What is "cell lineage tree reconstruction"?
Since all living organisms originate from a single cell (the fertilized egg), and grow by a series of cell divisions, the entire development of the organism can be represented by a binary tree. Analysis of the shape of this tree can uncover a vast amount of biological data pertaining to the development of the organism. We previously developed a method for cell lineage tree reconstruction, based on phylogenetic-like analysis of random mutations which accumulate in all cells during replication of DNA (see "Genomic variability within an organism exposes its cell lineage tree").

How does your method work?
The method proceeds by a series of sequential steps:

1. Individual cells are extracted from an organism or tissue
2. Genomic DNA is extracted from each cell
3. Each cell's DNA is analyzed at ~100 genomic positions, and mutations are recorded.
4. A phylogenetic-like algorithm analyzes the degree of similarity in the mutations of different cells and reconstructs a cell lineage tree.

What kind of information can be extracted from a cell lineage tree?
Analysis of the shape of a cell lineage tree can uncover different types of data related to the growth of the organism or tissue, such as detection of clonal relationships between cells, estimation of the numbers of stem cells, determination of the numbers of cell divisions, and uncovering of cell population dynamics.

What kind of information can not be extracted from a cell lineage tree?
Cell lineage reconstruction is a retrospective method, and as such it can infer the past existence of cells, but it cannot provide data regarding these cells phenotype or physical location within the organism or tissue.

Technical details:

Did the mouse tumor develop spontaneously?
The mouse used for the experiment was genetically modified by "knocking out" of a key gene involved in DNA repair. As a result of their DNA repair deficiency, such mice have high mutation rates and are pre-disposed to develop a variety of tumors, such as lymphomas, as well as gastro-intestinal and skin cancers. However, despite of the fact that cancer is inevitable in these mice, each tumor develops randomly, and it is not possible to determine beforehand the type, location, and timing of the tumor that will develop in a particular mouse. Therefore, the tumor that was analyzed can be thought of as simulating a spontaneous tumor (in this case, a lymphoma). 

Why did you analyze only 37 cells?
Each cell was individually isolated from the tissue, and processed further for DNA extraction and genome amplification. In addition, for each cell, 120 genomic positions were analyzed, most of which contain two alleles. Analysis of a greater number of cells requires greater resources, but may be beneficial since as the number of cells analyzed increases, more information can be extracted from the reconstructed tree, and the statistical significance of the results increases.

Can it be beneficial to analyze a greater number of genomic positions?
Yes. The amount of data increases with the amount of genomic positions that are analyzed per cell, resulting in a reconstructed tree with greater precision and confidence level. 

Was some data lost during preparation of the cell samples?
Yes. On average, about 38.5% of the genomic positions analyzed successfully. The remaining 61.5% of genomic positions suffered from "drop out" of the signal, as a result of truncation of cell nuclei during tissue preparation, and further loss of genetic material during the in-vitro duplication step.

Results and future applications:

Is the shape of the reconstructed tumor tree surprising?
Analysis of the reconstructed tumor tree revealed several aspects related to the growth of the tumor. Some of these aspects are in line with currently accepted models (such as the monoclonal origin of the tumor), whereas others represent new types of data that were not previously obtainable (for example, the "depth", or number of cell divisions, in cancer cells and in lung epithelial cells).  

Why are the tumor cells "deeper" than the normal cells?
The mean depth of cancer cells was 236 cell divisions, while the mean depth of normal lung epithelial cells was only 121 cell divisions. However, this large and significant difference in depth is not likely related to the cancerous vs. normal phenotype, but rather is likely to reflect the slower turnover rate of lung epithelial cells in relation to lymphocytes.

Can the method described in the article be applied to study human cancers?
The method can potentially be applied to investigate human tumors that are surgically removed from patients. However, since humans are not genetically modified, the rate of DNA mutations is likely to be smaller in most human cancers. The smaller rate of mutations can be overcome by several strategies, for example, by analyzing a greater number of genomic positions in each cell, or by analyzing special genomic positions with higher mutation rates.

What questions in cancer research do you intend to address with your method?
We intend to apply our method for cancer cell lineage tree reconstruction to answer key questions in human cancers, for example:

1. At what stage does metastasis occur?
2. Can the depth of tumor cells serve as a prognostic marker for cancer severity?