Image Cytometry Protocols

Dear Editor,

In the May 2013 issue of the Journal of Histochemistry & Cytochemistry, two papers were published based on quantification in light microscopic images or image cytometry (Garrido et al. 2013; Ruiz-Rosado et al. 2013). As Garrido and co-workers stated, valid quantification is essential to obtain meaningful numbers. However, the authors of this letter to the editor have serious questions regarding both publications and would like to outline these here. Our criticism is focused on different issues in the two publications, so we discuss them separately.

The article by Garrido et al. (2013) describes quantitative histological assessment of xenobiotic effects on rat liver, thyroid gland, and the pars distalis of the pituitary gland using whole-slide automated image analysis. We consider it as an important publication that allows preclinical screening of compounds for safety evaluation. This methodology combines quantitative and structural information on the effects of xenobiotics on tissues and we hope that it will have a considerable impact on drug screening research. However, the major flaw of this article is the lack of information on how the image cytometry was exactly performed. On the basis of the procedures described in the Materials & Methods section of the article, it is impossible to reproduce the study.

Information that is lacking concerns issues such as, (1) What were the exact fixation and staining procedures? (2) How was the staining [polymerized diaminobenzidine (polyDAB)] intensity quantified? (3) Hematoxylin and polyDAB have overlapping absorbance spectra, and thus, what is the light that has been used to illuminate the sections—white light or light of a specific wavelength? If the latter is the case, what wavelength? (4) How was the threshold set to distinguish between “dark staining” and “weak staining”? (5) The same questions concern the quantification in H&E stained sections. How did the authors segment the hematoxylin-stained nuclei from the eosin-stained cytoplasm? (6) What type of microscope, what magnifications, and what type of camera were used, including numbers of pixels of whole-section images and higher magnifications? (7) Digital cameras are very sensitive to infrared radiation while our eyes are not, and infrared blocking filters have to be used to reduce noise in the digital images. Therefore, the question arises whether infrared radiation was blocked or not.

In summary, essential details of the image cytometry procedures and, in particular, the image capturing procedures are lacking. We have recently published protocols for valid image cytometry in 2D and 3D that include all these essential aspects of the methodology (Chieco et al. 2013). We hope that image cytometry and quantification in microscopic images will acquire the status that is predicted in the July 2012 issue of Nature Methods (see Editorial 2012) and in Yuan et al. (2012), which describe an essential role for image cytometry technology in large-scale cancer studies. We hope that the protocols for the retrieval of meaningful quantitative data from microscopic images that we described (Chieco et al. 2013) will help to bring image cytometry into that position and help authors to provide details on methodology in publications that are necessary to check whether the image cytometry has been performed validly. Please note that we do not assume that Garrido et al. (2013) have used inadequate image cytometry technology, but we state that essential information on the methodology used is lacking in their article.

The second article in the May 2013 issue on image cytometry is Ruiz-Rosado et al. (2013), which describes rat muscle fiber typing using image analysis. The Materials & Methods section of this article does not provide any information to address the questions formulated above in relation to the Garrido et al. (2013) publication. Additionally, a number of other questions have to be raised. First, the incubations of the sections to demonstrate cytochrome c oxidase and ATPase activity both lasted 60 min, whereas control incubations are not mentioned anywhere. It has been shown by Kugler et al. (1988) that the linearity between incubation time and formation of the final reaction product, polyDAB, lasts only 10 min at 37C. Therefore, an incubation period of 60 min guarantees non-linearity and thus it is unpredictable whether the amount of the final reaction product reflects the cytochrome c activity (see Van Noorden and Butcher 1991). Second, control incubations showing formation of the final reaction product due to reactions other than that of cytochrome c oxidase are essential for valid enzyme histochemistry or metabolic mapping (Van Noorden and Frederiks 1992; Van Noorden 2010). Third, cytochrome c oxidase is a mitochondrial enzyme and mitochondria are preferentially localized at the periphery of skeletal muscle cells, as can be observed in Figs. 1M and 1N of Ruiz-Rosado et al. (2013). How was this heterogeneous intracellular localization of cytochrome c oxidase activity taken into consideration in fiber typing?

Thus, the article by Ruiz-Rosado et al. (2013) misses essential procedural information in the Materials & Methods section, like the Garrido et al. (2013) article. Furthermore, staining of the sections for cytochrome c oxidase activity is not valid because the incubation period was far too long to ensure linearity between incubation time and formation of the final reaction product. Moreover, proper control incubations have not been performed to determine the specificity of the color formation of the final reaction product, polyDAB.

Validity of histochemical and cytochemical staining procedures was an issue of high priority in the 1970s and 1980s, but since then, these validity issues have become less important. Image manipulation with Adobe Photoshop, for example, has become routine, even when used for quantitative histochemistry applying image cytometry, despite the fact that image cytometry has to be performed on images of microscopic objects that have been captured without any manipulation of the images (Chieco et al. 2013).

This letter to the editor has been written because we feel the need to advocate for methodological sound microscopy, histochemistry, and image cytometry, especially now that microscopy, including quantitative microscopy, has become such an important tool in the life sciences.