CD20, encoded by MS4A1, is a cell surface glycoprotein involved in calcium signaling, 1 expressed on normal and malignant B cells. Antibodies targeting CD20 (for example, rituximab) are used for the treatment of B-cell lymphoma, and mediate antitumor response by complement-dependent cytotoxicity (CDC), antibodydependent cellular cytotoxicity and direct induction of apoptosis. 2 Resistance to anti-CD20 antibodies remains a clinical challenge. Transcription factors like SPI1/PU. 1, 3 Oct2 (ref. 4) and Pax5 (ref. 5) have been described in MS4A1 regulation, however, other factors (eg, epigenetic changes, somatic mutations, noncoding RNA) could also affect the CD20 expression levels. To identify genes regulating CD20 surface expression, we performed a genome-wide RNAi screen using a library of barcoded small hairpin RNAs (shRNAs) delivered by lentiviral vectors. In total, the library targeted 15377 messenger RNAs (mRNAs) with 5–6 shRNAs per transcript in three separate modules. We chose the Raji cell line (Burkitt's lymphoma) characterized by intermediate CD20 levels (Supplementary Figure S1). After 6 days, infected cells were fluorescenceactivated cell sorting (FACS)-sorted according to CD20 expression generating CD20low and CD20high populations (Figure 1a, Supplementary Figure S2), and barcodes were sequenced (Supplementary Table S1). shRNA barcode counts were highly reproducible between replicates (Pearson correlation 0.91–0.96—for unsorted, 0.71–0.86—for sorted populations; Supplementary Figure S3). After filtering based on expression, candidate genes were identified using the weighted z-score method (Supplementary Materials and Methods, Supplementary Tables S2–S4). The positive control MS4A1 was ranked first in the CD20low population (Figures 1b and c). We identified two transcription factors that have previously been described as MS4A1 regulators: PAX5 (rank 11) 5 and SPI1 (rank 16). 3 To select novel candidates, we applied a weighted z-score threshold of 3.5, which yielded 37 candidate CD20 repressors and 51 potential activators across the three modules (Figure 1b; Supplementary Figures S3–S5). We chose 17 candidate regulators and 10 control genes for validation. Although 16 out of 20 tested shRNAs targeting controls did not change CD20 levels by> 15%, we observed a consistent decrease or increase of CD20 expression for 16 of 17 candidates, with at least 2 shRNAs (Figure 1c, Supplementary Figure S6). These data support the performance of the RNAi pipeline. CREM (cAMP-responsive element modulator) was identified as a top candidate for CD20 suppression (Figures 1b and c). Three nonoverlapping shRNAs targeting CREM efficiently decreased CREM mRNA and protein levels, and resulted in upregulation of surface CD20 levels (Supplementary Figures S6 and S7) across 6 out of 7 lymphoma cell lines (Supplementary Figure S8). To assess the specificity of CREM knockdown, we examined additional B-cell surface receptors (CD81, CD79b, CD19, CD22, CD38 and CD72). Only the level of CD72 expression also increased after CREM knockdown, albeit to a lower extent (Supplementary Figure S7c) suggesting a specific role of CREM for CD20 regulation. As the promoter region of CD20 contains three half-cAMP response elements sites, which were previously reported as binding motives for CREM in primary T cells6 (Supplementary Figure S9a), we investigated CREM-binding capacity to the CD20 promoter. We used two different CREM antibodies and SPI1 antibodies as a positive control. 3 Our ChIP-PCR showed that CREM binds the MS4A1 promoter region (Supplementary …