It’s worth the weight: mass spectrometry for bacterial identification

Aug. 18, 2014

Readers of this series will by now be familiar with the pattern: every month, we cover some technique of interest and clinical applicability in molecular diagnostics. The key here is the word “molecular,” and as you know if you’ve followed the series from the start, “molecular” in this context means specifically “nucleic-acid based.” 

This month, we’re going to change that by delving into an emerging laboratory diagnostic approach which is not strictly “molecular.” It is, however, one of the most promising alternatives to molecular diagnostics in the applications we will consider here, and it very much deals in the analysis of molecules. This technique is mass spectrometry as applied in the identification of bacterial samples. 

Mass spectrometry (or “mass spec” as it’s often abbreviated) is far from a new technology. The first “modern” instruments known by this name date back to 1918. They were invented by the physicist Arthur Jeffrey Dempster (1886-1950), who was building on earlier research into cathode ray tubes and the deflections of ionized gasses by electric fields within these tubes. While modern instruments incorporate a large number of improvements on these earliest machines, the core principle remains the same. It is most simply stated as an observation that even tiny molecules or fragments of molecules follow the basic laws of Newtonian physics with regard to force, acceleration, and motion. You may recall from physics class a couple of key equations: F=ma (a relationship between Force, mass, and acceleration); and d=1/2 at2 (a relationship between distance, time, and acceleration). 

If physics and math make you queasy, have no fear, for we need go no further than to observe that since these have acceleration as a common term, they can be combined to create an equation relating F, m, d, and t; and that knowing any three of these, we can solve for the fourth. A mass spec is really just an instrument into which we put a sample, where it is subjected to a known F, a fixed d, and a measured t; the instrument computes a mass m for the input molecules. (Bear in mind that this is a rather gross simplification of the many and varied complexities of mass spec hardware and the multitude of ways in which this can be done; it’s a large and complex field in its own right with complex terminology. This is, however, an accurate summary, and it is sufficient for our needs.)

In practice, and on the type of mass spec normally used for the application we are considering, the process works roughly as follows: A specimen material (the test sample) is mixed with a material referred to as “matrix,” dried, and placed inside the mass spec instrument. Air is pumped out of the instrument, and a powerful laser is used to fire a brief light pulse at the mixed sample/matrix target. The energy in the light pulse is effectively absorbed by the matrix material and transferred to the sample, causing molecules in the sample material to break apart into smaller fragments and come free from the sample surface. Two things are important here. First, the fragments produced are not random, but are characteristic of the molecule they originate from. That is, every molecule has specific “weak bonds” that tend to break under this form of stress, meaning the same fragments regularly form. Second, during the molecular fragmentation, the fragments tend to pick up a +1 electric charge. That is, they become positively ionized. This particular type of mass spectrometry is known by the acronym MALDI, for “Matrix Assisted Laser Desorption Ionization.”

At this point, we have a bunch of characteristic molecular fragments bearing electrical charges floating in a vacuum just above the sample surface. Simplistically speaking and ignoring some technical improvements such as reflectrons, this sample surface is at one end of a tube. At the other end is a grating bearing a highly charged negative electrical potential. Here’s where our physics summarized above comes into play; the opposite electrical charges on the grating and molecule fragments put a known attractive force “F” on the fragment, accelerating it toward the grating. Small molecules, having little mass “m,” accelerate rapidly and fly quickly to the grating, while bigger, more massive molecules accelerate more slowly and take a longer time to make the trip. At the grating is an electrical detector which can measure the impacts of the charged fragments. 

Figure 1a

This type of mass spec measures the time between the laser pulse and the impacts of fragments on the detector, the “Time of Flight” (making this instrument, in the parlance, a “MALDI-TOF”). Now knowing the distance between target and detector “d,” the known “F” and time “t,” the instrument can report the mass “m” of each of the fragments detected (Figure 1a). The instrument output is a graph which shows masses of fragments detected, with a peak height for each fragment representing a relative measure of how many of those fragments were detected: large peaks mean a lot of fragments, and small peaks mean fewer fragments (Figure 1b). Note that the individual peaks in a mass spectrum are not some Gaussian curve shape as in most techniques, but exact mass value lines, and that the instrument resolution can be sufficient to differentiate even isotopic differences. This precise ability to measure even tiny mass differences lies at the basis of our application here. 

Figure 1b

Mass spec has been in common use in the clinical lab for various blood chemistry applications for a long time, but how do we go about applying this to bacterial identification? Well, it turns out that if we use a relatively uniform sample of bacteria such as a single colony from a microbiological media plate as a target, we get a very complex mass spectrum pattern out of our MALDI-TOF. This isn’t surprising, when you think of the very large number of different kinds of molecules that are present in the bacteria. In fact, out of a complex mass spectrum pattern like this, we generally don’t even know what molecular fragment relates to each peak. 

What is perhaps surprising, however, is that the overall pattern of peaks, in terms of masses and relative heights, forms a recognizable “signature” for a bacterial species. Even closely related species will generally have a few very small but detectable fragment mass differences somewhere within the spectrum. If we run large numbers of bacterial isolates in a mass spec system, a library of these characteristic spectrum can be formed, and a pattern can be matched by a computer to the mass spec output generated by our laboratory sample. A good match is strong evidence for a sample being the same organism as the library spectrum. 

Note that this pattern matching is flexible enough to allow for some variation such as that which arises from the type of growth medium and growth phase the bacteria was in, while still providing consistent identification calls. (Generally, environmental variables will impact the relative abundance of different bacterial cell molecules as the organism responds to its environment. This leads to differences in relative peak heights under different conditions, but the exact masses of each of the peaks remain consistent.) 

This approach to bacterial identification is cheap (not counting the cost of the instrument, at least; the consumables in form of matrix are a few cents a sample) and very fast (sample preparation including drying down takes on the order of a half hour or less, and the actual instrument run, pattern matching, and identification call is performed in seconds). Because it’s a pattern-matching process, new organisms or strains can be incorporated by getting reference samples and adding their patterns to the library. 

This flexibility can also be a downside, however, as organisms can change over time and with geographic distribution, possibly leading to misidentifications as a species’ characteristic pattern changes. Curation and updating of the reference pattern library is therefore important. Other weaknesses of the method are that it needs a relatively large input sample and it doesn’t handle mixed specimens very well. (As one might imagine, just overlapping two complex spectra of similar intensities results in a net pattern that doesn’t really match anything in the reference library.) 

Another shortcoming is that mass spec for bacterial identification doesn’t directly address the antibiotic susceptibility of the sample tested; one would still at present need to determine this by classical means. Note, however, that I say “at present”; there are methods being developed, and showing promise, to add this capability to mass spec bacterial ID systems. One approach is to very specifically identify one or more characteristic mass peaks arising from bacterial products involved in an antibiotic resistance mechanism, and then determine if these peaks are present in a sample spectrum. This approach has significant technical challenges, in that not all resistance factors lead to recognizable ion fragments in a whole bacteria MALDI-TOF experiment; and the ones that do can, of course, undergo genetic changes which may or may not impact the antibiotic resistance phenotype but can alter the appearance of the mass spectrum. A second approach is to take the bacterial colony of interest, expose samples of it to a series of antibiotics, and mass spec analyze the samples after exposure, looking for characteristic mass spec changes associated with growth inhibition. This method is essentially akin to classical E-test or Kirby-Bauer testing, as a direct phenotypic measure of organism response to antibiotic agents, but can be performed much faster than the classical approaches (a few hours, as opposed to overnight growth). 

This combination of speed, accuracy, ease of use, and broad spectrum of identification makes mass spectrometry a powerful emerging tool for bacterial identification in the clinical lab, and its appearance in this role is likely to become more widespread in the near future. 

A final note, in case some of you are puzzled that some molecule fragments might not have a single positive charge. This is true, and instruments can be configured to detect either positive or negative ions; our example here uses just the positive ions. Some of these may in fact be more than singly ionized, which in fact gives the molecule greater “F,” and thus greater acceleration and a shorter “t.” Strictly, mass spectrums report “m/Z,” mass divided by charge, which corrects for this; and in our pattern matching type application, this proportion of fragments with multiple charges just adds additional spectral peaks which form part of the overall pattern.

John Brunstein, PhD, a member of the MLO Editorial Advisory Board, is President and CSO of British Columbia-based PathoID, Inc.