Spectral Karyotyping

In the SKBR3 (a.k.a. SK-BR-3) cell line metaphase above on the right I've drawn blue contours around the two chromosome 17's (hint: you may need to zoom the picture or ask me for a full size version). One chr. 17 is at 3 o'clock and has two small greenish dots facing upwards. The other chr. 17 is in a translocation. This metaphase also shows several of SKBR3's signature marker chromosomes (all resolved by SKY™).

An important point is that the FISH reagent was simply added to the SKY probe cocktail prior to hybridization. I have also had a report from a different customer that an ~200 Kb probe also worked fine. For maximum versatility I suggest a blue or cyan (aqua) fluorophore and monochrome image acquisition (hint: the blue coumarin tyramide from www.nenlifesci.com may do the trick).

My thanks to Dr. Evelin Schröck (was at NCI, now in Jena, Germany) for SKY samples, and the 1998 CSHL students for hyb'ing the slides (hopefully for 1999 more genes FISH'd at the course). Contact me for details and future SKY™&FISH applications.

We at Applied Spectral Imaging, Inc., have also been working on bright-field microscopy imaging applications. One such application is the analysis of stains used in histology, pathology and immunohistochemistry. The most common such stain is called H&E, which is an abbreviation for hematoxylin and eosin. H&E tissue sections are the basis for the majority of non-cervical cancer clinical diagnoses made by pathologists! (For that matter, the Pap smear, used in cervical cancer screening, contains H&E plus Orange G, Light Green SF and optionally Bismarck Brown…we've also looked at these stains). Since H&E only has three spectra (the background illumination, the hematoxylin stain and the eosin stain) we've also been looking into combining H&E with one or more immunohistochemical stains, such as Vector Red and BCIP, and perhaps DAB (if you don't know what these abbreviations mean, don't worry, since you probably don't need to).

As an example of how spectral imaging can be used to analyze H&E stained pathology samples we show below an example taken from the first section we at ASI analyzed with a pathologist. The image is an H&E stained prostate cancer tissue provided from Dr. Farhad Moatamed, of the West Los Angeles Veterans Administration Hospital.

The image looks like a typical RGB color camera (or color photograph) of the tissue section. The cells on the top left are cancer (the major diagnostic feature in the image are the nuclei which appear as purple rings with dark purple dots in them), the middle orange area is extracellular matrix, and the cells on the lower right are normal prostate epithelial cells (the diagnostic feature is that the normal cell nuclei are solid purple ovals).

The data is displayed in terms of optical density (O.D.), which has the nice property of normalizing for the light source spectrum. Optical density is a little hard to get used to, but has the nice property of being proportional to the amount of stain present (a feature not very important in H&E, but which has uses in immunohistochemistry and other staining applications). An optical density of 0.3 corresponds to 50% transmission of the light. An optical density of 0.6 corresponds to 25% transmission. An optical density of 1.0 corresponds to 10% transmission, while an optical density of 2.0 corresponds to 1% transmission (the advantage of O.D. is that a doubling of the amount of stain, i.e. a doubling of the absorbing material, corresponds to adding O.D.'s together: 0.3 + 0.3 O.D. = 0.6 O.D., or 1.0 + 1.0 = 2.0 O.D. which is easier -- for some of us -- to keep track of than multiplying the fractional amounts, i.e. 0.5 x 0.5 = 0.25, and 0.1 x 0.1 = 0.01.

 We can do something very novel and powerful by having the H&E (and illumination) reference spectra and a spectral image of the pathology tissue section: We can perform an mathematical operation we call Spectral Un-Mixing (or SUN), and display two new images showing the distribution of hematoxylin alone and eosin alone. Of course a pathologist could simply have their histotechnician stain other sections with each single stain, however this is not very useful diagnostically. We won't bore you with the mathematics (relatively straight-forward spectral analysis but with the novel feature of providing component images for each diagnostic stain). Here are the single stain component images for the above map:

The surprising (to some) result is that eosin is present in all the nuclei, and is mostly uniform distribution in the normal nuclei (purple ovals in the lower right of the H&E and hematoxylin component image) while eosin is not uniform in the cancer nuclei at top left.

The real power of SUN is that we can do this on more than just H&E, i.e. on a immunohistochemistry section stained for H&E plus Vector Red and BCIP. A clinically relevant example would be to perform double immunohistochemistry on a breast carcinoma tissue section for estrogen receptor (ER, a sex steroid growth factor receptor, i.e. stain with Vector Red, a red and fluorescent-red immunostain) and the Her-2/neu oncoprotein (a cancer marker, i.e. stain with BCIP, a turquoise immunostain). The goal being to identify what treatment (i.e. tamoxifen if ER positive, or Herceptin™ if Her-2/neu overexpressing) to give the patient, as well as to answer basic research questions that have not been answered to date (especially, what is the frequency of ER positive, Her-2/neu overexpressing breast carcinomas?). We expect that SUN and spectral imaging will provide new diagnostic tools for pathologists and clinicians.

Anyone interested in learning more about SKY™, spectral pathology, other spectral microscopy applications, or other uses of the SkyVision™ technology can contact an ASI representative.