Stability of patient-specific features of altered DNA replication timing in xenografts of primary human acute lymphoblastic leukemia

Abstract

Genome-wide DNA replication timing (RT) profiles reflect the global three-dimensional chromosome architecture of cells. They also provide a comprehensive and unique megabase-scale picture of cellular epigenetic state. Thus, normal differentiation involves reproducible changes in RT, and transformation generally perturbs these, although the potential effects of altered RT on the properties of transformed cells remain largely unknown. A major challenge to interrogating these issues in human acute lymphoid leukemia (ALL) is the low proliferative activity of most of the cells, which may be further reduced in cryopreserved samples and difficult to overcome in vitro. In contrast, the ability of many human ALL cell populations to expand when transplanted into highly immunodeficient mice is well documented. To examine the stability of DNA RT profiles of serially passaged xenografts of primary human B- and T-ALL cells, we first devised a method that circumvents the need for bromodeoxyuridine incorporation to distinguish early versus late S-phase cells. Using this and more standard protocols, we found consistently strong retention in xenografts of the original patient-specific RT features. Moreover, in a case in which genomic analyses indicated changing subclonal dynamics in serial passages, the RT profiles tracked concordantly. These results indicate that DNA RT is a relatively stable feature of human ALLs propagated in immunodeficient mice. In addition, they suggest the power of this approach for future interrogation of the origin and consequences of altered DNA RT in ALL.

Figure 1. RT profiles from ALL PDXs

A. Outline of a genome-wide RT assay. BrdU labeling of nascent DNA (E/L) is preferred whenever cells are actively proliferating. However, RT profiles can also be obtained on patient samples that have lost metabolic activity but retain their S-phase DNA content based on an analysis of their DNA copy number differences (S/G1). B. Comparison of log₂ ratios for both methods applied to the same spleen cells derived from a B-ALL PDX. Although the S/G1 method gives a dynamic range of <2-fold vs. up to >1,000-fold from the E/L method, the profiles are comparable. In the bottom panel, data from all 3 methods were quantile normalized to each other. C. Exemplary scaled and normalized plot of chromosome 1 from the same samples as in B.

Figure 2. Preservation of RT profiles in PDXs

A. Representative loci that showed patient-specific RT differences. RT of DNA from each of 3 primary patient samples and their PDXs has been conserved at the loci shown. B. Genome-wide Pearson correlation matrix using 204-kb windows. Cell types are color-coded as B-ALL (purple), embryonic stem (ES) cells (gray), and normal T- or B-cells (orange).

Figure 3. RT signatures of patients are preserved in PDX

K-means and hierarchical clustering of initial ALL samples, derived PDXs and multiple samples of non-leukemic cells (A), and the same analysis of patient and their PDX samples only (B). RT variable 204-kb windows were defined as early replicated in at least one sample (RT log₂ ratio ≥0.3) and late replicated in at least one other sample (RT log₂ ratio ≤-0.3) and processed by cluster analysis. The percentage of autosomal DNA segments that showed significant variation in RT is shown. Dendrograms were constructed based on the correlation values between distinct cell types (distance = correlation value -1). A correlation threshold of >0.6 was used to color label the major branches of the dendrograms. The distinct RT signatures identified are indicated by numbered grey boxes. Method of RT profiling is indicated (E/L vs. S/G1). Since the single T-ALL patient sample exhibits the most significantly different RT differences as compared to the BALL samples, most of the K-means clusters (RT signatures) are unique in the T-ALL patient sample. Figure 4 shows a clustering excluding this sample.

Figure 4. RT signatures distinguishing B-ALL patients are also preserved in PDX

A. Similar analysis as shown in Figure 3 except that 60kb RT variable regions were defined as the top 10% regions of standard deviation amongst only the B-ALL patients and their PDX and only those patients and PDX were subjected to K-means and hierarchical clustering analysis. Clusters 3,5, and 6 are specific for patient 11-064. Clusters 7 and 9 distinguish the clonal architecture of patient 4 (see Figure 5). Cluster 10 is shared between 11-064 and Case21. B. Exemplary plots of selected RT signature features (chromosomal segments or rows in Fig 4A) with the K-means cluster indicated in each panel. As expected, this unbiased analysis identified the regions shown in Figure 2A that were detected by visual inspection (panels i and ii). Panels vii and viii show signature features that distinguish the different clones in Patient 4 (P4) and those P4 profiles are extracted and shown separately below each panel. Note that examples of features that are significantly different for one particular patient are shown (panels iii, iv, vi, ix) with the outlier patient indicated but most RT features are variable between all patients. All features distinguish some patients from others while each patient has a unique pattern within each RT signature.

Figure 5. Changes in RT signatures in a serially passaged PDX mirror changes in genomically defined subclonal differences

(A) Exemplary chromosomal region where B-ALL #4 RT matched the 2^nd, but not the 1^st and 3^rd, serial PDX passage. (B) Two autosomal CNVs identified in the RT data of the original ALL cells were found to be present in the 2^nd, but not the first or third passage. PDX. Note that these CNVs did not result in a change in RT that would be detected as a significant RT variation but rather served as a genetic marker to track the alternating outgrowth of different subclones in sequential passages.