This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Maierhofer, C. Articles by Speicher, M. R. Search for Related Content PubMed PubMed Citation Articles by Maierhofer, C. Articles by Speicher, M. R.

(American Journal of Pathology. 2003;162:373-379.)
© 2003 American Society for Investigative Pathology

Technical Advance

Multicolor Deconvolution Microscopy of Thick Biological Specimens

Christine Maierhofer^*, Rainer Gangnus^*, Joachim Diebold and Michael R. Speicher^*

From the Institut für Humangenetik,^* Technische Universität München, München; the Institute of Pathology, Universität München, München; and the Institut für Humangenetik, Gesellschaft für Strahlenforschung Forschungszentrum für Umwelt und Gesundheit, Neuherberg, Germany

	Abstract

Top Abstract Materials and Methods Results Discussion References

 
One limitation in understanding disease at the cellular level has been the inability to efficiently analyze DNA on a cell-to-cell basis within the natural tissue context. However, DNA analyses at a single-cell resolution should be instrumental for the understanding of cancer cell biology, cancer evolution, for chromosomal mosaic analysis and rare cell events, and should provide otherwise inaccessible information on essential biological processes. Here we present a fluorescence in situ hybridization-based multicolor deconvolution technique for three-dimensional microscopy. We use up to seven different color channels for probe detection, which allows the simultaneous high-resolution localization of multiple point-like sources within a biological specimen with a thickness of up to 30 µm. In addition, a DNA counterstain is used for volume labeling of the nuclei offering the opportunity for a simultaneous segmentation of nuclei. Furthermore, as the instrumentation consists of a standard fluorescence microscope it represents a low-cost method as compared to confocal microscopy.

Current techniques for 3D microscopy generate 3D data by optical sectioning of the specimen. Up to date most 3D fluorescence microscopy is done using confocal microscopy, widely regarded as the gold standard. However, confocal microscopy has several drawbacks: the most serious drawback represents the amount of light required to produce a confocal image as the light dosage may bleach a dye. Furthermore, confocal systems that can detect a multitude of different fluorochromes using the entire color spectrum from the UV to the far infrared range are high-priced and therefore not affordable for most laboratories.

Deconvolution microscopy represents an alternative.⁹ Deconvolution refers to a wide-field image restoration by computational methods used to reduce out-of-focus fluorescence in 3D microscopy images.¹⁰ Here we demonstrate a new multicolor deconvolution microscopy approach. What distinguishes our work from previous studies is that up to eight different color channels can be used for the simultaneous analysis of multiple probes. As it uses a normal epifluorescence microscope with a motorized stage it represents, compared to confocal microscopy, a low-cost alternative. The data collection is faster than for most confocals so that bleaching problems are greatly reduced. We show multicolor fluorescence in situ hybridization (FISH) applications, which were applied for imaging deep into specimens, such as thick (30 µm) paraffin-embedded tissue sections. The potential of the technology is exemplified by the detection of portions within the same tumor sample with different chromosomal patterns and various degrees of chromosomal instability.

	Materials and Methods

Top Abstract Materials and Methods Results Discussion References

 
Tumor and Normal Tissue Sections

Histological sections were taken from paraffin-embedded tissue specimens, fixed in buffered formalin [4% phosphate-buffered saline (PBS)]. Normal ovarian tissue and tissue samples of two invasive ovarian carcinomas (13471B and 15078) were used. Case 13471B was a mixed mucinous/endometrioid ovarian carcinoma (grade 2, FIGO stage IA) and case 15078 was a serous ovarian carcinoma (grade 3, FIGO stage IIIC). For our analyses several serial thick sections from the same tissue block were prepared. In each case the first and last sections were stained with hematoxylin and eosin for histological analysis. In addition, 30-µm sections of a pTa bladder cancer, fixed in 85% ethanol, were included for further validation of our technology.

Fluorochromes

Fluorochromes tested included the dyes that we use for 7-fluorochrome multiplex FISH karyotyping:¹¹ 4'-6-diamidino-2-phenylindol, counterstains DNA and was used here for the volume labeling of nuclei (excitation, 350 nm; emission, 456 nm). For DNA probe labeling we used diethylaminocoumarin (426 nm; 480 nm); fluorescein isothiocyanate (490 nm; 520 nm); Texas Red (595 nm; 615 nm); and the cyanine dyes Cy3 (554 nm; 568 nm), Cy5 (652 nm; 672 nm), Cy5.5 (682 nm; 703 nm), and Cy7 (755 nm; 778 nm). In some experiments Cy7 was replaced by LaserProIR 790 (785 nm/801 nm). Filter sets were kindly provided by Chroma Technology Corp. (Brattleboro, VT).

Probe Sets

Three different probe sets were used. First, we assembled a 5-centromere probe set (Table 1) . These centromere probes were generously provided by Dr. M. Rocchi (University of Bari, Italy) (for detailed information see http://www.biologia.uniba.it/rmc/index.html).

Tissue Pretreatment, Hybridization, and Posthybridization Washes

Paraffin was removed by xylene for 60 minutes at 55°C and ethanol for 5 minutes. After air-drying the slides were incubated in 1 mol/L of NaSCN¹⁵ for 20 minutes at 72°C and subsequently washed twice in ddH₂O for 5 minutes each. Slides were digested with pepsin [15 µg/ml for various time periods (range, 15 minutes to 60 minutes) depending on cytoplasm and cell density] in 0.01 mol/L of HCl at 37°C. After washing with PBS, the slides were dehydrated over a graded series of ethanol (70%, 90%, and 99%) and air-dried. Before hybridization the slides were incubated in 70% formamide, 2x standard saline citrate (SSC) for at least 1 hour at room temperature. The slides were denatured in 70% formamide, 2x SSC, pH 7.0, at 73°C for 7 minutes.

The commercial probe kit was hybridized according to the manufacturer’s instructions. The centromeric probes were dissolved in the hybridization mixture (65% formamide, 0.5% dextran sulfate, 2x SSC), denatured at 78°C for 7 minutes and hybridized on pretreated slides without preannealing for 1 or 2 nights at 37°C. An aliquot (10 µl) of probe mixture containing labeled YAC and BAC probes (200 ng/each), human Cot-1 DNA (10 µg) in 50% formamide, 15% dextran sulfate, 1x SSC was denatured and applied on pretreated slides for 2 to 3 nights. After hybridization, the slides were washed three times (5 minutes each) with 4x SSC/Tween at 42°C and then three times (5 minutes each) with 1x SSC at 60°C. Blocking was done with 3% bovine serum albumin in 4x SSC/Tween for 30 minutes at 37°C. Afterward, the biotin- or digoxigenin-labeled probes were detected with avidin-Cy5.5 and sheep -digoxigenin fluorescein isothiocyanate, respectively. The slides were then washed three times (5 minutes each) in 4x SSC/Tween at 42°C, counterstained with 4'-6-diamidino-2-phenylindol, and embedded in p-phenylenediamine dihydrochloride anti-fade solution.

3D Image Acquisition, Processing, and Deconvolution Technique

To obtain a 3D data set, we recorded a number of two-dimensional images of the specimen in focal planes with a 250-nm distance. A motorized Zeiss Axioplan II Imaging epifluorescence microscope equipped with an Axiocam-MRm charge-coupled device camera was used. For image capturing and processing we used a newly developed software package. The mathematical details used in this software package were recently published.¹⁶ In brief, an accelerated maximum likelihood image restoration algorithm is used, which uses a conjugate gradient iteration scheme based on Poisson noise models. Regularization included modifications of the standard Tikhonov method, step size was determined using the Hessian matrix of the restoration functional. Here, we provide the first demonstration of the potential of the newly developed algorithms for the analysis of biological specimen. This new software tool was provided by Carl Zeiss Vision (Hallbergmoos, Germany) and has been dubbed AxioVision 3D deconvolution. A constrained, iterative deconvolution algorithm was used. Different procedures were used to select a point-spread function (PSF) as described in the text. The software package allows also the pseudo-coloring of fluorescent signals and 3D-reconstructions (AxioVision Inside4D).

The 4'-6-diamidino-2-phenylindol counterstain was used for volume labeling of the nuclei and thus for a simultaneous segmentation of nuclei. Hybridization signals were identified based on the signal intensities. Hybridization signals were assigned to segmented nuclei and counted.

Statistical Analysis

With the exception of tumor tissue 13471B, for which 64 nuclei were recorded, more than 100 nuclei/sample were evaluated. For each analyzed tissue or tissue area we calculated both the mean and the SD of the signal numbers per probe per cell. The mean values were compared with the Welch test (two-sample t-test assuming unequal variances), the standard deviations with the F-test. P values were subject to a two-sided significance level of 5%.

CGH Analysis

CGH was done according to previously published standard protocols.¹⁷

	Results

Top Abstract Materials and Methods Results Discussion References

 
We tested whether the co-localization and relative intensities of multiple components can be accurately studied in multiwavelength images. We were particularly interested in finding conditions to perform multicolor FISH in thick tissue sections so that hybridization signals from the intact cells in the interior portion of the section can be scored without being compromised by sectioning artifacts. Sections >20 µm thick contain a layer of cells in the center of the section, which have not been cut.¹⁸ We decided on the use of 30-µm-thick sections. Several experiments were designed in which single-labeled probes and combinatorially labeled probes were used.

Fluorochromes

We tested which fluorochromes are suitable for applications on thick tissue sections. The fluorochromes in the far infrared range, ie, Cy7 and LaserPro IR790, performed consistently poorly (data not shown) and were not used in further experiments. We could not achieve intensive fluorescence signals with diethylaminocoumarin in a reproducible manner on paraffin-embedded, formalin-fixed sections. However, good diethylaminocoumarin signals were achieved with a commercial probe kit on an EtOH-fixed specimen (Figure 1A) , suggesting that fixation protocols may have a considerable impact on the range of applicable fluorochromes. The subsequent experiments were done with the remaining fluorochromes, ie, fluorescein isothiocyanate, Cy3, Texas Red, Cy5, and Cy5.5.

View larger version (18K): [in this window] [in a new window]   Figure 1. 3D reconstruction of nuclei within 30-µm-thick tissue specimen after hybridization with multicolor probe sets, deconvolution, and false color assignment to the probes. The nuclei can be rotated on the computer screen and color channels can be inspected individually or in any possible combination. Note that the signals have a point-like appearance even in the optical axis. A: A commercially available probe set (UroVysion, Vysis) consisting of centromere probes for chromosomes 3 (red), 7 (green), and 17 (blue) and a probe for chromosome band 9p21 (yellow) was hybridized to a thick bladder cancer specimen. The reconstructed signals and the corresponding nucleus are depicted. B: A nucleus within a normal thick ovarian tissue specimen after hybridization with a five-fluorochrome centromere probe mix (Table 1) . Two signals of each probe were detected.

We determined PSFs empirically and theoretically using the following information: numerical aperture of the objective, working distance of the objective, wavelength of the emitted light, XYZ dimensions of the PSF, size of a pixel, spacing between Z slices, refractive index of the immersion medium and anti-fade, and thickness of both the specimen and the coverslip. We tested different deconvolution algorithms. This was necessary as imaging at some depth into a specimen invariably introduces some spherical aberrations. Furthermore, when multiple fluorophores are used to localize different components in the same sample, the image of the distribution of at least one of the fluorophores will be spherically aberrant.¹⁹ Accordingly, for some samples an additional correction for the spherical aberration into the PSF calculation improved the results, for others, no significant difference was apparent. Examples illustrating that the selected PSFs matched the aberrations in the images very well are shown in the figures.

Five-Color Centromere Probe Set

Normal Ovarian Tissue

The five color centromeric probe set (Table 1) was hybridized to normal 30-µm-thick ovarian tissue (Figure 1B) . As expected, two signals per probe were observed in most nuclei (Figure 2A) . We used the SD as a measure for the variability of the signal/probe/nucleus, which also reflects the degree of the chromosomal instability. As expected the SD was very low in the normal tissue. This low variation most likely was caused by hybridization variability and should not represent an actually existing low degree of aneuploidy within the normal tissue.

View larger version (72K): [in this window] [in a new window]   Figure 2. A: Mean number of signals per cell obtained with a five-fluorochrome centromere probe mix (Table 1) on normal ovarian tissue (compare to Figure 1B ) and on ovarian cancer specimens 13471B (B) and 15078. In the latter sample two separate areas (small regions within these areas are depicted in C), 15078 isolated and 15078 clustered were evaluated in addition. The SD reflects the observed signal variability and indicates the chromosomal instability (for details see text). B: Different views of a 3D-reconstructed nucleus within ovarian tumor tissue 13471B. The false color assignment corresponds to the column colors used in A, ie, the chromosome 7 signals are displayed in red, chromosome 8 signals in green, chromosome 11 signals in blue, chromosome 17 signals in yellow, and chromosome 18 signals in pink. C: Different regions based on their cell density were separately evaluated in sample 15078. Single planes of the z-stack from these two regions are depicted, the top row shows an area with a relative low density of cell nuclei (15078 isolated), the bottom row an area with tightly clustered nuclei (15078 clustered).

The centromere probe set was applied to two different ovarian tumor samples. Figure 2B illustrates the 3D-reconstruction of a nucleus from tumor specimen 13471B. In each sample the mean signal/probe/cell number was increased (Figure 2A) . In addition, a tremendous variability in the signal/probe/cell number was observed, which is reflected in the increased SD. Furthermore the data suggested that regions with differences in terms of their chromosomal composition and chromosomal stability existed within the same tumor specimen. We used the cell density as a morphological criterion and compared the signal/probe/nucleus numbers in a region with mainly isolated nuclei with a region containing tightly clustered nuclei in tumor specimen 15078 (Figure 2C) . We noted surprising differences for both the chromosomal composition and the chromosomal stability. The differences in the chromosomal makeup were most obvious for chromosome 7, which was relatively overrepresented in the isolated region but relatively underrepresented in the tightly clustered area (Figure 2A) . The mean signal/probe/cell numbers differed highly significantly for chromosomes 7, 8, and 18 (P values <0.008 for all three chromosomes) but were almost identical for chromosomes 11 and 15 (P values 0.87 for each chromosome).

An even more striking difference was observed for the chromosomal instability as estimated by the SD. The SD was extremely high in the isolated region (Figure 2A) . In this region the signal/probe/cell numbers varied so immensely that even in neighboring nuclei completely dissimilar signal numbers were counted, which suggests that chromosomes are passed to the daughter cells in a more or less random manner. In contrast, the signal patterns were relatively constant in the nuclei of the clustered region, reflecting a relatively stable segregation of chromosomes. These differences were highly significant (P values <10^-3 for all tested chromosomes).

Seven-Color Band-Specific YAC/BAC Probe Set

We hybridized a seven-color band-specific YAC/BAC probe set (Table 2) to the tumor sample 15078. As illustrated in Figure 3, A and B , there was an exact co-localization of combinatorially labeled probes. The mean and SD of the signals were calculated (Figure 3C) . As a control we corroborated the mean number of interphase signals with the CGH profile obtained with DNA extracted from the same tumor (Figure 3D) . This comparison showed a close correlation between the calculated ratio profiles and the signal/cell number for six of the seven probes. The only exception was the 20q13-probe, which yielded a slightly higher mean interphase number than expected from the CGH profile. This can be explained by the considerable heterogeneity within the tumor. Taken together the results suggest that combinatorially labeled probes, which represent spherically aberrant images, can be restored accurately by the constrained, iterative deconvolution applied here.

View larger version (57K): [in this window] [in a new window]   Figure 3. A panel of seven band-specific YAC and BAC probes (Table 2) was applied to ovarian cancer specimen 15078. A and B: Nuclei after hybridization and 3D-reconstruction. C: Mean number of YAC/BAC signals per cell. D: Corroboration between multicolor deconvolution microscopy results and CGH ratio profiles. The localization of the respective YAC/BAC clone is shown in the ratio profile, the number to the right indicates the mean signal per cell number as shown in C.

	Discussion

Top Abstract Materials and Methods Results Discussion References

Another advantage is that deconvolution microscope systems are cheaper than confocal microscopes. However, the deconvolution approach requires computational processing, which for our samples took typically somewhere from 30 minutes to 8 hours per stack. This constraint may be greatly reduced by an efficient use of faster computers with larger RAMs.

There may be constraints on the maximum thickness of a specimen that can be analyzed. The limitation is not within imaging but rather on the FISH side as signals with a sufficient intensity have to be achieved. Thirty-µm-thick sections represent a good compromise: they contain sufficient layers of cells that have not been cut but still allow obtaining of the hybridization signals with high-fluorescence intensities. The intensity of signals was already compromised when we tested the same protocol on 40-µm-thick sections (our unpublished data).

Another potential limitation, which applies to all tissue analysis approaches, represents regions where nuclei are tightly clustered so that there is little evidence from the images on where the borders of the nuclei are. The use of membrane-related protein markers, such as lamins or integrins, may reduce the problem.²⁰ Currently tests are underway to test whether the combination of such fluorochrome-tagged antibodies for nuclear surface staining together with our multicolor FISH is feasible.

We observed, as expected, a very homogenous signal pattern in the normal cell population. In contrast, the signal per cell variability was significantly higher in the tumor tissues. In tumor tissues, we observed the occurrence of regions with different aneuploid variants. Currently, we do not know which of these variants may suffer cell death or may be at selective proliferative advantage. However, this distinction may be important as it had been suggested that the predictive power of expression analysis may be compromised by this heterogeneity of tissue populations in primary tumors.⁴ Furthermore, it was hypothesized that tumors need a just-right instability to overcome new selection barriers. By contrast, if the level of genetic instability is too high, the accumulated damage in these too-high unstable cells may rise above the threshold for viability.²¹ It is tempting to speculate that cell patches as observed in the isolated area of tumor 15078 have a too-high instability pattern, which may not have any impact on the patient’s prognosis but instead may hamper expression analysis by introducing significant background noise to the data. In contrast, expression analysis with RNA extracted from a just-right unstable area, which may correspond to the clustered region, may improve the predictive power of expression analysis considerably.

Only a technology as presented here may contribute to the identification of areas with different chromosomal and instability patterns and opens avenues for new strategies for an improved resolution of expression data. A combination of interphase cytogenetics and gene expression analysis can be achieved by cutting serial sections from a tissue specimen. One section is used for FISH. The fluorescence signals can then be used as a template to guide dissection from the corresponding parallel tissue sections. After RNA isolation and microarray gene expression analysis, the results can, if desired, be interpreted again at single-cell resolution by application of one of the latest multilabeled mRNA in situ hybridization detection methods on further serial sections.^22,23 Further potential applications for our new single-cell analysis approach include the elucidation of the spatial order of the genome in differentiated cell tissue and the analysis of pathological processes in atypical proliferation, metaplasia, preneoplastic lesions, and carcinoma in situ.

	Acknowledgements

 
We thank Dr. S. Wagenpfeil for advice in statistical analysis; Drs. M. Cremer, T. Cremer, and R. Knuechel for critically reading of the manuscript and helpful discussion; and Margit Müller for expert technical assistance.

	Footnotes

 
Address reprint requests to Dr. Michael R. Speicher, Institut für Humangenetik Technische, Universität München, Trogerstr. 32 D-81675, München, Germany. E-mail: speicher{at}humangenetik.med.tu-muenchen.de

Supported by the Bundesministerium für Bildung und Forschung (NGNF KB PO6T5), the Wilhelm Sander-Stiftung, and TILL I.D., Gräfeling, Germany (stipend to C. M.).

Accepted for publication October 14, 2002.

	References

Top Abstract Materials and Methods Results Discussion References

 

1. Virchow RLK: Die Cellularpathologie in ihrer Begründung auf physiologische und pathologische Gewebelehre. A. 1858 Hirschwald, Berlin
2. Liotta L, Petricoin E: Molecular profiling of human cancer. Nat Rev Genet 2000, 1:48-56[Medline]
3. van ’t Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M, Peterse HL, van der Kooy K, Marton MJ, Witteveen AT, Schreiber GJ, Kerkhoven RM, Roberts C, Linsley PS, Bernards R, Friend SH: Gene expression profiling predicts clinical outcome of breast cancer. Nature 2002, 415:530-536[Medline]
4. Ince TA, Weinberg RA: Functional genomics and the breast cancer problem. Cancer Cell 2002, 1:15-17[Medline]
5. Potter H, Geller LN: Alzheimer’s disease, Down’s syndrome, and chromosome segregation. Lancet 1996, 348:66[Medline]
6. Migliore L, Botto N, Scarpato R, Petrozzi L, Cipriani G, Bonuccelli U: Preferential occurrence of chromosome 21 malsegregation in peripheral blood lymphocytes of Alzheimer disease patients. Cytogenet Cell Genet 1999, 87:41-46[Medline]
7. Geller LN, Potter H: Chromosome missegregation and trisomy 21 mosaicism in Alzheimer’s disease. Neurobiol Dis 1999, 6:167-179[Medline]
8. Sangha KK, Stephenson MD, Brown CJ, Robinson WP: Extremely skewed X-chromosome inactivation is increased in women with recurrent spontaneous abortion. Am J Hum Genet 1999, 65:913-917[Medline]
9. Carrington WA, Lynch RM, Moore ED, Isenberg G, Fogarty KE, Fay FS: Superresolution three-dimensional images of fluorescence in cells with minimal light exposure. Science 1995, 268:1483-1487[Abstract/Free Full Text]
10. McNally JG, Karpova T, Cooper J, Conchello JA: Three-dimensional imaging by deconvolution microscopy. Methods 1999, 19:373-385[Medline]
11. Azofeifa J, Fauth C, Kraus J, Maierhofer C, Langer S, Bolzer A, Reichman J, Schuffenhauer S, Speicher MR: An optimized probe set for the detection of small interchromosomal aberrations by 24-color FISH. Am J Hum Genet 2000, 66:1684-1688[Medline]
12. Knuutila S, Autio K, Aalto Y: Online access to CGH data of DNA sequence copy number changes. Am J Pathol 2000, 157:689[Free Full Text]
13. Lengauer C, Riethman HC, Speicher MR, Taniwaki M, Konecki D, Green ED, Becher R, Olson MV, Cremer T: Metaphase and interphase cytogenetics with Alu-PCR-amplified yeast artificial chromosome clones containing the BCR gene and the protooncogenes c-raf-1, c-fms, and c-erbB-2. Cancer Res 1992, 52:2590-2596[Abstract/Free Full Text]
14. Zhou H, Kuang J, Zhong L, Kuo WI, Gray JW, Sahin A, Brinkley BR, Sen S: Tumour amplified kinase STK15/BTAK induces centrosome amplification, aneuploidy and transformation. Nat Genet 1998, 20:189-193[Medline]
15. Hopman AH, van Hooren E, van de Kaa CA, Vooijs PG, Ramaekers FC: Detection of numerical chromosome aberrations using in situ hybridization in paraffin sections of routinely processed bladder cancers. Mod Pathol 1991, 4:503-513[Medline]
16. Schaefer LH, Schuster D, Herz H: Generalized approach for accelerated maximum likelihood based image restoration applied to three-dimensional fluorescence microscopy. J Microsc 2001, 204:99-107[Medline]
17. Speicher MR, Jauch A, Walt H, du Manoir S, Ried T, Jochum W, Sulser T, Cremer T: Correlation of microscopic phenotype with genotype in a formalin fixed, paraffin embedded testicular germ cell tumor using universal DNA amplification, comparative genomic hybridization and interphase cytogenetics. Am J Pathol 1995, 146:1332-1340[Abstract]
18. Thompson CT, LeBoit PE, Nederlof PM, Gray JW: Thick-section fluorescence in situ hybridization on formalin-fixed, paraffin-embedded archival tissue provides a histogenetic profile. Am J Pathol 1994, 144:237-243[Abstract]
19. Scalettar BA, Swedlow JR, Sedat JW, Agard DA: Dispersion, aberration and deconvolution in multi-wavelength fluorescence images. J Microsc 1996, 182:50-60[Medline]
20. Ortiz de Solorzano C, Malladi R, Lelievre SA, Lockett SJ: Segmentation of nuclei and cells using membrane related protein markers. J Microsc 2001, 201:404-415[Medline]
21. Cahill DP, Kinzler KW, Vogelstein B, Lengauer C: Genetic instability and Darwinian selection in tumours. Trends Cell Biol 1999, 9:M57-M60[Medline]
22. Zaidia AU, Enomotob H, Milbrandt J, Roth KA: Dual fluorescent in situ hybridization and immunohistochemical detection with tyramide signal amplification. J Histochem Cytochem 2000, 48:1369-1376[Abstract/Free Full Text]
23. Levsky JM, Shenoy SM, Pezo RC, Singer RH: Single-cell gene expression profiling. Science 2002, 297:836-840[Abstract/Free Full Text]

This article has been cited by other articles:

				S. Rauser, R. Weis, H. Braselmann, M. Feith, H. J. Stein, R. Langer, P. Hutzler, M. Hausmann, S. Lassmann, J. R. Siewert, et al. Significance of HER2 Low-Level Copy Gain in Barrett's Cancer: Implications for Fluorescence In situ Hybridization Testing in Tissues Clin. Cancer Res., September 1, 2007; 13(17): 5115 - 5123. [Abstract] [Full Text] [PDF]

				J. B. Geigl, S. Langer, S. Barwisch, K. Pfleghaar, G. Lederer, and M. R. Speicher Analysis of Gene Expression Patterns and Chromosomal Changes Associated with Aging Cancer Res., December 1, 2004; 64(23): 8550 - 8557. [Abstract] [Full Text] [PDF]

				J. C.L. Schuh Trials, Tribulations, and Trends in Tumor Modeling in Mice Toxicol Pathol, January 1, 2004; 32(1_suppl): 53 - 66. [Abstract] [PDF]

This Article Abstract Full Text (PDF) Purchase Article View Shopping Cart Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Maierhofer, C. Articles by Speicher, M. R. Search for Related Content PubMed PubMed Citation Articles by Maierhofer, C. Articles by Speicher, M. R.