Beyond Gating

In the last 40 years, flow cytometry has evolved from a cumbersome but promising technology to a routine and indispensable diagnostic tool, descended but different from the original fluorescence activated cell sorting technology pioneered by Herzenberg and others1 in the 1960s. Despite its transformation in recent years, the potential of clinical flow cytometry for the immunophenotyping of leukemia and lymphoma is still not fully exploited in many clinical laboratories. Early operative challenges mandated by limitations in hardware design, reagent composition, and computational power during the formative years of flow cytometry have been unnecessarily carried forward to an age in which these limitations no longer exist. In addition, many laboratories apply criteria appropriate for primarily quantitative assessments such as absolute T-helper counts to leukemia and lymphoma immunophenotyping, despite the substantially more qualitative nature of the latter application. Information provided by flow cytometers is still often interpreted as static dots on paper rather than as unique points in multidimensional space, easily interrogated and viewed in multiple combinations and from multiple vantage points by trained interpreters. In this issue of the Journal, Fromm and colleagues2 describe an innovative application of 9-color (11-parameter) flow cytometry to the diagnosis of classical Hodgkin lymphoma (CHL). Their study is remarkable for 2 fundamental advances: First, it refutes the adage—dogmatically recited since the birth of this technology—that flow cytometry has no useful role in the diagnosis of CHL. Second, it requires a modern, iterative approach to flow cytometric immunophenotyping; the traditional sequence of restrictive gating and histogram interpretation in this setting would likely miss a substantial portion of the key diagnostic findings illustrated in their study. To understand the importance of the attitude embodied in the study by Fromm et al,2 it is useful to review some history. The amount of computer memory required to save every measured parameter from every cell in a given flow cytometric assay (so-called list-mode data) is modest by current standards. There was a time, however, when this amount of computer memory was simply too great for routine storage of all information gleaned in a single analysis. Acquired flow cytometric data were, therefore, manipulated via real-time gating strategies so that 2-dimensional histogram projections of the data could be saved and printed. These histograms existed as projected images, not actual data, and, therefore, consumed comparatively little computer memory. In the absence of routine list-mode analysis and color “eventing,” the only way to exploit the dimensionality of the flow cytometric data in those days was via restrictive, real-time gating strategies that excluded certain cells based on one or more measured parameters, followed by 1- or 2-dimensional projections of remaining measured parameters onto printed or saved histograms. In addition, the lower dimensionality of old flow cytometric data—1- or 2-color analysis was the norm back then—mandated that surrogate markers such as light scatter characteristics, consistent through every assay, were necessary gating parameters to ensure assessment of comparable cell populations through multiple analyses.3 Thus, the 1-way exclusionary process still known as gating was born of necessity from a less technologically advanced era. Unfortunately, these attitudes toward flow cytometric data analysis persist in the minds of many pathologists, despite advances that have made these approaches obsolete. Restrictive gating is now unnecessary. (I even go so far as to discourage house officers from using the word gating because it bears such little resemblance today to the processes of years past that gave rise to the term.) With modern cytometers and ordinary personal computers, a single routine assay produces the equivalent of thousands of unique points embedded in high-dimensional space (11-dimensional space in the case of the 9-color flow cytometry used by Fromm et al2). The potential information contained in that space can only be harvested by iterative approaches that freely analyze and reanalyze numerous regions from numerous vantage points—a process not realistically feasible 20 years ago, but easily achieved today.4,5 The common paper-printed, gated histograms so widely used today are an unwieldy and antiquated medium compared with direct and repeated manipulation of data on a computer. Just as our colleagues in radiology and nuclear medicine have turned away from the limited medium of film and toward the enhanced dimensionality of direct computer visualization, we should also abandon paper in favor of computers for the routine interpretation of flow cytometric data. Beyond the fact that Fromm et al2 come closer than most to fully exploiting the potential power of clinical flow cytometry, the question remains as to the fundamental role of even the most effective flow cytometric immunopheno-typing methods in the diagnosis of CHL. There are certain diseases (low-grade B cell lymphomas and acute leukemias) for which flow cytometry is a highly functional diagnostic modality, able to act alone as an acceptable confirmatory test for histologic or cytologic findings in most cases. There are other diseases (plasma cell myeloma and mastocytosis) for which flow cytometry has been fully validated, but its relative contribution to routine diagnosis is controversial because existing diagnostic criteria can generally be met by using other methods that yield comparable information. Given the distinct histopathologic appearance of most CHL cases and the relatively small number of immunohistochemical markers required for diagnostic confirmation, one could argue that the status quo will suffice in a large majority of CHL cases. However, in an era in which the gray zone between Hodgkin and non-Hodgkin lymphoma is attracting unprecedented scrutiny, it is refreshing to know that studies such as the study by Fromm et al2 are adding an advanced and apparently useful tool to the diagnostic armamentarium. Approaches to flow cytometry that treat data as objects embedded in high-dimensional space (rather than hierarchical sets of 2-dimensional histograms) hold the promise of converting flow cytometry from a simple immunophenotyping technology to one of true cytomic profiling.6–15 Just as the revolution in functional genomics and proteomics has led to unsupervised class discovery by the simultaneous analysis of thousands of expressed genes or proteins, flow cytometry can act as a cell-based proteomic discovery platform. Although in comparison with other proteomic analysis methods flow cytometry reduces the number of analyzed proteins from thousands to dozens, the added advantage of flow cytometry includes high-dimensional marker expression patterns being traceable to the level of the individual cell. Such an approach adds a truly spatial characteristic to the dimensionality of flow cytometry that is not found in other proteomic or genomic methods.9 Realizing this potential, however, requires a shift in attitude among diagnosticians toward the application of this powerful technology.