For long-term storage, Cryptococcus isolates were kept at -70°C in cryobank stocks (Mast Diagnostica, Reinfeld, Germany). After thawing, strains were propagated on Sabouraud’s (SAB) agar slants supplemented with 0.5% peptone (casein), 0.5% peptone (meat), and 2% glucose. Before sample preparation, strains were cultivated on SAB agar overnight at 30°C.

For the purpose of text clarity, only the species nomenclature according to Hagen et al. (Hagen et al., 2015) is adopted from here. As references, thirteen strains of the CBS collection (Westerdijk Fungal Biodiversity Institute) were used: three C. neoformans (CBS 8710 (molecular type VNI), CBS 10084 (VNII), CBS 10085(VNI)), two C. deneoformans (CBS 6900 and CBS 10079 (VNIV)), two C. gattii (CBS 6289,and CBS 10078, VGI), two C. deuterogattii (CBS 10082, and CBS 10514, VGII) two C. bacillisporus (CBS 6955 and CBS 10081, VGIII), one C. tetragattii (CBS 10101, VGIV), and C. decagattii (CBS 11687, VGIV).

A test set of 153 isolates was assembled from previously characterized collections. This included all Thai strains from Worasilchai et al. (2017), augmented with rare species isolates from Kenya (Kangogo et al., 2015) and the Birmingham laboratory collection, which include strain from various studies [e.g. (Voelz et al., 2014)]. All isolates were typed either previously (Kangogo et al., 2015; Worasilchai et al., 2017) or specifically for the purpose of this study using the URA5-RFLP method. The final set contained n=96 C. neoformans, n=6 C. deneoformans, n=5 C. gattii, n=18 C. bacillisporus, n=20 C. deuterogattii, and n=8 C. tetragattii isolates. A negative control group was assembled from mass spectra randomly chosen from those obtained during bacterial (n=86) of fungal (n=403) routine diagnostics.

Restriction fragment length polymorphisms were performed as described previously (Kangogo et al., 2015; Worasilchai et al., 2017). Briefly, genomic DNA was extracted from cells using phenol/chloroform and the URA5 gene was amplified using URA5 forward (5-ATGTCCTCCCAAGCCCTCGACTCCG-3) and SJ01 reverse (5-TTAAGACCTCTGAACACCGTACTC-3) primers (Meyer et al., 2003). The amplicons obtained were either simultaneously digested with HhaI (20 U/μl) and Sau96I (10 U/μl) or StuI (10 U/µl) alone for 8 hours (all from New England Biolabs). The digestion products were purified using a PCR clean-up kit (NucleoSpin, Macherey-Nagel, Düren, Germany) and visualized on a 3% agarose gel.

For regular harvest and formic acid-extraction [preprocessing protocol (A) (Bader, 2017)], cells were taken from agar plates by scraping approximately a 1µl loop full of cells and re-suspending them in 300 µl water. 700 µl absolute ethanol was added to a final concentration of 70% (v/v) and vortexes. Cells were spun down at 8500xg for 5 min, the supernatant completely discarded and the cells lysed first with 50 µl 70% (v/v) formic acid, and 50 µl pure acetonitrile. Modifications to this protocol tested were for preprocessing protocol (B) that cells were collected in 300 µl 5% (v/v) DMSO_ad, for preprocessing protocol (C) that DMSO was included in the 70% ethanol washing step to a final volume of 5% (v/v), and for preprocessing protocol (D) that cells were collected in 300 µl water already including an equivalent of ~100 µl glass beads (0.5 mm diameter, Roth, Karlsruhe, Germany). Here, cells were mechanically disrupted in a FP120 fast prep machine (Bio101, Thermo Savant) at setting 4, for 30 sec during the formic acid step.

MSP references for the MALDI Biotyper were generated according to the manufacturer’s guidelines (Kostrzewa and Maier, 2017), using preprocessing protocol A. Spectra from 24 individual spots were gathered on a freshly calibrated (BTS reference standard) Autoflex III system (Bruker Daltonics, Bremen, Germany) using the automated acquisition mode of the Biotyper 3.1. Spectra were processed using the inbuilt MSP generation method, using the standard parameters.

The literature reports that both, removal of cryptococcal capsule can (Thomaz et al., 2016) or does not (Hagen et al., 2015) positively influence spectrum quality. Since the capsule material is soluble in DMSO, we devised pre-processing protocols that would deplete the capsule prior to the regular formic acid/acetonitrile extraction protocol. Both pre-processing protocols, B ( Figures 1A, B ) and C (not shown), efficiently removed capsules in all strains. However, subsequent measurement of mass spectra did not reveal any additional mass signals, or major differences in spectrum quality ( Figure 1C ).

Next, we tested if mechanical disruption of the cells yielded more informative spectra using mechanical disruption (preprocessing protocol D). Indeed, mass spectra recorded from mechanically disrupted cells resulted in more evenly distributed peak intensities across the major mass signals. However, no additional mass signals of high intensity were found ( Figure 1D ).

In our hands removal of the capsule did not result in spectra with higher information content, at any time. Mechanical disruption did reveal some additional masses, but in favor of the lower hands-on-time the original pre-processing protocol A was subsequently used for MSP creation and testing.

Next, we created MSPs for 13 reference strains encompassing seven molecular types of the C. neoformans/gattii complexes (Meyer et al., 2003; Hagen et al., 2015), using the standard extraction procedure (pre-processing protocol A). Cluster analysis of the MSPs generated suggested sufficient distance to clearly distinguish between C. neoformans complex molecular types VNIV (C. deneoformans) and VNI/II, and possibly also between VNI and VNII themselves, but less so among molecular types within the C. gattii complex ( Figure 2 ).

Mass spectra for all test isolates were obtained using preprocessing protocol A. Were MALDI-TOF results using the new MSP set deviated from previous data, URA5-RFLP typing was repeated as the gold standard ( Figure 3A ). All but two deviations could be resolved (see below). To discriminate between C. tetragattii and potential C. decagattii strains, we sequenced the URA5-amplicon obtained from CBS 11687 (C. decagattii, deposited at Genebank under the accession number MH605184) and compared it to the respective sequence of CBS 10101 (C. tetragattii, gene bank accession AY973155). Restriction with StuI was found, and experimentally confirmed, to discriminate the two species ( Figure 3B ). However, there were no further C. decagattii isolates among our strains. C. decagattii remains a rare species, and only a single isolate of this molecular type (CBS 11687) was available for this study, which was already included in the reference set. Therefore, the final test collection encompassed only six of the seven species used for generation of references.

From the test collection, we were able to correctly identify 143/153 (93.5%) of the isolates on species-level using duplicate spots, with the top log score ≥ 2.000 ( Figure 3C ), as recommended by the manufacturer. Of the remaining ten isolates, eight still gave correct species matches at scores between 1.700 and 1.999, considered only genus-level by the manufacturer. Among the negative control set, there were no results higher than a log score of 1.300, indicating no false positives are to be expected under routine diagnostic conditions ( Figure 3C ). Inconsistent identifications were only observed for two C. tetragattii isolates where repetitively top matches of different spots of the same preparation were C. tetragattii, C. gattii, or C. deuterogattii, all at values above 1.999.

Because of this, and the close relations found during cluster analysis ( Figure 2 ), we also inspected the log score difference from the correct to the highest scoring false match for each spot ( Figure 3D ) for those tests where a second species matched above the significance threshold. Only 3% of all tested spots (14 out of 428) matched a second MSP with a log score >1.999. As expected from the cluster analysis, these “best false” second matches were found only among species in the C. gattii complex. This was the case for three C. bacillisporus isolates giving a second best match with C. decagattii, with a log score difference between 0.1 to 0.4. In addition to the two inconsistent C. tetragattii isolates discussed above, one additional C. tetragattii isolate also gave a second best match with C. decagattii. The score values for both matches were near 2.000. The close relationships of the different species will likely also have implications on properly identifying hybrid isolates.