB datasets.

The MASCOT search engine (Matrix Science, USA) was used to analyse the PMF and peptide fragmentation data. The MASCOT search against the genome sequence of C. parvum revealed a list of contigs with significant scores for individual peptides. The ion scores of the individual peptides were recorded from the MASCOT search output page and the BLAST searching of any putative protein sequence was performed through the linked web from the same page. However, this was not suitable in cases where the significant peptides were few in number, or located some distance apart in a long contig.

The MASCOT search against the NCBI database and locally downloaded Cryptosporidium genome sequences revealed a list of contigs with significant peptides. The sequence containing those peptides was then BLAST searched (protein-protein BLAST or BLASTp) to identify sequence similarity with proteins from other organisms. The interpretation of the score and sequence similarity from BLAST searching eventually led to the identification of putative or homologous protein sequences. The whole sequential steps of this data analysis towards the identification of putative or homologous sequences are illustrated in Fig. 1.

Briefly, the PMF data and peptide fragmentation data from MS analysis was searched against the NCBI database and a number of contig hits were revealed, for which it matched one or more peptides with a specific ion score for each of them. Once the Mascot score was found significant (as manifested by a direct match with a protein or EST in the database with a significantly high individual peptide score for which the entry was already submitted in the database) the identity was confirmed for that protein or its homologs. However, if the MASCOT score was not significant and the identified peptides had a high ion score or if they were closely located together (indicating peptides from one protein), they were further searched against the CryptoDB database. The search again revealed some contig hits and the relevant peptides, with or without a significant MASCOT score. The peptides with insignificant ion scores were then discarded while those with high MASCOT scores were used for further MS BLAST analysis.

The MS BLAST strategy involved the use of BLAST search tools and the putative protein sequence containing the significant peptides identified by mass spectrometry. The sequence string was carefully chosen for the MS BLAST approach. Usually, two or more peptides which were located closely enough to be a part of a single protein were submitted for BLASTp search. This was done by submission of the sequence string from the beginning of the first peptide until the end of the last peptide for a BLASTp homology search. The output of the BLASTp search was then further analysed to identify putative protein hits. The length of the query sequence was recorded for each search. Once the significant hits were identified, the number of search peptides (in the query sequence), the GenBank accession number of homologous sequences, names of the proteins, the percentage of sequence similarity and the position of the query sequence in the contig were recorded.

The gene ontology (GO) analysis provides valuable information to assign a putative function for any identified protein [8]. The three general principles of GO were molecular function, biological process and cellular component. As the gene product had one or more molecular functions and was used in one or more biological processes, it is likely to fall into subcategories for one or more of these broad ontology groupings. Using the GO databases (AmiGO, USA), the GO number was checked for any protein or its homolog in other species.

The FunCatDB (MIPS, Germany) is an annotation scheme for the functional description of proteins from different prokaryotes, unicellular eukaryotes, plants and animals. It consists of 28 main functional categories, including different functional categories such as cellular transport, metabolism, cellular communication/signal transduction, etc.

The Bioinformatic Harvester EMBL Heidelberg [9] is a protein database which collects and displays bioinformatic data and predictions for human proteins from various databases. The database collects text-based information from the a number of public databases and prediction servers which includes Uniprot, SOURCE, Genome Browser, BLAST, SMART, SOSUI, PSORT II, CDART, MapView, NCBI-BLAST, SOSUI, STRING, Genome Browser and EMBL. Once the data are downloaded and saved, it is subsequently presented as text or inframe, depending on the data presentation of the original server. Therefore, it provides similar result as in the original database. For this experiment, the gene ontology number and related information of any significant entries (from BLAST searching) were derived from the reference proteome published in this database.

The gene ontology number and related information for each individual entry (found after MS BLAST searcing) were derived from the human reference proteome published in the Bioinformatic Harvester EMBL Heidelberg database.

The MASCOT search of LC-MS/MS data (while searched against the non redundant NCBI database) from all 20 samples in 1D-SDS-PAGE gel bands (Fig. 2) revealed 33 hits of Cryptosporidium. To obtain further information from the same MS data, the MS BLAST strategy was applied for a sequence similarity based protein homology search. While the mass fragmentation data of each individual band from LC-MS/MS were searched against the locally downloaded Cryptosporidium ORF (open reading frame) dataset, a total of 196 significant hits against contigs were recorded from 20 searches. When the contig sequences were analysed, each contig was found to contain at least one significant peptide hit, while some contained as many as 20 significant peptides hits (data not shown). In many instances the identified peptides with significant scores were found closely situated in a long continuous contig. The predicted ORF sequences (with significant peptides within each sequence string) were then used for BLAST search (protein-protein BLAST) for homology based protein identification. A total of 165 Cryptosporidium proteins were identified by this MS-BLAST approach. However, those hits included both C. parvum (n = 84) and C. hominis (n = 81) entries. In nearly all cases, the C. hominis homologous proteins were found with the same query in MS BLAST and the peptides were almost identical to C. parvum. Incorporating the two protein lists from the 1D-SDS-PAGE experiment (derived by MASCOT searching against the non redundant NCBI database and a MS BLAST search with peptides from Cryptosporidium ORF dataset) identified 100 C. parvum proteins (Table 1). Comparing the two approaches, the MS BLAST search strategy was found to provide 5 times greater (33 to 165) information than the NCBI search alone.

Many hypothetical proteins (n = 37) were identified by bioinformatic analysis of 1D-SDS-PAGE experimental data and the high MASCOT score along with higher percent identity confirms their physical existence in the proteome. Again, a number of metabolic enzymes have been identified, which include protein disulphide isomerase (gi.32398654), glyceraldehyde-3-phosphate dehydrogenase (gi.46229140), phosphoenolpyruvate carboxylase (gi.46227248), phosphoglucomutase (gi.46227774), glucose methanol choline oxidoreductase, enolase (gi.46227284), fructose, 1,6 biphosphate aldolase (gi.46227620), pyruvate kinase (gi.46227634) and phosphoglycerate kinase (gi.46229859). Several membrane associated proteins (gi.32398735, gi.46228663, gi.46227005) and oocyst wall protein (gi.46226838) were also identified. Other groups of proteins include many ribosomal proteins (n = 24), heat shock proteins (gi.2894792, gi.17385076, gi.46229711), and several uncharacterised proteins with unknown functions.

The functional categorization of 84 identified C. parvum proteins from the Cryptosporidium ORF dataset were made according to MIPS functional catalogue database (Fig. 3). The protein hits were matched with the human protein database, with the GO number and relevant functions of the homologous protein hits recorded for further analysis. A third (33%) of the identified proteins constituted hypothetical proteins while another third (29%) were responsible for protein biosynthesis. A significant proportion (20%) of total hits were proteins involved in intermediate and energy metabolism, while other groups were involved in DNA maintenance (7%), protein/RNA transport (6%) and proteins responsible for cell polarity and structure (5%).

The MudPIT analysis of sporozoite protein revealed a total of 42 proteins of Cryptosporidium while searched against NCBI database (Table 1). However, the number of submitted Cryptosporidium entries (i.e. previously identified and characterized) in NCBI is limited which possibly limits the success of such analysis. Therefore, the MS BLAST strategy was applied for sequence similarity based protein homology searching from the peptide fragmentation data derived after MudPIT analysis. While the MudPIT data were searched against the locally downloaded Cryptosporidium ORF dataset, a total of 150 hits of significant contigs were recorded. As previously observed in 1D-SDS and MS BLAST, the number of significant peptides in each contig also ranged from 1 to 20. The Cryptosporidium ORF sequence strings containing those significant peptide(s) were then used as a query sequence for BLASTp homology searching. The protein-protein BLAST searching revealed a total of 259 proteins of Cryptosporidium sp. which included a wide range of proteins. However, they included proteins from both C. parvum (n = 133) and C. hominis (n = 126). As with the MS BLAST analysis following 1D-SDS-PAGE, the homologous proteins of C. hominis were found in the same query for MS BLAST where the peptides were identical as in C. parvum. Notably, a similar level of redundancy was observed between C. hominis and C. parvum proteins. Incorporating the two protein lists from the MudPIT experimental data analyses provides a total of 140 proteins of C. parvum. Comparing the two approaches, the MS BLAST search strategy was found to be more informative in that it provided 6 times higher information than MASCOT search alone (42 to 259). A number of hypothetical proteins (n = 17) were identified by MS-BLAST search while many metabolic enzymes were recorded during the analysis. Some of the important enzymes are protein disulphide isomerase (gi. 32398654), enolase (gi.46227284), alcohol dehydrogenase (gi. 46228815), glycogen phosphorylase (gi.46229042), lactate dehydrogenase (gi.46229853), glyceraldehyde-3-phosphate dehydrogenase (gi.46229140), fructose,1,6 biphosphate aldolase (gi.46227620), pyruvate kinase (gi.46227634), NADP+ oxidoreductase (gi.13897519), phosphoglycerate kinase (gi.46229859). Several oocyst wall proteins and mucin like surface glycoproteins were also revealed from the study. In addition, a quarter of the total proteins (n = 35) identified from MudPIT analysis consists of 40S and 60S ribosomal proteins.