|
|
Biotechniques. Author manuscript; available in PMC 2005 August 2. Published in final edited form as: Biotechniques. 2004 February; 36(2): 240–247. | PMCID: PMC1182179 NIHMSID: NIHMS2032 |
Computer-assisted image analysis protocol that quantitatively measures subnuclear protein organization in cell populations Ty C. Voss, Ignacio A. Demarco, Cynthia F. Booker, and Richard N. Day Many nuclear proteins, including the nuclear receptor co-repressor (NCoR) protein are localized to specific regions of the cell nucleus, and this subnuclear positioning is preserved when NCoR is expressed in cells as a fusion to a fluorescent protein (FP). To determine how specific factors may influence the subnuclear organization of NCoR requires an unbiased approach to the selection of cells for image analysis. Here, we use the co-expression of the monomeric red FP (mRFP) to select cells that also express NCoR labeled with yellow FP (YFP). The transfected cells are selected for imaging based on the diffuse cellular mRFP signal without prior knowledge of the subnuclear organization of the co-expressed YFP-NCoR. The images acquired of the expressed FPs are then analyzed using an automated image analysis protocol that identifies regions of interest (ROIs) using a set of empirically determined rules. The relative expression levels of both fluorescent proteins are estimated, and YFP-NCoR subnuclear organization is quantified based on the mean focal body size and relative intensity. The selected ROIs are tagged with an identifier and annotated with the acquired data. This integrated image analysis protocol is an unbiased method for the precise and consistent measurement of thousands of ROIs from hundreds of individual cells in the population. The interphase cell nucleus is a fluid, but extremely organized, cellular organelle ( 1– 3). The assembly of higher-order protein structures at particular sites in the nucleus plays a critical role in the control of gene expression. Knowledge of these processes is being gained through the combination of biochemical, genetic, and molecular approaches. Importantly, these in vitro approaches are now being complemented by noninvasive imaging techniques that allow direct visualization of protein activities in their natural environment within the living cell. This became possible with the cloning of genes that encode fluorescent proteins (FP) from marine organisms ( 4) and extensive mutagenesis to yield proteins that fluoresce from the blue to the red range of the visible spectrum ( 5). We are using multispectral imaging to characterize the association of FP-labeled transcription factors and coregulatory proteins at particular sites in the living cell nucleus ( 6– 8). Many nuclear proteins are enriched in distinct subnuclear domains, ranging from spherical bodies to more diffuse and irregular speckles ( 9). For example, the transcriptional co-repressor proteins, nuclear receptor co-repressor (NCoR) and silencing mediator for retinoid and thyroid hormone receptors (SMRT), are organized with their histone deacetylase partners in discrete nuclear bodies called matrix-associated deacetylase (MAD) bodies ( 10, 11). Our imaging studies using expressed co-repressor proteins labeled with FPs revealed a substantial heterogeneity from cell to cell in the organization of these subnuclear bodies. This heterogeneity creates a problem for image analysis, where the interpretation of protein distribution in high-resolution microscopy images is subjective and may not be representative of the cell population. The rigorous quantification of these subcellular features requires high-resolution images of cells that are objectively selected from the population, followed by an unbiased analysis. Addressing the latter requirement, computer algorithms have been developed to achieve more rigorous and quantitative extraction of data from digital microscopy images (reviewed by Reference 12). Recent improvements in both computer hardware and software have facilitated the automated manipulation of large multidimensional data sets consisting of many images ( 13). However, given the power of fully automated image analysis techniques, there are remarkably few reports detailing the design of these procedures for specific problems in cell biology. Further, there are few methods available for the unbiased selection of FP-expressing cells. Here, we use cells co-expressing yellow and red FPs (YFP and RFP, respectively) to demonstrate an integrated method for unbiased cell selection and subsequent analysis of the acquired images. This approach will provide more detailed and precise information about mechanisms that control subcellular protein organization. Expression Constructs and Transfection of Cell Lines Optimized plasmids encoding the YFP (BD Biosciences Clontech, Palo Alto, CA, USA) and the monomeric RFP (mRFP) variant of Discosoma sp., kindly provided by Dr. R. Tsien (University of California at San Diego) ( 14), were used for expression of the fusion proteins. The sequence encoding mRFP was substituted for the FP encoding sequence in the EYFP-C2 vector (BD Biosciences Clontech) to generate the mRFP expression vector. The cDNA encoding the C-terminal region (amino acids 962–2454) of the mouse NCoR ( 10, 15) was inserted in-frame to generate the YFP-NCoR expression vector. The expression vectors were verified by automated nucleotide sequencing. Mouse embryonic pituitary GHFT1–5 cells were transfected by electroporation and cultured for 24 h on glass coverslips as described previously ( 7). Digital Imaging of Protein Organization in Living Cells The cover glass with the monolayer of cells was transferred to a medium-filled chamber that was fitted to the stage on the microscope ( 7). The wide-field fluorescence microscopy (WFM) images were acquired using an inverted Olympus IX-70 microscope (Olympus America, Melville, NY, USA) equipped with a 1.2 numerical aperture, 60× aqueous-immersion objective lens. A 150-W xenon/mercury combination lamp was used to illuminate the samples. A Model 1962 long-term stabilizer (Opti Quip, Highland Hills, NY, USA) was used to keep light intensity constant for accurate quantitative data collection. The YFP filter combination was 510/20 nm excitation and 560/40 nm emission, and the RFP filter combination was 560/40 nm excitation and 630/60 nm emission. Grayscale images with no saturated pixels were obtained using an Orca-200 cooled digital interline camera (Hamamatsu, Bridgewater, NJ, USA). All images were collected at a similar gray-level intensity by controlling the excitation intensity with constant neutral density filtration and by varying the on-camera integration time (0.1–8 s). Reference images of standard fluorescent beads were acquired to monitor consistency of microscope performance for all quantitative imaging experiments. All image files were processed for presentation using ISee software (ISee Imaging Systems, Raleigh, NC, USA) and Canvas 8.0 software (Deneba, Miami, FL, USA). Automated Image Analysis A series of computerized image analysis functions were integrated into a single algorithm using the ISee graphical programming software (ISee Imaging Systems). The first subroutine uses a histogram-based statistical method to optimally threshold the image acquired in the RFP channel to identify the whole cell region of interest (ROI). The mean intensity of the area outside the whole nucleus ROI was measured in both the red and yellow channel images to define the background fluorescence. Optimal thresholding of the yellow fluorescence channel image was then used to select the whole nucleus ROI. Measurements of mean fluorescence intensity in the whole cell ROI and whole nucleus ROI were used to estimate the relative expression of mRFP and YFP-NCoR in each cell. The normalized relative fluorescence intensity for all images was expressed as gray level per second exposure time. Another subroutine then takes the whole nucleus ROI as input and optimally thresholds this region using an iterative method to separate areas of bright fluorescence from surrounding regions. Next, the subroutine measures the shape of each identified bright ROI using several parameters. For the studies of NCoR focal bodies described here, ROIs were empirically defined as statistically significant regions of elevated fluorescent intensities that have a contiguous size between 10 and 3000 pixels. Additionally, the spherical NCoR foci were defined by a roundness value between 0.9 and 1.5 [roundness = (4 × π × total area)/(perimeter2), roundness of perfect circle = 1] and an axial ratio value between 1 and 1.3. The algorithm automatically selects the ROIs that meet empirically determined shape parameters of NCoR protein focal bodies for further analysis. If the ROI does not meet the requirements of spherical NCoR foci, then the ROI is reanalyzed by automatic thresholding and shape measurements to determine if NCoR foci are located within the original ROI. The process is repeated until all ROIs are evaluated. The area and fluorescence intensity of each selected focal body ROI is automatically measured and recorded. The center position of the selected NCoR focal body is then used to place a second rectangular ROI that measures the fluorescence intensity of the nucleoplasm surrounding the focal body. The size of the surrounding square ROI is four times that of the selected focal body. All the selected ROIs are marked in the image, and each is annotated with the acquired data. All the measurements were automatically exported to text files, and further analysis was performed using spreadsheet software (Microsoft® Excel®) to determine the relationship between the labeled protein expression levels and subnuclear organization. Unbiased Selection of Transfected Cells Using mRFP When cells are cotransfected with plasmids encoding two or more protein fusions to FP color variants, almost all of the transfected cells express each different color protein—although the relative levels of the different FPs may vary substantially from cell to cell ( 6– 8). Here we exploit this observation in the development of a nonbiased method to select transfected cells for analysis of subcellular protein distribution and organization. Our approach was to cotransfect cells with an expression plasmid encoding the optimized mRFP of Discosoma sp. ( 14) and a plasmid encoding the protein of interest, YFP-NCoR. This allowed selection of cells for imaging based upon the expression of the diffuse cellular mRFP without prior knowledge of the subnuclear organization of the co-expressed YFP-NCoR. The images shown in demonstrate the selection of cells based upon the expression of mRFP and the subsequent detection of YFP-NCoR among the cells in the transfected population. Importantly, examination of over 100 randomly selected mRFP-expressing cells revealed that over 95% also contained a detectable nuclear YFP-NCoR fluorescence signal. This confirmed that images of transfected cells expressing a protein of interest could be obtained using the mRFP channel without user bias to particular patterns of YFP-labeled protein distribution or expression level. Integration of Automated Image Analysis Protocols To rigorously quantify the FP-NCoR subnuclear organization in the cell population, we have integrated several automatic rule-based subroutines that consistently select and measure subnuclear protein bodies from images of many cells (see Materials and Methods). The logic diagram shown in illustrates the flow of information through the algorithm. Both YFP-NCoR and mRFP fluorescence data are utilized by the algorithm to quantify protein expression level and protein organization. In this way, all images were rapidly analyzed using the same rule-based system with no user intervention. The algorithm selected the whole cell and measured the mean mRFP fluorescence intensity. Similarly, the YFP-NCoR intensity in the whole nucleus was also automatically measured. The mRFP and YFP fluorescence intensity data were then combined with image exposure times to estimate the relative protein expression level in each cell. The examples shown in illustrate the autoselected whole cell or whole nucleus, with the ROI outlined in yellow. As is typical of the transient transfected cell population, these example cells expressed different relative levels of both mRFP and YFP-NCoR, and the images also illustrate the highly variable subnuclear organization of YFP-NCoR. A second subroutine in the algorithm characterized the organization of YFP-NCoR. ROIs were identified for YFP-NCoR in the nucleus of both cells (), and the morphometric data describing each ROI was automatically acquired (summarized in Table 1). Each bright focal body is a region of highly concentrated NCoR protein, which is surrounded by regions that contain a lower concentration of NCoR as illustrated by the intensity profile plot (). The ratio of fluorescence signal originating from the foci to that from the surrounding region (white squares, ) defines the enrichment factor (EF), which is the steady-state concentration of protein maintained in the focal body. This relative intensity-based analysis approach revealed that the average EF for the foci in cell B was 3-fold higher than that for cell A ( and Table 1). Finally, the algorithm determined the organization factor (OF), which is the product of foci size and the EF. Thus, greater mean OF values correspond to cells with larger and more distinct focal bodies ( and Table 1). | Table 1Summary of Morphometric Data Extracted from Example Cell Images |
Rigorous Image Analysis of Living Cell Populations To validate this method, images of over 100 cells were collected based solely on the mRFP signal and were then analyzed for YFP-NCoR organization using the integrated automated algorithm. Graphical analysis showed that increased levels of cellular YFP-NCoR expression were clearly related to increased OF values (). Statistical regression and analysis of variance (ANOVA) confirmed that both foci size and EF parameters were significantly related to the YFP-NCoR expression level in each cell (ANOVA F test P-value <5.0 × 10 −2 considered significant; YFP-NCoR versus foci size, ANOVA F test P-value = 8.9 × 10 −12; YFP-NCoR versus EF, ANOVA F test P-value = 8.4 × 10 −12), with the OF value being most strongly correlated (YFP-NCoR versus OF, ANOVA F test P-value = 1.8 × 10 −13) (). This indicates that the OF value, which combines the measurements of both foci size and intensity, provides the most robust assessment of co-repressor subnuclear organization. By establishing the quantitative relationship between YFP-NCoR fluorescence intensity and OF values, this method confirms earlier qualitative observations that focal body formation depends on the amount of co-repressor protein that is expressed in each cell ( 10). Surprisingly, the population analysis also revealed that the expression levels of the cotransfected YFP-NCoR and mRFP were only weakly correlated to one another (ANOVA F test P-value = 5.0 × 10−2) (). This indicated that while almost all transfected cells expressed both mRFP and YFP-NCoR, the mRFP expression was not a good predictor of YFP-NCoR expression level. To address concerns that mRFP expression might in some way influence YFP-NCoR, we also compared mRFP fluorescence intensity to the YFP-NCoR organization in each cell of the population. Statistical analysis revealed that there was no significant relationship between these two parameters (ANOVA F test P-value = 3.3 × 10−1) (), eliminating the possibility that mRFP expression influenced the organization of YFP-NCoR. This paper describes a strategy for the consistent and unbiased analysis of the subcellular organization of FP-fusion protein in the cell population. Although we have tested many strategies, this integrated method was chosen because it provided the most robust quantification of subnuclear bodies in a large population of cells, using the most straightforward image acquisition and analysis protocols. The approach begins with the random selection of the transiently transfected cells for image analysis. The advantage of the transient transfection approach is that it allows the rapid examination of many different FP-fusion proteins without the labor-intensive generation of stable cell lines. However, the efficiency of transfection is variable, and not all cells contain detectable levels of FP-fusion proteins. Therefore, it is desirable to have a method to efficiently screen for the transfected cells within the population. Here, we demonstrated that the co-expression of mRFP provided a useful marker for cells that also expressed a FP-labeled protein of interest. Importantly, we showed that the co-expression of mRFP did not affect the subcellular organization of the YFP-NCoR. Together, these results demonstrate that the mRFP signal can be used efficiently to select for cells co-expressing proteins labeled with other spectral variants of the FPs without biasing the analysis of the protein of interest. After selection of the cells based on mRFP, we used an automated image analysis algorithm to rigorously quantify the subcellular distribution of the protein of interest. Similar analytical functions have been used to extract quantitative data from images of living cell nuclei. For example, automated segmentation has been used to identify and measure intranuclear cajal bodies and RNA splicing factor speckles ( 16, 17). However, our method is novel in that it integrates several individual analytical subroutines into a single rule-based system that measures multiple aspects of the cells; namely the foci size, the foci relative intensity, and the approximation of expression level for both fusion proteins. This provides a more detailed quantification of subcellular protein organization with minimal user input, increasing throughput, and decreasing the subjectivity of the analysis. Using a population of YFP-NCoR-expressing cells as an example, this method precisely quantified a significant relationship between fusion protein expression level and higher-order protein organization. This concentration-dependent behavior is not idiosyncratic of co-repressor proteins, since we have qualitatively observed similar behavior with several other transcriptional regulators ( 6, 18). Further, our immunostaining studies indicate these concentration-dependent effects occur at near physiological expression levels. In the case of co-repressor protein, even modest increases in expression level (approximately 5-fold over endogenous) result in the increased formation of subnuclear foci (data not shown). Therefore, it is critical that imaging studies of subcellular protein organization carefully account for protein expression levels, and our method does this. Further, the unbiased selection of transfected cells using mRFP allows for the inclusion of cells expressing very low levels of the protein of interest (see ) that would be missed using standard approaches. This method will likely have important applications to many cell signaling studies. For example, the effect of hormone treatment on the subcellular organization of proteins of interest can be readily determined for the transfected cell population. In addition, other FP color variants or vital stains could also be included in the analysis to address how different proteins may function to regulate the subcellular organization of a protein of interest. In both of these cases, it would be critical to confirm that observed changes in organization do not result from differences in fusion protein expression level. This general approach to image analysis that precisely accounts for FP-fusion protein expression variability in cell populations is widely applicable to study the subcellular organization of other proteins. The modular nature of our algorithm allows modification for the analysis of other subcellular features. For example, we have empirically chosen a statistical method to determine the optimal thresholds used in defining NCoR subnuclear foci. However, if other image segmentation methods are found to be superior for selecting additional cellular features, these subroutines could easily be substituted into our integrated algorithm. Likewise, the preliminary implementation of this algorithm uses only a single focal plane to approximate subnuclear organization, but future versions will be modified to take advantage of multiple focal plane image sets for evaluation of the full nuclear volume. Emphasizing this trend in the evolution of more complex analytical tools, it has been suggested that automated algorithms will be indispensable for understanding large multidimensional imaging data sets ( 13). Although we have implemented this algorithm using specific commercial software, other software packages also allow integration of complex image analysis functions. Thus, this type of analysis could be performed in many laboratories. Therefore, we suggest this integrated method will be useful as a template for the development of additional automated image analysis protocols. Because there can be substantial heterogeneity between protein organization in individual cells, rigorous analysis requires that large sets of unbiased images be acquired from the cell population. Once acquired, the accurate quantitative analysis of the high-resolution microscopic images requires the consistent extraction of thousands of measurements, which is only made practical with the use of computer automation. Here, we demonstrated an integrated approach for the unbiased selection of cells and the consistent quantitative analysis of protein distribution within those cells, providing an improved method with the precision necessary to address many critical questions in cell biology. We thank Dr. Roger Tsien for the plasmid encoding the optimized mRFP. We also wish to thank Dr. Ammasi Periasamy, Director of the W.M. Keck Center for Cellular Imaging, for his help with these studies. This work was supported by the National Institutes of Health (NIH) grant nos. DK43701 (R.N.D.) and F32DK60315-01 (T.C.V.). 1. Carmo-Fonseca M. The contribution of nuclear compartmentalization to gene regulation. Cell. 2002;108:513–521. [PubMed] 2. Lamond AI, Earnshaw WC. Structure and function in the nucleus. Science. 1998;280:547–553. [PubMed] 3. Misteli T. The concept of self-organization in cellular architecture. J Cell Biol. 2001;155:181–185. [PMC free article] [PubMed] 4. Labas YA, Gurskaya NG, Yanushevich YG, Fradkov AF, Lukyanov KA, Lukyanov SA, Matz MV. Diversity and evolution of the green fluorescent protein family. Proc Natl Acad Sci USA. 2002;99:4256–4261. [PMC free article] [PubMed] 5. Patterson G, Day RN, Piston D. Fluorescent protein spectra. J Cell Sci. 2001;114:837–838. [PubMed] 6. Schaufele F, Chang CY, Liu W, Baxter JD, Nordeen SK, Wan Y, Day RN, McDonnell DP. Temporally distinct and ligand-specific recruitment of nuclear receptor- interacting peptides and cofactors to subnuclear domains containing the estrogen receptor. Mol Endocrinol. 2000;14:2024–2039. [PubMed] 7. Day RN, Periasamy A, Schaufele F. Fluorescence resonance energy transfer microscopy of localized protein interactions in the living cell nucleus. Methods. 2001;25:4–18. [PubMed] 8. Enwright JF, Kawecki-Crook MA, Voss TC, Schaufele F, Day RN. A PIT-1 homeodomain mutant blocks the intranuclear recruitment of the CCAAT/enhancer binding protein alpha required for prolactin gene transcription. Mol Endocrinol. 2003;17:209–222. [PMC free article] [PubMed] 9. Spector DL. Nuclear domains. J Cell Sci. 2001;114:2891–2893. [PubMed] 10. Soderstrom M, Vo A, Heinzel T, Lavinsky RM, Yang WM, Seto E, Peterson DA, Rosenfeld MG, Glass CK. Differential effects of nuclear receptor co-repressor (N-CoR) expression levels on retinoic acid receptor-mediated repression support the existence of dynamically regulated co-repressor complexes. Mol Endocrinol. 1997;11:682–692. [PubMed] 11. Downes M, Ordentlich P, Kao HY, Alvarez JG, Evans RM. Identification of a nuclear domain with deacetylase activity. Proc Natl Acad Sci USA. 2000;97:10330–10335. [PMC free article] [PubMed] 12. Eils R, Athale C. Computational imaging in cell biology. J Cell Biol. 2003;161:477–481. [PMC free article] [PubMed] 13. Swedlow JR, Goldberg , Brauner E, Sorger PK. Informatics and quantitative analysis in biological imaging. Science. 2003;300:100–102. [PubMed] 14. Campbell RE, Tour O, Palmer AE, Steinbach PA, Baird GS, Zacharias DA, Tsien RY. A monomeric red fluorescent protein. Proc Natl Acad Sci USA. 2002;99:7877–7882. [PMC free article] [PubMed] 15. Horlein AJ, Naar AM, Heinzel T, Torchia J, Gloss B, Kurokawa R, Ryan A, Kamei Y, et al. Ligand-independent repression by the thyroid hormone receptor mediated by a nuclear receptor co-repressor. Nature. 1995;377:397–404. [PubMed] 16. Eils R, Gerlich D, Tvarusko W, Spector DL, Misteli T. Quantitative imaging of pre-mRNA splicing factors in living cells. Mol Biol Cell. 2000;11:413–418. [PMC free article] [PubMed] 17. Platani M, Goldberg , Swedlow JR, Lamond AI. In vivo analysis of Cajal body movement, separation, and joining in live human cells. J Cell Biol. 2000;151:1561–1574. [PMC free article] [PubMed] 18. Day RN, Nordeen SK, Wan Y. Visualizing protein-protein interactions in the nucleus of the living cell. Mol Endocrinol. 1999;13:517–526. [PubMed]
|
See other articles in PMC that 
