We’re all familiar with ISH (in-situ histochemistry) and FISH (fluorescent in-situ hybridization). Aside from these being common tools – perhaps even better described as essential – for the anatomic pathologist, they can provide visually stunning images. It would be hard to imagine a challenger for title of “most common journal or book cover image type” other than multicolor confocal microscopy images. The beauty of these is purely a bonus on top of the utility in both diagnostic and research applications to visualize spatially resolved cellular components. In this edition of The Primer, we’re going to look at some emerging new methods for generating this type of data.
These types of images can be generated by localization of particular antigens, such as proteins or complex sugars, representing the end products of expression of genetic pathways, or they can directly target nucleic acid elements. Since in most normal cells of an organism the DNA content is identical, targeting DNA is generally going to lead to some pretty boring, uninformative, monochrome images.
Exceptions to this are, of course, where explicit DNA variations are being probed, such as known chromosomal translocations (a common target of FISH) or to localize cellular or even subcellular occurrence of a pathogen. What’s more likely to be of variation and interest, however, is RNA, where expression is going to be different from cell to cell, based not only on the cell type but also a complex set of cell signalling pathways that regulate gene expression (RNA transcription). Let’s consider each of these approaches – antigen- and nucleic acid-based – in turn, highlighting for each what some of the limitations of common current methods are, and what some new techniques might be to address these shortcomings.
Antigen-based approaches – traditional methods
Antigen-based methods most commonly rely on using specific antibodies to tag the target(s) of interest. The chemistry part of ISH comes about when these antibodies are linked (usually indirectly, via a secondary antibody, for reasons of reagent supply-chain simplicity) to an enzyme, which, in turn, catalyzes the localized deposition of a stain at the site(s) of antibody binding. Stains are generally monochrome in nature, meaning a single target can be visualized per image.
Alternatively, the secondary antibody can be fluorescently labeled, which opens this up a bit by allowing for the possibility of simultaneously and separately distinguishing up to about four different targets per image. More than that generally isn’t feasible just due to the physics of fluorophores, which tend to have fairly broad excitation and emission spectra. Maintaining differentiable signals requires spacing these – at least the emissions side – out across the visible spectrum, and then using dedicated filter sets to observe them independently before recombining into that image destined for the next journal cover. This places practical limitations on the number of colors (targets) per image.
Emerging antigen-based approaches
Imagine if there were a method to selectively image antigens (biological macromolecules) that neither required the availability of a suitable antibody, nor was effectively restricted in the number of different targets it could separately label within a single image. What sounds magical not only exists, and it’s not even new; mass spectrometry imaging (MSI) data has been showing up in conferences for decades.
Essentially, it works by preparing a thin section much as for ISH, then placing that tissue section on a movable stage within a mass spectrometer. A mass spectral data set is collected from one point on the section, then the stage is moved to allow another, spatially separated data set to be obtained. (There are two different methods of achieving this, microprobe or ion microscope based, but for purposes of our overview the net results are similar). Within each mass spec data set, characteristic ion masses are used to indicate the presence and relative abundance of targets of interest.
Very large numbers of targets can be separately classified in this manner, with each target artificially color-coded onto a spatial image where each data set is represented at its location within the sample. A further benefit of this approach is that the data can be retroactively queried for additional targets of interest – if it’s something with a characteristic M/Z (mass over charge) ratio, you can look in old data sets for it and its distribution.
At this point, the most obvious question is, “why isn’t this more widely used?” The answer almost certainly is because the instrumentation required is very expensive (particularly when contrasted against routine IHC). It can also suffer from poor spatial resolution in comparison to more traditional methods. Although complexities exist in that different MSI approaches have different resolution capacity, the one of most interest here for suitability to lipid and polypeptide targets, MALDI-(Matrix Assisted Laser Desorption Ionization) based MSI, has around 20 µm resolution.
It’s also rather slow in comparison to direct visual imaging of IHC and has larger data storage and processing requirements. It’s not inconceivable that future iterations of the technology could become less costly and become more widely adopted. For now, it’s an incredible technical solution available in a limited number of labs in search of a compelling use-case scenario.
Meanwhile, back in the nucleic acids…
On the nucleic acid side, we considered above that most applications would involve RNA targets. “Spatially resolved transcriptomics,” as it is sometimes referred to, has been done in a large number of different ways, which range in the number of targets that can be analyzed and the physical size of the region tested; in general, there is a trade-off between these two (for an interesting graphic presentation of this, readers are directed to reference 1 below).1
One generic path to molecular in situ analysis is through the use of laser capture microdissection (LCM) to visually select and isolate single cells or groups of cells from within a tissue section for nucleic acid extraction. Once this is done, any and all molecular tests can be run on this template, up to and including whole genome sequencing (or more likely, whole exome/transcriptome sequencing). While potentially providing great depth of information, this approach doesn’t lead to nice graphic images as it’s destructive in nature. It also requires some reason or rationale to select what portions of a tissue section are captured for further analysis; likely requiring some prior, more traditional staining method such as IHC already.
At the other extreme of the spatial area versus depth of analysis lie techniques such as in-situ PCR (usually for a single target) or, more commonly now, in-situ hybridization (ISH). Both of these work by nucleic acid base pairing to complimentary sequences (by labeled primers for PCR, and by labeled probe for ISH). The PCR-based approach then requires the addition of enzyme and optimized thermocycling, in return for which high sensitivity is achievable, while ISH trades less sensitivity for simplicity.
Both of these share the visualization modes (and thus inherent challenges) of IHC described above: enzymatic dye deposition is monochrome, and fluorescent methods are usually just a few colors per section. Making in-situ PCR multiplex involves further significant technical challenges, as all targets should amplify under a shared set of thermocycling conditions – a requirement not always easily met. Overall, these methods can provide good spatial resolution (and quantitative data in the case of ISH) for a limited number of targets per tissue section.
Emerging nucleic acid methods
On this side of our topic, some of the emerging approaches involve substituting enzyme-based signal amplification (PCR or tyramide signal amplification) for direct hybridization-based amplification. One variant of this approach is known as Hybridization Chain Reaction. In this method, target-specific probes share regions of complementarity to a complimentary pair of labeled, partially palindromic oligonucleotides (let’s call them A and A’). These normally form into self-annealed hairpin structures, but in an isothermal process not requiring any enzymes, natural transient unfolding of one of these (A) allows it to hybridize to a tail portion of the target probe in a manner that blocks refolding while localizing label to probe site.
The exposed unfolded section of (A) then attracts hybridization by its pairing partner (A’), adding more label and driving unfolding and hybridization of another copy of (A), and so on.2 Publications using this method have shown good sensitivity and specificity with low background signal levels, while allowing for multiplexing within the constraints of fluorescence reporting already noted.
While this space generally sticks to discussions of methods and technologies without reference to specific products or vendors, the above method leads inescapably to mention of not just one, but two specific products. The first, known as RNAscope,3 also employs hybridization amplification, albeit by a different strategy (target-specific nucleic acid probes are bound through tag sequence matches to preamplifier probes, which in turn hybridize multiple amplifier probes and labels, essentially a form of what’s called branched chain DNA amplification).
In its fluorescent label incarnation, this method has seen increasingly wide use at least within research publications. An interesting twist on this comes when the fluorescent label is exchanged for a form of more complex, machine-scorable “barcodes.” Unlike fluors with limited multiplexing due to spectral overlap, large numbers of these labels remain uniquely differentiable with proper instrumentation. One such platform is from nanoString and pairing of their labeling and imaging approach with RNAscope for the target acquisition is the basis for a product they call the GeoMX Digital Spatial Profiler. Product literature indicates greater than 1,500 RNA target probes are available on this platform, which can simultaneously provide good spatial imaging resolution combined with very high target multiplicities.
Conclusions
Decades of training, experience, and familiarity with traditional IHC (including, critically, a good understanding of its own particular processing artifacts) is not going to be going away any time soon – it’s too useful, too cost effective, and too entrenched. An additional hurdle is that application of methods such as those outlined here, particularly on the nucleic acid side, are generally best performed on tissue prepared with “molecular friendly” fixative strategies as opposed to traditional formalin fixation (note use of “best,” particularly for short target sequences, hybridization-based approaches, and controlled formalin fixation times; traditional formalin fixed paraffin embedded (FFPE) tissue is amenable to molecular analysis, albeit less than ideal).
The flexibility and multiplexing capacity of these emerging methods on both the antigen and nucleic acid sides are, however, exciting. The images under analysis in the anatomic pathology lab in the not too distant future may well start to be even more dramatic, multicolored, and rich in data than those images gracing today’s covers.
References
- Liao J, Lu X, Shao X, Zhu L, Fan X. Uncovering an organ’s molecular architecture at single-cell resolution by spatially resolved transcriptomics [published online ahead of print, 2020 Jun 3]. Trends Biotechnol. 2020;S0167-7799(20)30140-2. doi:10.1016/j.tibtech.2020.05.006.
- Bi S, Yue S, Zhang S. Hybridization chain reaction: a versatile molecular tool for biosensing, bioimaging, and biomedicine. Chem Soc Rev. 2017;46(14):4281-4298. doi:10.1039/c7cs00055c.
- Wang F, Flanagan J, Su N, et al. RNAscope: a novel in situ RNA analysis platform for formalin-fixed, paraffin-embedded tissues. J Mol Diagn. 2012;14(1):22-29. doi:10.1016/j.jmoldx.2011.08.002.
