Compensation is not dependent on signal intensity or on number of parameters

Abstract

A recent paper by Stewart and Stewart on compensation (1) provides useful information and examples of proper and improper compensation. Four-color (and more) analysis is becoming more common, if not routine in many laboratories, with the advent of relatively inexpensive desktop cytometers that collect this many fluorescence signals. As manufacturers extend the technology to perform six or more color experiments easily, the demands on users to understand the complexities of fluorescence compensation become greater. Therefore, I wish to correct a mistake in this manuscript to prevent serious misconceptions about compensation for multicolor applications. The authors observed that multiply-stained cells sometimes appear to be undercompensated, when compensation was properly set based on singly stained compensation controls. Based on this observation, they assert that single-color compensation controls are inadequate for setting proper instrument compensation. Indeed, they provide (in their Appendix) some mathematical modeling to explain why, as more and more colors are added, there appears a pervasive undercompensation in the multiparameter data. In their example, they assert that the fluorescence of a cell that is doubly stained and compensated would be less than the fluorescence of a cell that is singly stained and compensated, going through the mathematics to illustrate this. The first problem is that their model of compensation, even if it were correct, would result in overcompensation, and not the undercompensation the authors are trying to explain. Fundamentally, however, the basic model they use to illustrate how compensation works is incorrect. In their electronic schema, the compensated outputs for each detector are used as the inputs for the compensation for the other detectors. Such a circuit would be unstable, because the values change depending on the order in which they were calculated. In fact, for the single-stained cells example, if the authors were to calculate the detector outputs in the opposite order, they would find that the process yielded proper values! In reality, all inputs to compensation, whether done electronically or by software, are the uncompensated measurements (indeed, this is how the authors performed the calculation for the doubly stained example, in contrast to their singly stained example). When the correct procedure for compensation is performed (whether electronically (2) or by software (3, 4)), the compensated values will always be the same whether or not cells are singly or multiply-stained. This is consistent with our experience that even 11-color data can be properly compensated using singly-stained compensation controls (5). There is no doubt, however, that multicolor data can appear undercompensated even when the compensation is set properly, and that this appearance is exacerbated by the presence of multiple other colors on the same cell. Thus, the question remains as to why such data appear undercompensated, especially when the number of colors used increases (or when the signal intensity increases). As I reported in a manuscript in Cytometry (6), the answer is that this apparent undercompensation is a consequence of photon-counting statistics. The paper explains this in detail, and further demonstrates that we can no longer expect bivariate distributions of multicolor data to be rectilinear, as we have come to expect from our experience with two-color flow cytometry. In conclusion, compensation itself is not dependent on the number of fluorochromes used in an experiment, and neither is it dependent on the signal intensity of any given fluorochrome. Of course, the ability to estimate proper compensation is compromised by using compensation controls that are less bright than experimental stains, whether compensation is performed electronically or by software. Most importantly, the use of singly stained compensation controls are sufficient for properly setting compensation. Mario Roederer*, * Vaccine Research Center, National Institutes of Health, Bethesda, Maryland 20892-3015