Profound Impact of Sample Processing Delay on Gene Expression of Multiple Myeloma Plasma Cells

Introduction:

Gene expression profiling (GEP) has significantly contributed to the elucidation of the molecular heterogeneity of multiple myeloma plasma cells (MMPC) and only recently it has been recommended for risk stratification. Prior to GEP MMPC need to be enriched resulting in an inability to immediately freeze bone marrow aspirates or use RNA stabilization reagents. As a result in multi-center MM trials sample processing delay due to shipping may be an important confounder of molecular analyses and risk stratification based on GEP data.  In order to determine the impact of “shipping delay” on MMPC gene expression we analyzed a set of 573 newly diagnosed German MM patients including 230 in-house and 343 shipped samples.

Materials and Methods:

We included publicly available GEP data of newly diagnosed MM patients treated in the GMMG HD4 and MM5 trials. All samples had been processed in a central laboratory in Heidelberg and include 85 HD4 and 145 MM5 in-house and 97 HD4 and 246 MM5 shipped samples. Prediction of sample status was done on publicly available GEP, including data from the UK, UAMS and MMRC. Differential gene expression was assessed using empirical Bayes statistics in linear models for microarray data. Predictor for shipment status was generated on the MM5 cohort using prediction analysis for microarrays. Pathway enrichment analysis was done using WebGestalt. Risk signatures and molecular subgroups were obtained as previously described. Fisher’s exact test was used to compare the subgroup distribution between cohorts. If applicable, results were corrected for multiple testing using the Benjamini-Hochberg method. In all statistical tests, an effect was considered statistically significant if the P-value of its corresponding statistical test was not greater than 5%.

Results:

Applying the Goeman’s global test on the MM5 set showed that “shipping delay” significantly impacted global gene expression (P<0.001). Compared to 145 in-house samples, we detected 3301 down-regulated and 3501 up-regulated genes in 246 shipped samples. For 4280 genes we confirmed differential expression in an independent set of 85 in-house and 97 shipped samples. Of these genes 2040 had a >1.5-fold and 826 a >2-fold difference in expression level. Differentially expressed genes were enriched in processes like ribosome biogenesis, cell cycle, and apoptosis. We observed significantly lower proliferation rates in shipped samples (P<0.001). We did not detect significant differences in the distribution of molecular subgroups between in-house and shipped samples in the combined set of HD4 and MM5. Among GEP based risk predictors the IFM-15 seemed to underestimate high risk in shipped samples, whereas the GEP70 and the EMC-92 gene signatures were more robust. In order to provide a tool to assess the "shipping effect" in public repositories, we generated a 17-gene predictor for shipped samples with a 10-fold cross validation error rate of 0.06 for the training set and an error rate of 0.15 for the validation set. Applying the predictor to further publicly available data sets we detected the "shipping effect" signature in 11% of cases of the UAMS set, 94% of the UK set and 57% of the MMRC set.

Conclusion:

Our study shows that “shipping delay” widely influences gene expression of MMPC with different impact on molecular classification and risk stratification. Based on available data, currently no clear circumvention of the shipping impact on MMPC can be recommended. It should be avoided if possible or at least be taken into account.