With MALDI-TOF-based microbiology assays making significant headway in terms of clinical adoption, researchers and vendors have begun turning their sights on new applications for the technology.
Among these new applications is mold identification, where MALDI mass spec could offer significant improvements in both accuracy and turnaround time compared to conventional methods.
Last year, both Bruker's MALDI Biotyper and BioMérieux's Vitek MS systems – the two market leading mass spec-based microbiology platforms – received US Food and Drug Administration 510(k) clearance, making them available for US clinical use.
The platforms identify microbes by matching the protein profiles of sample organisms generated via MALDI mass spec to profiles contained in a proprietary database. Compared to traditional biochemical methods of microbe detection, MALDI-based systems can offer significant improvements in speed, price, and accuracy.
Thus far, MALDI approaches have focused primarily on identification of bacteria and yeast. BioMérieux's 510(k) clearance, for instance, covered databases for Gram-negative bacteria, Gram-positive bacteria, yeast, and anaerobes. Bruker's 510(k) clearance, meanwhile, covered a Gram-negative bacteria database, and the company has since launched clinical trials for databases covering Gram-positive bacteria and yeast that it plans to submit to FDA this year.
Recently, though, the field has begun to explore the potential of MALDI for mold identification. Mold infections are considerably rarer than bacterial infections; however, they have extremely high mortality rates, particularly in immunosuppressed individuals like cancer patients or transplant recipients. Additionally, the paucity of good tools for identification of these organisms suggests that MALDI could perhaps prove even more useful for mold ID than it has for bacterial work.
Mold identification "relies heavily on morphological identification," Susan Butler-Wu, associate director of the clinical microbiology laboratory at the University of Washington, told ProteoMonitor. "When growing in culture you examine how they look on the plate, how they are growing, what the structural characteristics are, etc."
This approach is prone to error, however, due to the structural similarities among many varieties of mold, she said.
Indeed, Nathan Ledeboer, medical director for the clinical microbiology and molecular diagnostics laboratories at Milwaukee's Froedtert Hospital, estimated that, depending on the experience of the technician making the identification, conventional mold ID methods range in accuracy from around 50 percent to 90 percent.
Initial studies using MALDI-TOF for mold identification, on the other hand, report accuracy in the 90 percent range, Ledeboer told ProteoMonitor. And, he said, this number should go up as more organisms are added to the databases used for these IDs.
"Using our microscopic and morphological identifications of these organisms, we are calling a lot of things incorrectly as a field," he said. "What we are finding as we sequence more of these isolates and learn more about these molds, is that organisms that look very similar are, in fact, very different organisms. They have very different outcome profiles, susceptibility profiles. And one of the things that I think mass spec is going to allow us to do is to be much more consistent on what we call."
The main reason MALDI has been slower to move into mold ID than bacterial work is the relatively small number of mold cases clinicians deal with, Ledeboer said, noting that his lab does "probably 100 to one bacterial IDs to mold IDs."
Another factor that he said has likely contributed to the lag is the relative difficulty of extracting proteins from the organism for identification.
"With the molds, using just the traditional formic acid extraction technique [that is used for bacteria], you won't be able to get the proteins out of the organism to analyze them," he said. "So you have to do some sort of harsh technique to really break it open."
Additionally, Ledeboer noted, as a eukaryote, molds exhibit significant variation in protein expression depending on where they are in their growth cycle. "If you look at a mold colony, the growth at the center of the colony versus the growth at the edge will have a very different protein profile."
This means that researchers must standardize their extraction somehow so that they are testing specimens at the same place in their growth cycle. One way to do this is to culture the mold in a liquid broth, a technique that has been used by Bruker in the development of the mold database for its Biotyper instrument.
One disadvantage of this method, however, is turnaround time, Ledeboer said, noting that growing the mold in broth can add another 24 to 48 hours to the workflow.
An alternative method is direct extraction from solid media, in which researchers take samples from the fast growing outer edge of the colony. This is the method used by National Institutes of Health researcher Anna Lau, who, with her colleagues, has built one of the more extensive MALDI-TOF mold databases generated to date, consisting of 220 species, 19 genera, and 410 strains.
According to Lau, to her knowledge, the NIH runs the only lab in the US that uses MALDI mass spec for routine identification of mold, and, she said, they are continuously adding new isolates to their database. The database is freely available to academic centers around the world, and Lau and her team have shared it with 21 labs thus far.
Lau told ProteoMonitor that the researchers have considered taking the database through FDA so that it could be used by clinical labs, but, she said, "at the moment we're happy just to get it out there and share it" on an RUO basis. She said the NIH database is an RUO supplement to Bruker's database, which contains 366 strains from 129 species, and 45 genera. The Bruker mold database is also RUO currently.
In a presentation this month at the Mass Spectrometry Applications to the Clinical Laboratory annual meeting in San Diego, Lau reported on a study she and her team had conducted comparing the effectiveness of databases generated using the solid media and broth-based extraction techniques.
Her findings indicated that the NIH database outperformed Bruker's database and that the best results were obtained when researchers used the same extraction method for identifying their isolates as was used to generate the databases they were searching against. For instance, if researchers used direct extraction from solid media, they got the best results searching against a database – like the NIH database – that was built using direct extraction, and vice versa.
Assuming one extraction method becomes dominant in the field, Lau's findings could have implications for platform adoption. For instance, if direct extraction becomes the preferred method – an occurrence Ledeboer said he considered most likely given its advantage in turnaround time – that could impact the utility of Bruker's broth-based database. Like NIH, BioMérieux, Bruker's main commercial competitor in this space, is pursuing the direct extraction route.
Lau did note, though, that while Bruker's database doesn't work well with direct extraction for some organisms, it does perform well on some of the more common molds like Aspergillus fumigatus.
In an email to ProteoMonitor, Markus Kostrzewa, Bruker's vice president, clinical mass spectrometry R&D, echoed this comment, noting that while "generally it is preferable to use the method which has been used for creation of the database when identification is performed ... this does not mean that generally no identification is possible with solid medium cultivation against the Bruker database."
A common workflow being used by labs using the Bruker database, he said, is to try direct extraction first to see if a good match can be made that way, followed by culturing in broth if this initial attempt doesn't provide a good result.
"As no misidentifications because of using one method with the other database [have been] reported, nothing argues against combining both approaches," he said.