This paper represents a significant advance that was achieved by putting together a number of existing concepts to produce what may be considered an approach to a cell biological equivalent of nucleotide or cDNA arrays. The basic design involved culturing HeLa cells in 384-well plates, treating the cells with 13 threefold dilutions of 100 compounds, and then staining cells in wells with one of 11 different fluorescent probes. The authors termed this as “hypothesis-free molecular cytology.” A variety of parameters such as cell size and shape, as well as fluorescent intensity, were quantified automatically. These data were then reduced in a variety of ways, including what the authors termed a titration-invariant similarity score (TISS) in order to characterize effects of the tested compounds on the cell type cultured.
Although the paper is titled as “multidimensional drug profiling” and the choice of 11 probes was “hypothesis-free,” the potential uses of the approach outlined extend far beyond drug profiling based on “hypothesis-free” cytology. It is not difficult to imagine any reasonable number of probes selected to test defined hypotheses regarding a wide variety of cell biological phenomena. Rather than restrict the analysis to one cell type, multiple cell types could be included in the design. And, although the authors emphasized the importance of reducing multiple descriptors of cell phenotype to a single number, one can also envisage examining multiple parameters individually or using other methods of multivariate analysis to characterize results.
Although the use of arrays has provided many valuable insights into tissue responses to manipulation, disease, development, aging, etc., it is, after all, the cellular responses to induced transcription alterations that constitute the ultimate, perhaps even the defining, events. The approach outlined in this paper suggests many ways in which the steps between transcription and cellular phenotype may be better understood.
In a recent technological advance published in the journal Science by Z.E. Perlman et al., the group provides convincing evidence of the use of high-throughput cytological profiling by automated microscopy. This system is essentially a combination of microscopy-based single cell analysis with multi-level system biology computation, which enables the automated assessment of cells in culture. The paradigm the researchers have chosen initially as a proof of concept is to evaluate dose-dependent phenotypic changes of specific drug effects. The analysis scheme that the group has developed is really spectacular in the sense that previously automated microscopy has lagged behind other functional genomic and proteomic high-throughput technologies to assess multiple transcripts, peptides, or relevant pieces of information at the cellular level.
The present report uses a method that is devoid of a hypothesis in that these are unbiased measurements that employ individual descriptors as characteristics to identify aspects of a cell’s behavior under experimental conditions. The authors use different dyes, histochemical, and immunocytochemical preparations to identify the nucleus and cytoplasm of HeLa cells plated in culture. This is really a very interesting paradigm because the authors are able to statistically combine, compare, and contrast 93 descriptors leading to approximately 109 data points, which is bewildering considering most cytological examinations are using only single, double, or triple labeling techniques. The data is very convincing and enables the identification of effects of different drugs on these cells in an unbiased fashion, i.e.: the drugs can be blinded, and effects can be assessed at the cellular level without any prior knowledge of the mechanism of action. Simply stated, these types of technologies are certainly the wave of high-throughput analysis in the future. Cytological examination of cells has been around since the development of modern molecular and cellular biology, and these new single cell/systems biology techniques will be a boon to neuroscientists who are looking to identify differences in drug effects and mechanisms of action in vitro.
The group preformed this work within the context of cancer biology with well-described in vitro assays using human cell lines. It will be interesting to see from a neuroscience perspective whether or not neuronal cell lines are as viable in this paradigm. Moreover, the fact that neurons are postmitotic cells with polarized regions may present problems for automation. Furthermore, whether or not in vivo tissue sources can be used in single cell assays using multiple descriptors needs to be examined.
In summary, this technological report is a clear breakthrough for cell-based cytological examinations and phenotypic profiling at the single cell level. It will certainly augment current genomic and proteomic investigations, especially as they relate to drug discovery and assessment of mechanisms of action of specific drug targets.
Comments
Banner Research Institute
This paper represents a significant advance that was achieved by putting together a number of existing concepts to produce what may be considered an approach to a cell biological equivalent of nucleotide or cDNA arrays. The basic design involved culturing HeLa cells in 384-well plates, treating the cells with 13 threefold dilutions of 100 compounds, and then staining cells in wells with one of 11 different fluorescent probes. The authors termed this as “hypothesis-free molecular cytology.” A variety of parameters such as cell size and shape, as well as fluorescent intensity, were quantified automatically. These data were then reduced in a variety of ways, including what the authors termed a titration-invariant similarity score (TISS) in order to characterize effects of the tested compounds on the cell type cultured.
Although the paper is titled as “multidimensional drug profiling” and the choice of 11 probes was “hypothesis-free,” the potential uses of the approach outlined extend far beyond drug profiling based on “hypothesis-free” cytology. It is not difficult to imagine any reasonable number of probes selected to test defined hypotheses regarding a wide variety of cell biological phenomena. Rather than restrict the analysis to one cell type, multiple cell types could be included in the design. And, although the authors emphasized the importance of reducing multiple descriptors of cell phenotype to a single number, one can also envisage examining multiple parameters individually or using other methods of multivariate analysis to characterize results.
Although the use of arrays has provided many valuable insights into tissue responses to manipulation, disease, development, aging, etc., it is, after all, the cellular responses to induced transcription alterations that constitute the ultimate, perhaps even the defining, events. The approach outlined in this paper suggests many ways in which the steps between transcription and cellular phenotype may be better understood.
View all comments by Paul ColemanNathan Kline Institute/NYU Langone School of Medicine
In a recent technological advance published in the journal Science by Z.E. Perlman et al., the group provides convincing evidence of the use of high-throughput cytological profiling by automated microscopy. This system is essentially a combination of microscopy-based single cell analysis with multi-level system biology computation, which enables the automated assessment of cells in culture. The paradigm the researchers have chosen initially as a proof of concept is to evaluate dose-dependent phenotypic changes of specific drug effects. The analysis scheme that the group has developed is really spectacular in the sense that previously automated microscopy has lagged behind other functional genomic and proteomic high-throughput technologies to assess multiple transcripts, peptides, or relevant pieces of information at the cellular level.
The present report uses a method that is devoid of a hypothesis in that these are unbiased measurements that employ individual descriptors as characteristics to identify aspects of a cell’s behavior under experimental conditions. The authors use different dyes, histochemical, and immunocytochemical preparations to identify the nucleus and cytoplasm of HeLa cells plated in culture. This is really a very interesting paradigm because the authors are able to statistically combine, compare, and contrast 93 descriptors leading to approximately 109 data points, which is bewildering considering most cytological examinations are using only single, double, or triple labeling techniques. The data is very convincing and enables the identification of effects of different drugs on these cells in an unbiased fashion, i.e.: the drugs can be blinded, and effects can be assessed at the cellular level without any prior knowledge of the mechanism of action. Simply stated, these types of technologies are certainly the wave of high-throughput analysis in the future. Cytological examination of cells has been around since the development of modern molecular and cellular biology, and these new single cell/systems biology techniques will be a boon to neuroscientists who are looking to identify differences in drug effects and mechanisms of action in vitro.
The group preformed this work within the context of cancer biology with well-described in vitro assays using human cell lines. It will be interesting to see from a neuroscience perspective whether or not neuronal cell lines are as viable in this paradigm. Moreover, the fact that neurons are postmitotic cells with polarized regions may present problems for automation. Furthermore, whether or not in vivo tissue sources can be used in single cell assays using multiple descriptors needs to be examined.
In summary, this technological report is a clear breakthrough for cell-based cytological examinations and phenotypic profiling at the single cell level. It will certainly augment current genomic and proteomic investigations, especially as they relate to drug discovery and assessment of mechanisms of action of specific drug targets.
View all comments by Stephen D. GinsbergMake a Comment
To make a comment you must login or register.