Immune Signatures in Follicular Lymphoma

Dave et al. recently published an interesting paper in the New England Journal of Medicine. However, we had concerns about their conclusions, and these were subsequently published as Correspondence in the same journal. Unfortunately, the editors truncated our letter to such an extent that our point was no longer clear. Therefore, we are posting the complete letter below.

Original letter submitted to NEJM

To the Editor:
Dave et al. (November 18 issue) suggest that survival in follicular lymphoma correlates with two gene signatures, "Immune-response 1" (IR1) and "Immune-response 2" (IR2). Because most of the genes in these signatures were expressed more highly in CD19- cells, they conclude incorrectly that survival is based on features of the nonmalignant cells in the tumors. It must be noted that many of the genes were expressed at only modestly higher levels in CD19- cells. We estimate that the CD19- cells contributed more than 50% of the mRNA for 33 of 43 IR1 genes, but only 2 of 24 IR2 genes (Figure 1). Importantly, when each signature was tested separately, survival in the test set correlated only with the IR2 signature (Rob Tibshirani, private communication). Thus, it is unwarranted to conclude that survival correlates with features in the nonmalignant cells, as claimed in the title of the paper.



Fig 1: Cellular origin of IR1 and IR2 signatures. The heat map shows the fraction of the expression profile arising from CD19- cells, given that an average of 25% of tumor cells are CD19- (Refs. 1 and 2). Yellow bars indicate an estimated average contribution of more than 50% of the mRNA in the tumor samples.

References

1. Lowder J, Meeker T, Campbell M, Garcia C, Gralow J, Miller R, Warnke R, Levy R. Studies on B lymphoid tumors treated with monoclonal anti-idiotype antibodies: correlation with clinical responses. Blood 1987; 69: 199-210.
2. Swerdlow S, Habeshaw J, Murray L, Stansfeld A. T cells in follicular centroblastic/centrocytic (cleaved follicular centre cell) lymphomas at diagnosis, relapse and transformation. Hematol Oncol 1989; 7: 355-363.

Rebuttal

Dave et al. published a response, claiming that we made three errors. First, they state that their CD19- fraction contained only 12.6 percent contamination with follicular lymphoma cells. We assumed no contamination in their CD19- fraction. We simply pointed out that CD19- cells constitute about 25% of the cells in a typical follicular lymphoma biopsy specimen. Thus, more than 50% of the mRNA for a gene in the biopsy specimen would arise from CD19- cells only if the gene is expressed at least 3-fold more highly in CD19- cells compared to CD19+ cells.

Dave et al. claim that we erred in dismissing the importance of the immune-response 1. However, this signature fails to correlate with survival by itself, and must be combined with the immune-response 2 signature in order to show any correlation. Furthermore, Rob Tibshirani has noted that the algorithm used to find these signatures is not robust, raising the distinct possibility that the signatures were found by over-fitting the data. (http://www-stat.stanford.edu/~tibs/FL/report).

Dave et al. state that our third error was a failure to recognize that many of the immune-response signature genes are selectively expressed in T cell, monocytes, or dendritic cells. We do not dispute this fact. We are simply pointing out that “selective” expression of a gene must be more than 3-fold before one can even suggest that mRNA from CD19- cells might correlate with survival in follicular lymphoma patients.

In any case, the conclusion by Dave et al. is testable. The immune-response signatures can be evaluated on an independent set of follicular lymphoma biopsy specimens. We believe that there is a high likelihood that the signatures will fail to predict survival.

Wan-Jen Hong
Gilbert Chu