All free-living amoebae (FLA) isolates were a gift from Dr. G.S. Visvesvara (DPD, CDC). All chemicals used during this study were purchased from Sigma-Aldrich (St. Louis, MO) except where indicated. Buffers and culture media were obtained from the Scientific Resources Program at the Centers for Disease Control & Prevention (CDC).

We included in this study a total of 24 Naegleria isolates, comprising seven species, including 18 N. fowleri genotypes I, II, and III from different sources listed in Table 1. These isolates were previously identified by using conventional phenotypic tests and further characterized using the internal transcriber spacer and mitochondrial small subunit rRNA gene (Zhou et al. 2003). For comparison, we included an additional six species described within the genus Naegleria. All Naegleria spp. used in this study were grown in double-modified Nelson’s medium. Growth temperature for N. jadini and N. gruberi was 25 °C; the other Naegleria species were grown at 37 °C (Zhou et al. 2003). All isolates were washed with amoeba saline as described before (Zhou et al. 2003), pelleted and stored at −80 °C. To account for growth variability, we grew, harvested and analyzed separately each organism on different occasions over a 2-wk (batch 2) and 4-wk (batch 3) interval.

We prepared protein extracts for MS analysis as described previously (Moura et al. 2003, 2008; Shaw et al. 2004; Williamson et al. 2012) with a few modifications. Briefly, extracts were obtained from 100 μl frozen amoeba pellets containing 10⁸ trophozoites after several freeze-thaw cycles and extracted with a 10% formic acid-50% acetonitrile solution. Five microliters of each extract were aliquoted and then premixed with equal volumes of the respective matrix solutions just before spotting on the MALDI target. We followed the protocol described previously (Moura et al. 2008) employing MALDI matrices (saturated solutions of α-cyano-4-hydroxycinnamic acid, or 3,5-dimethoxy-4-hydroxycinnamic acid [sinapinic acid]), 192-well stainless steel sample target plates (Applied Biosystems [AB], Framingham, MA) and mass standards for calibration (Sequazyme Peptide Mass Standards Kit, AB) prior to MALDI-TOF MS analysis. Details of conditions and parameters for MALDI-TOF MS analysis using a MALDI-TOF/TOF mass spectrometer (AB 4700 Proteomics Analyzer) are those described before (Moura et al. 2008).

We processed mass spectra from three harvestings as described before with a few modifications (Moura et al. 2008; Satten et al. 2004). Briefly, we exported profile spectral data as text-format m/z-intensity lists. The text data were further processed and viewed by use of a suite of custom Microsoft Visual Basic.NET (VB.NET) programs, the VBA macros and VB.NET programs “MultiSpec Processor” and “MultiSpec Viewer”. We normalized these spectra to the base peak, smoothed and finally standardized and denoised by use of a custom Fortran program (Satten et al. 2004). We used PAST software v1.34 (http://folk.uio.no/ohammer/past/doc1.html) for hierarchical cluster analysis, with the single summed spectra (one spectrum representing each organism) for input. We used Random Forest (RF) v5.1 (http://www.stat.berkeley.edu/users/breiman/RandomForests/cc_home.htm) for classification and identification, in this case, with ~9 summed spectra from three separated harvestings (2 wk apart) of each organism as a training set and ~3 summed spectra as unknowns. The recompilation of the Fortran RF code for each experimental condition was automatically driven by VB.NET programs, and custom viewing applications were developed to aid in the interpretation of the RF results.

In addition, we imported the summed spectra into the BioNumerics software v7.1 (Applied Maths Inc., Sint-Martens-Latem, Belgium) that were preprocessed using default parameters. A similarity matrix was generated using Ranked Pearson Correlation and provided the basis for constructing Neighbor Joining trees. Next, we submitted the extracted peak lists from all studied Naegleria to an in-house script for tentative protein identification. We used a total of 31,833 Naegleria spp. proteins entries available at GenBank (http://www.ncbi.nlm.nih.gov/genbank/) as of January, 2014. The Pepstats software from the European Molecular Biology Open Software Suite (EMBOOS) (Olson 2002) was used to retrieve protein database information and to calculate statistics of protein properties. We subtracted the molecular mass of the N-terminal methionine (131.00 Da), as described by before (Moura et al. 2008) and subtracted the proton charge (1 Da) as well. We matched the observed peaks (m/z minus 1 Da) and the predicted protein MW proteins in the database using an in-house script considering ±3.00 Da and posttranslational modifications such as the oxidation of methionine (+16 Da). Finally, the list of possible proteins for each peak was manually curated in order to exclude partial or fragment proteins and hypothetical proteins.

Extracts from three batches of the FLA species and isolates described in Table 1 were analyzed by MALDI-TOF MS. Spectra obtained revealed complex spectral patterns with 20–40 peaks in the m/z range corresponding to 1,000–6,000 Da or 2,000–14,000 Da. The mass range corresponding to 2,000–14,000 Da was preferred for this study because the peaks in this region were more abundant and more consistent. Spectra of amoeba protein extracts revealed reproducible patterns and differences among species could be detected visually after analyzing several runs. As expected, only minor variation from batch to batch and from well to well was detected in the analysis. Careful analysis of the spectra in Fig. 1 reveals that several peaks desorbed from extracts of Naegleria species and isolates are markedly similar to each other. However, there are specific peaks that can be used to easily differentiate among the different species although isolate differentiation would require more effort.

Because spectral visual analysis is not an easy task, we preferred to employ a statistical analysis of the MALDI-TOF MS data to quickly reveal significant differences among the strains analyzed. Naegleria fowleri could be discriminated from each other by Random Forest using summed spectra from three harvestings of each organism as a training set and ~3 summed spectra as unknowns. With low estimated classification error (0.75%) the RF algorithm successfully classified all the Naegleria species and the majority of the isolates in the training set (Fig. S1). Two isolates (CDC:V019 and CDC:V551) shared several peaks and the RF algorithm was unable to differentiate this subset. Results with three clustering algorithms within PAST, such as paired group, single linkage, and Ward’s method, using similarity measure methods such as Euclidean, Dice, and Jaccard, among others, were analogous and quite consistent regardless of the analysis method applied, so that the different Naegleria species and isolates studied could be reliably and reproducibly separated from each other (Fig. S2). As expected, N. fowleri isolates were clustered together, and even sub-clusters could be detected. Interestingly, two isolates (CDC:V020 and CDC:V518) were clustered with other species within the genus Naegleria. Naegleria lovaniensis was clustered with N. fowleri isolates CDC:V551 while N. australiensis is the closest species to the N. fowleri isolates. Even more, isolates were mainly clustered according to the species in Neighbor Joining trees constructed using BioNumerics (Fig. 2). Naegleria fowleri isolates could be distinguished from related species, such as N. dunnebackei, N. gruberi, N. jadini, and N. italica. In this analysis, N. australiensis was closely related to a N. fowleri isolate recovered from water (CDC:V518) and N. lovaniensis was closely related to a N. fowleri isolate recovered from CSF (CDC:V551). Among N. fowleri strains, various subgroups were observed.

The peak lists compared with proteins of Naegleria spp. in the database revealed 51 distinct matches among all Naegleria, including peaks corresponding to 13 ribosomal proteins (Table 2). Considering the peaks observed among all 18 N. fowleri isolates studied plus the strain HB3, a mean of 37 proteins were identified. The isolate that generate the smaller number of peaks was CDC:V511 with only 27 proteins identified. A total of 42 of the 51 proteins identified were found in N. fowleri HB3. Nevertheless, none of the 19 N. fowleri studied presented all the 51 proteins identified, although 29 proteins were identified in 14 isolates.