Methods for Sample Acquisition and Processing of Serial Blood and Tumor Biopsies for Multicenter Diffuse Large B-cell Lymphoma Clinical Trials

Biopsies

Of the 40 patients enrolled in the trial, 27 (67.5%) underwent a pretreatment biopsy, and 14 patients (35%) also underwent the day 15 biopsy. Most biopsies were performed by an interventional radiologist using ultrasound guided techniques. No serious adverse events occurred in relation to the biopsies. However, 3 patients had biopsy-related bleeding, all of which were self-limited; 2 were minor bleeds not requiring intervention, and one bleed required medical intervention (grade 1 and grade 2, respectively, by the Common Terminology Criteria for Adverse Events version 4; ref. 7).

Biopsies were all processed at the central processing laboratory (JGH), where tumor B cells were separated and nucleic acids extracted. We assessed whether the quality of biomaterial differed depending upon the research site where collection was performed and by transport to the central processing laboratory. We isolated fewer tumor cells from samples collected at non-JGH sites (see Table 1). This may be attributed to the biopsy system (8) or to shipping. The JGH used larger (16 gauge) needles for biopsy collection and recovered more cells than other participating sites (4.75 vs. 1.98 million, P = 0.05). The time from tissue acquisition to tissue processing was also significantly shorter for JGH biopsies compared with other sites (27 minutes vs. 23 hours). There were only four shipped biopsies in which the temperature indicator tag showed exposure above 10°C. While these tended to have lower quality by the 260/280 ratio, it was not significant (1.875 vs. 1.788; P = 0.0889). Despite the decreased cell number, there was no difference in the DNA quantity or quality between research sites (Table 1). In contrast, we recovered significantly less RNA when biopsies were shipped from non-JGH research sites. However, while we recovered less RNA, it was of excellent quality with a RIN average of 9. Furthermore, nucleic acids isolated from day 15 biopsies were also of excellent quality, independent of collection site. Thus, we conclude that high-quality tumor-derived nucleic acids were isolated from all enrollment sites.

Blood samples

We collected pretreatment blood samples from all enrolled patients (N = 40) and day 15 blood samples in all but 1 patient, who went off study due to a drug-related adverse event. The yield and quality of DNA isolated from PBMCs were comparable among sites (Table 1), but the yield of RNA was greater when the blood did not require shipping. Despite this, the RNA quality was high.

Immunohistochemistry

The first NCB was preserved in formalin immediately upon procurement from the patient, embedded, and subsequently used for immunohistochemical analyses at the central processing laboratory. In our experience, the core obtained from the first pass provides the best tissue integrity, important for histologic assessment. Representative images of H&E stains and IHC (stained with anti-CD20) are shown (Fig. 2A). We found that all biopsies, regardless of collection site, were of high quality. Thus, shipping did not affect the quality of fixed tissue.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Examples from downstream applications. A, representative H&Es and anti-CD20 IHC are shown for one patient at the site of the central processing laboratory (JGH) and one from an outside hospital. B, metabolites from two samples were run 1 year apart and compared by Pearson coefficient.

Sequencing and microarrays

Of the 27 patients with pretreatment biopsies in this study, we submitted DNA on 26 tumor/germline pairs (n = 52 samples) for exome sequencing, in which 1 patient was excluded because of low DNA yield from the tumor biopsy. For the first 13 patient pairs, libraries were constructed using TruSeq in pools of 4 to 5 samples. Unfortunately, due to technical difficulties unrelated to DNA quality, this resulted in an imbalanced distribution of reads among the individual samples. The second batch of libraries was prepared using the Nextera Rapid Human Exome Kit and yielded more balanced libraries. Overall, we obtained at least 30× redundant sequence coverage across targeted regions for 41 of 52 exomes prepared from these samples. The average coverage achieved across all exomes in targeted regions was 71.5×. As a measure of quality, we assessed the number of duplicate reads in the libraries prepared using the Nextera Kit. When duplicate reads represent less than 5% of the library, this indicates high quality. Of these 26 libraries, only one had higher than 5% duplicate reads (6.7%). There was no significant difference between the libraries derived from blood and biopsy DNA (3.94% vs. 3.98%, P = 0.8841). Furthermore, there was no difference in the quality of libraries derived from JGH or non-JGH hospitals (4.15% vs. 3.80%, P = 0.2131).

For cDNA microarray analyses, 0.2 μg of RNA is required. Two pretreatment biopsies (both from non-JGH) did not yield enough RNA. We found that both the Agilent and Affymetrix microarrays yielded high-quality data. We used Agilent Feature Extraction Software V10.7.3.1 for 1 Color Gene Expression to assess the overall data quality. The GE1_QCMT_Sep09 thresholds were used for quality control criteria of the microarray data. Any arrays with values outside of the threshold ranges were further evaluated (see Supplementary Table S1). Only one sample showed poor results, with values outside of the acceptable range for 8 of the 10 metrics reported. Microarrays using RNA extracted from biopsies collected at JGH and non-JGH sites had equal quality (see Supplementary Table S1).

Metabolomics

All samples passed the technical quality control criteria and were normalized in terms of raw area counts. Each raw data metabolite was then rescaled to set the median equal to 1 and missing values were imputed with minimum values to allow inter sample comparisons. A total of 605 metabolites were identified and quantified in all the samples (389 named and 216 unnamed metabolites). To estimate metabolite measurement variability from sample storage conditions and batch effects, we processed two sets of duplicate samples in two runs, 1-year apart, and found excellent correlation (Pearson correlation R = 0.9; Fig. 2B).