Precise reconstruction of the TME using bulk RNA-seq and a machine learning algorithm trained on artificial transcriptomes.

Cellular deconvolution algorithms virtually reconstruct tissue composition by analyzing the gene expression of complex tissues. We present the decision tree machine learning algorithm, Kassandra, trained on a broad collection of >9,400 tissue and blood sorted cell RNA profiles incorporated into millions of artificial transcriptomes to accurately reconstruct the tumor microenvironment (TME). Bioinformatics correction for technical and biological variability, aberrant cancer cell expression inclusion, and accurate quantification and normalization of transcript expression increased Kassandra stability and robustness. Performance was validated on 4,000 H&E slides and 1,000 tissues by comparison with cytometric, immunohistochemical, or single-cell RNA-seq measurements. Kassandra accurately deconvolved TME elements, showing the role of these populations in tumor pathogenesis and other biological processes. Digital TME reconstruction revealed that the presence of PD-1-positive CD8+ T cells strongly correlated with immunotherapy response and increased the predictive potential of established biomarkers, indicating that Kassandra could potentially be utilized in future clinical applications.

Distinct Requirements for HIV-1 Accessory Proteins during Cell Coculture and Cell-Free Infection.

The role of accessory proteins during cell-to-cell transmission of HIV-1 has not been explicitly defined. In part, this is related to difficulties in measuring virus replication in cell cocultures with high accuracy, as cells coexist at different stages of infection and separation of effector cells from target cells is complicated. In this study, we used replication-dependent reporter vectors to determine requirements for Vif, Vpu, Vpr, or Nef during one cycle of HIV-1 cell coculture and cell-free infection in lymphoid and nonlymphoid cells. Comparative analysis of HIV-1 replication in two cell systems showed that, irrespective of transmission way, accessory proteins were generally less required for virus replication in 293T/CD4/X4 cells than in Jurkat-to-Raji/CD4 cell cocultures. This is consistent with a well-established fact that lymphoid cells express a broad spectrum of restriction factors, while nonlymphoid cells are rather limited in this regard. Remarkably, Vpu deletion reduced the level of cell-free infection, but enhanced the level of cell coculture infection and increased the fraction of multiply infected cells. Nef deficiency did not influence or moderately reduced HIV-1 infection in nonlymphoid and lymphoid cell cocultures, respectively, but strongly affected cell-free infection. Knockout of BST2-a Vpu antagonizing restriction factor-in Jurkat producer cells abolished the enhanced replication of HIV-1 ΔVpu in cell coculture and prevented the formation of viral clusters on cell surface. Thus, BST2-tethered viral particles mediated cell coculture infection more efficiently and at a higher level of multiplicity than diffusely distributed virions. In conclusion, our results demonstrate that the mode of transmission may determine the degree of accessory protein requirements during HIV-1 infection.

Isolation of gene-edited cells via knock-in of short glycophosphatidylinositol-anchored epitope tags.

We describe Surface Oligopeptide knock-in for Rapid Target Selection (SORTS), a novel method to select mammalian cells with precise genome modifications that does not rely on cell cloning. SORTS is designed to disrupt the target gene with an expression cassette encoding an epitope tag embedded into human glycophosphatidylinositol (GPI)-anchored protein CD52. The cassette is very short, usually less than 250 nucleotides, which simplifies donor DNA construction and facilitates transgene integration into the target locus. The chimeric protein is then expressed from the target promoter, processed and exposed on the plasma membrane where it serves as a marker for FACS sorting with tag-specific antibodies. Simultaneous use of two different epitope tags enables rapid isolation of cells with biallelic knock-ins. SORTS can be easily and reliably applied to a number of genome-editing problems such as knocking out genes encoding intracellular or secreted proteins, protein tagging and inactivation of HIV-1 provirus.

Gene Editing in Human Lymphoid Cells: Role for Donor DNA, Type of Genomic Nuclease and Cell Selection Method.

Programmable endonucleases introduce DNA breaks at specific sites, which are repaired by non-homologous end joining (NHEJ) or homology recombination (HDR). Genome editing in human lymphoid cells is challenging as these difficult-to-transfect cells may also inefficiently repair DNA by HDR. Here, we estimated efficiencies and dynamics of knockout (KO) and knockin (KI) generation in human T and B cell lines depending on repair template, target loci and types of genomic endonucleases. Using zinc finger nuclease (ZFN), we have engineered Jurkat and CEM cells with the 8.2 kb human immunodeficiency virus type 1 (HIV-1) ∆Env genome integrated at the adeno-associated virus integration site 1 (AAVS1) locus that stably produce virus particles and mediate infection upon transfection with helper vectors. Knockouts generated by ZFN or clustered regularly interspaced short palindromic repeats (CRISPR/Cas9) double nicking techniques were comparably efficient in lymphoid cells. However, unlike polyclonal sorted cells, gene-edited cells selected by cloning exerted tremendous deviations in functionality as estimated by replication of HIV-1 and human T cell leukemia virus type 1 (HTLV-1) in these cells. Notably, the recently reported high-fidelity eCas9 1.1 when combined to the nickase mutation displayed gene-dependent decrease in on-target activity. Thus, the balance between off-target effects and on-target efficiency of nucleases, as well as choice of the optimal method of edited cell selection should be taken into account for proper gene function validation in lymphoid cells.

Determining antigen specificity of a monoclonal antibody using genome-scale CRISPR-Cas9 knockout library.

An essential step in monoclonal antibody (mAb) development is the characterization and final identification of the specific target antigen and its epitope. Antibody validation is rather straightforward when immunization is carried out with peptide or purified protein, but is more difficult when whole cells or other complex antigens are used for the immunization. Determining antigen specificity of a mAb is further complicated, when reactivity of an antibody is not detected in Western blotting and/or immunoprecipitation assay. In addition to protein-based methods used for antibody characterization, a number of gene-based techniques, such as cDNA expression or short-interfering RNA (siRNA) knockdown have been applied for validation of antibodies with restricted reactivities. Earlier we have generated, characterized, but not identified the BF4 mAb that specifically stains viral biofilms on the surface of the Human T-lymphotropic Virus Type I (HTLV-1) infected T cells. In this study, using the recently developed genome-scale CRISPR-Cas9 knockout (GeCKO) library vectors, we have established the CEM T- and the Raji B cell lines with pooled libraries. After immunofluorescent staining of these cells, negative cell sorting, and guide-RNA (gRNA) sequencing, we have identified BF4 as an anti-CD82 mAb. A deep sequence analysis of GeCKO library transferred to the cells shows that the chance to succeed in the selection of antibody-negative cells and, therefore, to identify a mAb depends on the quality of cell library preparation. We believe that the described method is applicable for identification of many other hybridomas and represents a good alternative to the current protein- and gene-based methods used for mAb validation.