Letter to the Editor

Dear Editor,

The paper of Bagyalakshmi et al.¹ on rRNA gene region-based identification of molds contained a number of surprising results. That as many as 12% of ocular isolates were non-sporulating is the first point of interest, suggesting that these fungi appear more frequently than is generally appreciated. But the identification of Pythium insidiosum from patients who healed with medical treatment was truly surprising. Of course, such analyses for molds are open to interpretation; the degree of confidence for an identification depends on the quality of the sequencing and the accuracy of the phenotypic observations. Ideally, the database will contain useful sequence from the type strain or at least from strains obtained from a reference laboratory. As the authors deposited nucleotide sequences from their isolates into GenBank, we decided to repeat the searches. The results differed from some of those of the authors.

First, we have several queries and comments. If the primers were designed to “amplify only fungal DNA” (p. 140) and shown to not amplify human and Acanthamoeba DNA, why should they amplify genome from the Stramenopile P. insidiosum? “Fungi with 99% homology were identified to species level, and the genus level was assigned when there was homology of 95 to 98%” (p. 141). But homology with what, and at what coverage? “The multalin analysis performed on P. insidiosum isolates with GenBank accession number EF016789 …” (sic, p. 142). EF016789 is a sequence from a field mouse.

We performed sequence similarity searches on 5/8/08 using the authors' accession numbers, the nr/nt database, the Megablast program, filtering for low-complexity regions and fungal species-specific repeats, and masking for the lookup table. These parameters are appropriate when making species identifications based on conserved regions such as rRNA genes and their spacers. Both 1/-2 and 1/-3 match/mismatch scores were tested; the results were similar except in one instance noted below. The results for scoring at 1/-3 are given.

We accept the following identifications:

1. EF446286, Aspergillus fumigatus. 100% coverage and identity with sequences from many isolates from a range of laboratories. Only 36% coverage with sequence from the type strain ATCC1022.

2. EF446290, Fusarium proliferatum. 100% coverage and 99–100% identity with many isolates of F. proliferatum or its teleomorph.

3. EF446281, Botryosphaeria rhodina. ≤ 95% coverage and 99% identity with other isolates of B. rhodina or its anamorph.

It is not clear how the authors made the following identifications:

1. EF446287, Thielavia tortuosa. The only identity above 94% was 96% to an uncultured ascomycete. There was only 50% coverage and 93% identity with the one other T. tortuosa in GenBank.

2. EF446291, Botryosphaeria dothidea. There were no significant alignments.

We consider the following identifications uncertain:

1. EF446289, Glomerella cingulata. 92–94% coverage and 97–98% identity with isolates of G. cingulata or its anamorph from many laboratories—probably an acceptable genus identification.

2. EF446283, Aspergillus terreus. 100% identity with many isolates, including type strain ATCC1012, but no better than 71% coverage. For 1/-2 match/ mismatch scoring, 98% identity at 75% coverage with ATCC1012.

3. EF446284, Bipolaris species. 95–99% identity with many Bipolaris isolates, but no better than 67% coverage.

4. EF446282, Rhizoctonia bataticola. 100% identity but only 81% coverage with one R. bataticola. Recently, sequences from isolates of Lasioplodia theobromae (CBS collection) and its teleomorph B. rhodina have been deposited that show 100% identity at 79–81% coverage with EF466282. Thus, EF446282 could be from another B. rhodina.

5. EF446288, Macrophomina phaseolina. Only 82% coverage and 98% identity at best. M. phaseolina is an anamorph of R. bataticola, but EF446288 shared only 92% identity at 57% coverage with EF446282 suggesting one or both of these identifications are incorrect.

6. EF446285, P. insidiosum. 90% coverage and 97% identity at best with other isolates designated as P. insidiosum, 80% coverage and 92% identity at best with a sequence from type strain ATCC58643. It is unusual that there was no mention of the patients being exposed to fresh water and that six of the nine infections healed with antifungal treatment. For the 22 cases of Pythium keratitis in the literature in which the outcome is known,², ³, ⁴, ⁵ 16 eyes were lost and 6 eyes were salvaged by a corneal graft. It is strange that the success reported here passed without discussion. Indeed, the challenges the rest of us have faced when dealing with Pythium keratitis are well described in lines 3–9 on p. 146, a passage taken (unreferenced) from another paper.⁵

One needs to be judicious when identifying molds from sequences in GenBank. Unlike for many bacteria, biochemical data are usually lacking to support the identification. Good identity with high-quality sequence from isolates from expert laboratories is essential.

Sincerely,

P. Badenoch, B. Wetherall,

M. Woolley and D. Coster

Department of Ophthalmology, Flinders Medical Centre,

Bedford Park, Australia