A commonly encountered situation in diagnostics is the desire to run multiple tests on a specimen. In the case of infectious disease diagnostics, this occurs particularly when there are several possible etiological agents with similar presentation, but with different best treatment methods (and/or infection control considerations), such as in acute respiratory infections. A similar desire for multiple tests can arise in genetic disease testing. (Consider, for instance, cystic fibrosis, where several point mutations in the CFTR gene should be tested for on a single specimen.) A similar example from oncology treatment-related testing would be examining the KRAS gene for any of several activating mutations; certain treatment strategies assume normal signalling function of this molecule and are rendered ineffective by these mutations.
In these cases one could apply methods such as those covered in this series to date to examine multiple portions of a sample separately, one at a time, for each molecular marker of interest; however, the use of such “simplex” (single test per reaction) approaches in these cases is inefficient in terms of reagents, instrument space, and technologist time. Furthermore, it may often occur that a specimen is of limited volume; there just isn’t enough of it to go around for all the different tests the laboratory wishes to run on it.
In molecular diagnostics, the answer to this is to combine multiple independent tests in a single reaction. This is known as “multiplexing.”
For multiplexing to be feasible, the tests must share an underlying similar methodology so that the reagents in the reaction support all the different tests going on at the same time. PCR (and its RNA-focused sibling, RT-PCR) is well suited for this, as the only target- (assay) specific component in a reaction is the primers. Imagine, for instance, a PCR designed to amplify target A, and the reaction set up to do so; then consider that in order to also be able to amplify a different target B, one would only need add in the target B specific primers. In an ideal situation, either targets A or B (or both) will lead to production of specific amplicon(s), yielding information on the presence or absence of their respective targets. The process has shared one reaction tube, one master mix, one reaction setup, and one portion of sample. The benefits of this over multiple simplex reactions are obvious, and not surprisingly, multiplex molecular diagnostic methods are increasing in popularity.
Challenges to successful multiplexing
As the saying goes, however, “there is no free lunch”; the situation will not likely be ideal. Multiplex PCR methods bring with them their own set of complexities. First, successful multiplexing will require that each set of target-specific primers has very similar Tm (annealing temperatures); since the reactions are all occurring in a single tube, it follows that they are sharing a single thermal cycling profile. If the annealing temperature step in this cycle is above the Tm of a primer set, the sets won’t anneal well (or at all), leading to little or no product production when template is present. If the cycle annealing temperature is below the Tm of a primer set, then they may start to anneal in places of less than perfect homology, leading to spurious false-positive amplicon production. Generally, this problem can be dealt with in the assay design stages, by lengthening or shortening primer sequences as needed to get close Tm agreement among all of them.
A second issue may arise if there is any potential for primers to anneal to one another, again leading to artifactual “primer dimer” products. As a multiplex reaction contains more and more primer sets, the chances for such interactions grow exponentially and the assay design work of considering all primer pairs in silico for such interactions, and designing around them, becomes increasingly challenging. Methodical approaches and specialized multiplex PCR design software can help in overcoming this.
A third difficulty relates to thermodynamic behavior of different amplicons, that of their amplification efficiencies. By necessity, the thermal cycling profile will be one which has to work for all targets, and so may be something of a compromise—less than fully optimized for any particular assay target. Use of varying primer concentrations (biased towards more of those primers for the harder-to-amplify targets) across the sets can help with this, and must often be done to ensure that a better amplifying target does not effectively monopolize a reaction, leading to a lack of resources (polymerase, dNTPs, etc.) for its less robust tube mates—a process known as “squelching.” Again, a methodical approach to equalizing yield across reactions to be combined in a multiplex can generally yield good results, but may be a time-consuming process, and more so with increasing numbers of targets for an assay.
A fourth difficulty—by no means the only remaining one for multiplexing, but the one we shall consider for the remainder of this month’s installment—is that of separately detecting each of the possible amplicons arising from a multiplex PCR reaction.
Meeting the challenges with real-time PCR
Since our earlier installments of “The Primer” dealt with real-time PCR, let’s consider multiplexing in that format first. In the case of DNA binding dye-based real-time PCR, multiplexing is in theory possible if the target amplicons have sufficiently different melt peaks, so that a final melt curve step can show the presence or absence of characteristic products. The author has seen this successfully employed in a two-target “duplex” reaction case; however, this approach is rarely practical. One reason is that the difference in amplicon melt points relates to differences in length and/or GC content, which in turn generally mean appreciably different amplification kinetics and a tendency toward the third class of problems described earlier. Another is that the presence of multiple peaks in a melt curve makes it increasingly difficult to discern appropriate product formation from spurious side reactions.
FRET-probe based methods fare somewhat better, in all their forms as covered last month. Here, in order to multiplex one merely needs to employ reporter dyes for each of the target-specific probes such that each dye has a spectrally discernible emission wavelength, and then use appropriate filters to separately detect signals from each probe in a reaction. Unfortunately, most fluorophores have relatively broad emission spectra and, despite using advanced signal processing techniques such as deconvolution (beyond the scope of discussion here), no currently available real-time PCR instruments are capable of more than six “channels” of detection. One channel is frequently reserved for use for an internal control signal (more on controls in an upcoming installment), leaving at most five possible targets to be multiplexed. In practice, few assays with more than three multiplexed targets plus an internal control are seen with probe-based real-time PCR systems, but the assays generally perform very well and are a highly practical way of multiplexing small numbers of tests.
What about historic, gel-based methods? From a multiplexing standpoint, these work quite well, as the only requirement is that the amplicons be of distinguishable sizes. Multiplex PCRs of 10 or more simultaneous targets with agarose gel resolution exist in the literature; however, the previously described practical problems of agarose gel methods (contamination risk, low throughput) relegate these primarily to research use and not clinical applications. More advanced automated methods of amplicon size separation in multiplex settings, such as use of capillary sequencers (another upcoming topic), is possible and is used in some cases, but still lacks convenience and speed and is primarily a niche application for functions such as short tandem repeat (STR) genotyping for forensic applications.
A method which is very well suited to the endpoint detection and classification of PCR products is array detection. Array-based methods are highly varied, and space constraints here will limit us to a brief description of their form and application in this context alone. A traditional array is a two-dimensional grid of nucleic acid spots on a solid surface (usually chemically derivatized glass). Each spot can be unique, usually an oligonucleotide complementary to and thus capable of hybridizing to a sequence internal to a potential amplicon in a test. Individual array spots are identified by physical position; thus every possible product can be labeled with the same fluorophore, and an array reader optimized for this. In use, one of the PCR primers for each target is synthesized with the reporting fluorophore. Post-PCR, the reaction mixture is washed over the array surface and complementary sequences are allowed to hybridize and then gently washed to remove non-specific binding. Where PCR products in the reaction anneal to an array spot, the primer-conjugated (and thus product-incorporated) fluorophore is now localized, and optical scanning of the array will reveal which spot(s) fluoresce. These are then indexed back against their physical position, and the identity of the product(s) produced in the reaction is determined.
The advantages of this method are that it is amenable to automation and extremely high levels of multiplexing; arrays the size of a postage stamp can contain hundreds of thousands of unique spots, or even more. These numbers greatly exceed practical limits of PCR multiplexing from squelching and primer interaction concerns, keeping product detection and classification from being the limiting factor in these techniques.
A variation on this is the use of liquid phase, bead based arrays. In these methods, rather than a solid surface for printing all of the complimentary “probe” sequences on, each unique type of probe is coupled in large numbers to a specific optically tagged “microbead.” This optical tagging can be in the form of a color code, a tiny black and white barcode, or something similar. Large numbers of each bead type, bearing a single type of probe sequence, can be “mixed and matched” together in suspension and mixed with the PCR reaction products, labeled as in the case for solid phase arrays. Again, hybridization between PCR products and their matching probes occurs if such products are present, although with the advantage of occurring with 3D liquid phase kinetics; faster hybridizations are possible than with solid phase arrays. Detection and classification of which beads hybridized to their matching PCR product(s) is done optically.
One approach to this is an adaptation of flow cytometry methods, whereby the bead mix is sampled through a very small glass capillary, forcing the beads to flow through single-file; as they pass a detector, the bead “identity code,” such as color, is read out, defining the bead type being read and thus its coupled detector. Simultaneously the bead is analyzed for fluorescence, indicating bound PCR product. Instruments of this type read and classify hundreds of beads per second, making for fast and accurate classification of results. Another approach allows the beads to disperse over the surface of a small well plate, and then the surface is imaged by a CCD camera through a microscope lens. Each bead in the image is identified (color or barcode), again followed by an overlapping fluorescence image, allowing for identification of which product(s) were amplified.
Simplex test methods will remain in use for some situations, but expect to see more and more multiplexing services in molecular testing as better platforms and test panels are developed and come into use.
John Brunstein, PhD, a member of the MLO Editorial Advisory Board, is President and CSO of British Columbia-based PathoID, Inc.