Impact of Tissue Shipping on Plasma Cell Isolation, Viability, and RNA Integrity in the Context of a Centralized Good Laboratory Practice Certified Tissue Banking Facility

Abstract
The Multiple Myeloma Research Consortium (MMRC) has established a tissue bank for the deposition of bone marrow (BM) samples from patients with multiple myeloma (MM) to be mailed and processed under good laboratory practices (GLP). We performed a pilot study on 14 samples drawn at Mayo Clinic Rochester (MCR). Individual samples were pooled and split into 2 equal aliquots. One-half of each sample was processed following GLP compliant standard operating procedures (SOPs) to isolate the CD138+ cells, immediately after procurement, at MCR. The CD138+ cells were stored at -80 in TRIZOL. The other half of the aspirate was sent overnight to Mayo Clinic Scottsdale (MCS) where they were processed using identical SOP’s. The RNA was extracted and analyzed in a single batch at MCR. Samples were assayed for the following quality determinants: Viability was assessed in 9 of the 14 samples using a 3-color flow cytometric method (CD45, CD38 and 7AAD). Post sort purities were checked and demonstrated a slightly higher percentage in the immediately processed samples. RNA recovery and integrity was assessed using the Agilent Bioanalyzer with ample quality RNA available for gene expression profiles (GEP) in 50% of the samples. GEPs were compared to determine the signature emanating from the shipment of samples. Samples processed at 24 hours showed fifty-one probe sets up-regulated and 31 down-regulated. These genes were enriched for gene ontology (GO) processes that represent cellular response to environmental stress. In aggregate our results suggest that shipment of samples did not significantly affect these quality determinants.

Introduction
In modern clinical drug development, validating the efficacy of therapeutic agents entails conducting randomized trials. These trials require the participation of a large number of patients to ensure adequate statistical power for meaningful conclusions. As such, they are usually conducted in the setting of multi-center, and sometimes international, cooperative groups. Additionally, the need for obtaining high quality purified tumor cells for ancillary research studies from these trials is ever increasing. To this end, the Multiple Myeloma Research Consortium (MMRC) has established a tissue bank to collect bone marrow from patients with multiple myeloma. To assure the quality of these samples we have implemented standard operating procedures (SOP) for the collection, isolation and storage of the tumor cells as well as striving towards Good Laboratory Practices (GLP) standards. This has led to an internal quality assurance unit (QAU) and an information technology infrastructure for the registration, the sample tracking and the sample attributes.

Unlike many other hematological malignancies the tumor cells from multiple myeloma patients coexist with many normal hematopoietic cells in the bone marrow and thus need to be purified using some type of selection method. Our experience with shipped samples within the context of the Eastern Cooperative Oncology Group (ECOG) has demonstrated the ability to obtain good plasma cell recoveries from mailed bone marrow aspirates. Using these samples we have demonstrated the use of slides for FISH, DNA for mutation analysis, and RNA for GEP. However, these studies did not compare results with a split sample being processed immediately. The RNA in particular was of the most concern due to its instability. To directly address this question, a pilot project was set up to compare the isolation of plasma cells immediately after bone marrow aspiration versus having the aspirate shipped priority overnight at 4 degrees and processed the following day with particular focus on the RNA and the subsequent expression profile.

MATERIALS AND METHODS
Samples
Bone marrow aspirates from 14 patients were obtained after informed written consent and collected into tubes containing ACD as the anticoagulant. The bone marrow was pooled and gently mixed to ensure uniform cell distribution.

To limit the number of variables in this study and yet mimic mailing conditions we decided to use one other site, Mayo Clinic Rochester (MCR). MCR was supplied with and utilized the same SOPs as Mayo Clinic Arizona (MCA) and has years of experience with these plasma cell separation methods. We designed the study the following way: MCR pooled the bone marrow aspirate (up to 40 cc), split in half, processed one half of it immediately and shipped the other half to the MMRC tissue bank located at MCA. Post plasma cell purification samples were placed in TRIZOL, frozen and the MCA samples were sent back to MCR on dry ice. MCR then extracted the RNA from the TRIZOL and performed both quantitative and integrity analysis. This analysis included the total RNA recovered as determined by spectrophotometer readings and the quality of the RNA determined by the 28S/18S ratio ascertained from the Agilent Bioanalyzer. Gene expression profiling was performed on all samples where sufficient RNA quantities and quality were available. Slides from the CD138+ selected cells were also sent to MCR for plasma cell purity analysis. This was done to avoid possible variability among readers. All other quantitative and qualitative determinants were performed at the respective sites, which included the following: recovery of CD138 positive plasma cells as well as viability of both the whole bone marrow and the CD138 selected fractions (Figure 1).

Flow Cytometry
The percentage of live, apoptotic, and dead cells were determined on ACK lysed whole bone marrow and CD138+ cell fraction by flow cytometry. These parameters were determined using a three-color apoptosis assay as previously described. Briefly, cells were stained using CD 45 conjugated to fluorescein isothiocyanate (CD45-FITC Becton Dickinson, Mountain View, CA), CD 38 conjugated with Phycoerythrin (CD38-PE Becton Dickinson, Mountain View, CA) to identify the plasma cells(45-/dim38++) and 7AAD to identify the apoptotic/dead fractions. All samples were run using the BD FacScan flow cytometer (Becton Dickinson) and the data was analyzed using the Cell Quest software program (Becton Dickinson). Regions were drawn to identify the percentage of cells in each of the three possible populations; alive, dead or apoptotic. Plasma cells which were negative for 7-AAD were considered alive as the membranes were intact enough to exclude the dye; cells that were bright 7-AAD positive were considered dead and very permeable to the dye; cells undergoing apoptosis had 7-AAD staining between these two values. The percentage of each fraction was calculated by the software program.

Plasma Cell Isolation
Plasma cells were isolated from the whole bone marrow utilizing the immunomagnetic bead selection. We used a monoclonal mouse antihuman CD138+ antibody micro beads and the AutoMACS cell separator (Miltenyi Biotech, Auburn, CA). The red blood cells were lysed using an ammonium chloride lysing (ACK) procedure, and cells counts were performed using a Coulter counter. Antibody bead conjugates were incubated and the cells were washed using PBS containing 2% BSA and 1 mM of EDTA (bead buffer). Cells were re-suspended in 4 ml of bead buffer and the sample loaded onto the AutoMACS. We used POSSELDS on all separations. This program utilizes two columns which increases the plasma cell purity at the expense of some plasma cell loss. The cells from the POS 2 port were removed, counted, aliquoted for flow cytometry and the remainder placed in TRIZOL at a concentration no greater then 10 million/ml and frozen at –80 until all samples were collected.

Post-sort purity check
The purity of all sorts was confirmed using a 3 color immunofluorescent slide based method. Approximately 10,000 cells were removed from the positive fraction after sorting and spun onto a slide using a cytospin centrifuge. The slides were allowed to air dry. A circle was drawn around the cells using a Super PAP Pen (The Binding Site, SanDiego, CA), dried and placed into a coplin jar containing 95% ethanol for 5 minutes. The slides were removed, dried and placed into a new coplin jar with APK wash solution (Ventana Medical Systems, Tucson, AZ). The slides were removed and dried. 100 ul of antibody mix containing 10ul anti Kappa-AMCA, 10ul Anti lambda-FITC and 80ul of RPMI containing 10% FCS was added to each slide and incubated in the dark for 30 minutes at room temp. The slides were washed 3 times (3 minutes/wash) by placing the slides in a coplin jar with APK and gentle agitation. After the last wash, the slides were, air dried, 10ul of antifade with PI (Vector Labs) added, cover slip added and then 100 cells were scored using a fluorescent microscope and a triple pass filter. The percentage of FITC, AMCA and PI positive only cells are recorded and checked against the known isotype to ensure quality of the sort.

RNA Purification and Integrity Assessment
RNA was isolated from the TRIZOL using a chloroform extraction protocol. Briefly, TRIZOL samples were homogenized using a 20 gauge needle, chloroform added and the tubes centrifuged. The aqueous phase, containing the RNA, was removed and isopropyl alcohol was added to precipitate the RNA. The RNA pellet was washed with 75% ethanol and the pellet was allowed to dry. The dried RNA pellet was then suspended in RNAse free water. The RNA was further "cleaned-up" using the Qiagen RNAeasy columns. The concentration of the RNA was determined by using a ratio of the nucleic acid absorbance at 260nm (A260) to the protein with the absorbance at 280nm (A280) on the spectrophotometer. Additionally, the RNA integrity was assessed on the Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA) using the 28S/18S ratio. High quality total RNA samples have distinct features which include the 18S and 28S ribosomal peaks. The 28S:18S ratio should be > 1.0 for successful GEP. Also, the base line between 29 seconds and the 18S ribosome should be relatively flat and free of small round peaks. Lastly, the baseline between the two ribosomal peaks should not have well defined peaks.

Gene Expression Profiling
Gene expression analysis was performed on RNA from CD138+ selected plasma cells using the Affymetrix U133A chip (Affymetrix, Santa Clara, CA). Microarray hybridization has been previously reported. Probe level data was normalized using the Affymetrix Microarray Suite 5.0 algorithm. Gene expression intensity values were log transformed, median-centered and analyzed using GeneSpring 7.3.1 GX (Agilent Technologies, Palo Alto, CA). For quality of the gene expression data, we assessed the GAPDH 3'/5' ratio and the Actin 3'/5' ratio. Samples with poor quality results were identified if the ratios were greater than 1.25 and greater than 3.0 respectively. The comparability of the gene expression data of the paired samples was assessed in several ways. First the raw expression values of the pairs were plotted on an X-Y plot to check for concordance. Second, the samples were clustered in an unsupervised manner using genes that varied across the individual samples. These genes were first identified by Welch ANOVA using variance computed by applying the Cross-Gene Error Model based on Deviation from 1 available within GeneSpring. This overcomes the lack of replicates and variance associated with the individual samples and is similar in principle to variance filtering. Unsupervised clustering was performed using the hierarchical agglomerative algorithm. Pearson's correlation coefficient and centroid linkage were used as similarity and linkage methods respectively. To detect possible differences between samples processed immediately and those that were shipped, we extracted genes that had 1.5 fold difference in expression and were statistically significant at a corrected p–value of 0.05 by student's t-test with Benjamini-Hochberg multiple testing corrections. These differentially expressed genes were then assessed for gene ontology (GO) enrichment using Genespring.

RESULTS
Despite some variability, there were no consistent or significant differences in the quality determinants when the two sets of samples were compared (Figure 2 and Table 1). The Cell viability, as determined by a three-color flow cytometric assay, was evaluated in 9 of 14 paired samples and was similar in both sets of samples (Figure 2a). The majority of the cases demonstrated equivalent percentages of cells in the apoptotic/dead fraction. In 3 of 9 samples the apoptotic fraction was higher in the mailed in samples while 5 of 9 samples from the immediately processed samples were higher and 1 of 9 were equivalent. The Median apoptotic fraction for the immediately processed sample was 17.4% with a range from 2.1% to 29.1 % as compared to the mailed in samples with a median of 8.0% and a range of 3.7 to 62.8% (p–value 1.0). The whole bone marrow samples showed a similar pattern of apoptotic/dead cells as the plasma cell fraction (data not shown). We were able to collect highly enriched plasma cell population (Figure 2b). The mean percentage of purity was 95.5% with a range between 75 and 100 % for the immediately processed samples and 89% with a range of 56 to 99 % for the shipped samples (p–value = 0.05). In the majority of the samples the percent plasma cells were equivalent. There were 5 samples which showed some discrepancy. Of the immediately processed samples 3 of these had better purity while 1 of the 12 mailed in samples demonstrated a higher purity. The plasma cell yield was slightly lower in the shipped samples (Figure 2C) with a median of 4.3 million on the immediately processed samples as compared to 2.2 million on the mailed samples. The difference in recoveries may be possibly due to some shipment-associated apoptosis and subsequent loss of cell surface CD138 antigen. However, as mentioned, the purity of the samples shipped was very similar to that of the locally processed ones. Subjective analysis of the RNA was similar between both groups. The median RNA recovery normalized to ug/million cells was 5 ug (range of 2.1 to 9.7) for the MCR processed samples and 4.6 ug (range of 0.8 to 10.3) for the MCS samples (p–value = 0.33) (Figure 2D). There was no evidence of additional degradation in the shipped samples after purification when the 28S/18S ratios were compared (data not shown). Adequate RNA, a minimum of 3 ug of total RNA at a concentration of at least 375 ug/ul and a 28S/18S ratio greater than 1.4, was available for GEP studies in 7 paired samples. The other 7 pairs failed our quality control metrics due to either degraded RNA (5 pairs) or an insufficient amount of RNA (2 pairs). All 14 GEP results passed quality control as assessed by GAPDH and actin 3'/5' ratios (minimum values of 1.25 and 3.0 respectively). The concordance in the GEP results between the paired samples is also extremely good. Five of the pairs demonstrated more than 99% of genes with expression within 2–fold of each other. On unsupervised clustering, all the paired samples were clustered next to each other with very similar expression profiles, although in several pairs, a small cluster of genes were over expressed in the samples that were processed 24 hours after aspiration (Figure 3A). No differentially expressed genes between samples processed immediately and after 24 hours passed our statistical filter but this could be attributed to relatively low sample size. Eighty-two probe-sets have a 1.5 fold difference in expression level between those samples processed immediately and 24 hours later. Fifty-one probe sets were up-regulated and 31 down-regulated in the samples processed at 24 hours. (Figure 3B) Interestingly, these genes are enriched for gene ontology (GO) processes that represent cellular response to environmental stress (Table 1).

DISCUSSION
When establishing a biorepository one must not only meet the needs of the immediate studies at hand but also have the foresight for what may need to be stored for projects at a much later time. There are many variables one must consider in establishing a tissue bank. Some of these include the harvest of the tumor, sample processing, and storage of the samples. We have previously demonstrated that we were able to utilize slides from mailed in tissue for IHC/IF as well as FISH studies. Decisions about what storage media to use for future isolation of RNA, DNA, Protein, can also be problematic. We chose for our tissue bank to store the tumor cells in TRIZOL reagent. This has allowed us to obtain high quality RNA, even in shipped samples, as demonstrated here. We were also able to obtain high quality DNA for aCGH and sequencing with very good results. Others have previously shown the use of protein from TRIZOL extractions for western blots.

There are not many studies published that have looked at RNA stability and shipment of cells isolated from bone marrow. These issues were highlighted in a recent pharmacogenomics study involving patients entered into multi-center international trials for bortezomib, a recent FDA approved treatment for multiple myeloma. There was significant attrition at various steps along the way from consenting patients, obtaining samples to various quality control measures (RNA quality, microarray hybridization quality and contamination issues) such that on average only 34% of collected samples have analyzable gene expression data. In these studies, samples are collected and processed (isolation of malignant cell of interest by negative selection) at the individual collection sites before being shipped to a central laboratory for gene expression studies. Of note, in their study the greatest attrition occurs at the step of RNA quality (ranging from 37% to 65%). Furthermore this attrition appears worse in the international studies compared to the US-only studies (62% attrition versus 39% attrition). This suggests that one of the main reasons for not being able to obtain good quality RNA may relate to heterogeneity in sample collection and processing procedure or the need to transport samples. These issues are extremely important in the current age of large-scale multi-center and often international drug trials.

In this study, we focused on the feasibility of utilizing a single centralized processing and storage center versus multiple local centers in regards to the quality of the overall specimen with emphasis on the RNA as transcriptional profiling is integral to pharmacogenomics study towards individualized therapy. The ability to use a centralized bank would have significant cost savings and quality assurance issues would be much more manageable. However, all of these advantages would quickly be diminished if the quantity and quality of the tumor were such that it rendered it unusable. We therefore wanted to compare the quality of the RNA from the tumor cells both prior to and after shipping and assess the impact on gene expression profiling. In doing so, we investigated the viability and recovery of the plasma cells and the quantity as well as quality of the RNA. We did not see any consistent differences in any of these parameters between samples processed within a few hours of the aspirate to those shipped at 4 degrees overnight. When assessing the pairs, >90% of genes have expression within 1.5 fold of each other. In fact only 1% of probe sets have on average more than 2 fold differences in expression between samples processed immediately and shipped. Our results are in contrast with a previous publication, which found that only 8.5% of probes never exceeded 1.5 fold in all experiments and 38.4% of probes have greater than 2 fold difference in expression. One major difference is that in our study, only RNA from CD138 positive malignant plasma cells, as opposed to unselected bone marrow cells, are isolated for gene expression studies. It is conceivable that transcriptional programs of plasma cells or malignant cells are less prone to fluctuations or environmental changes. However, despite having a much lower number of genes with significant changes in gene expression, the genes whose expression does change are mainly involved in stress response as corroborated by the gene ontology analysis. This result is similar to that reported in a previous study. More importantly, the gene expression profile was not altered in a biologically relevant manner as the paired samples essentially cluster together and also express genes relevant to their biology in similar manner such that their molecular classification is not affected.

Further analysis of RNA recoveries revealed that samples where we recovered less than 2 million plasma cells rendered less than ideal amounts of RNA and the quality of this RNA was more likely to be degraded. This is demonstrated in fig 2D where the greatest differences (ug/million PC) is observed in sample pairs 8-14. Samples 8-14 started with fewer plasma cells in TRIZOL then samples 1-7 (median 1.5 million [range 0.8-9.2] versus median 7.4 million [range 2.0-19.2]). Additionally, the RNA was degraded in 5 of 8 of these samples. Although we performed a flow cytometric apoptosis assay the number of samples were too small and the variability to great to draw any good conclusions as to whether or not we could predict these RNA failures. These results suggest that the quantization of apoptosis with this assay be performed on fresh samples. From these observations and additional testing, we have determined samples yielding less then 2 million plasma cells are less likely to yield enough high quality RNA for expression profiling (regardless of whether or not they are processed immediately or not). Within the context of the MMRC biobank we have had very good results with samples where plasma cell recoveries were greater than 2 million plasma cells. To date we have extracted quality RNA/DNA with successful profiles in approximately 85% of these cases.

Bio-banking is becoming an important tool for assisting in the categorization of disease, prognostic markers, as well as potential drug discovery leading to personalized medical care. Studies such as this should be done to insure the quality of what is put into a bank. The MMRC tissue bank collects and stores peripheral blood mononuclear cells, plasma from both the blood and the bone marrow, ammonium chloride lysed whole bone marrow, cytospin slides and CD138 positive and negative cells. To insure the MMRC would have the best tissue bank for multiple myeloma we have put together approximately 70 standardized operating procedures for the shipment, receipt, isolation and storage of these samples. In this study we affirmed that a centralized bank could obtain and isolate tumor cells without a substantial reduction in RNA quality or gene expression profiling outcome.

References

Legend

Figure 1. Schematic representation of study design

Figure 2. Comparison between (A) apoptosis, (B) plasma cell purity, (C) plasma cells recovered and (D) RNA recovered per million plasma cells in samples processed immediately at MCR and those shipped and processed 24 hours later in MCS

Figure 3. Gene expression profiling of paired samples with adequate good quality RNA. (A) The paired samples clustered together when unsupervised clustering of samples was performed using genes variably expressed across the dataset, suggesting that changes in gene expression, does not significantly affect overall gene expression profile and samples essentially cluster according to biology. Red indicates over-expressed genes, blue under-expressed genes and white genes with median expression. (B) 82 probe-sets were differentially expressed between samples processed immediately (Rochester) and 24 hrs later after shipping (Scottsdale).

Figure 1

Figure 2

Figure 3a

Figure 3b

Table 1. Comparison of parameters between samples processed immediately at Rochester and those shipped to Scottsdale

Table 2. GO enrichment analysis for differentially expressed genes between samples processed immediately and after 24 hrs