Challenges and best practices in omics benchmarking.

Technological advances enabling massively parallel measurement of biological features - such as microarrays, high-throughput sequencing and mass spectrometry - have ushered in the omics era, now in its third decade. The resulting complex landscape of analytical methods has naturally fostered the growth of an omics benchmarking industry. Benchmarking refers to the process of objectively comparing and evaluating the performance of different computational or analytical techniques when processing and analysing large-scale biological data sets, such as transcriptomics, proteomics and metabolomics. With thousands of omics benchmarking studies published over the past 25 years, the field has matured to the point where the foundations of benchmarking have been established and well described. However, generating meaningful benchmarking data and properly evaluating performance in this complex domain remains challenging. In this Review, we highlight some common oversights and pitfalls in omics benchmarking. We also establish a methodology to bring the issues that can be addressed into focus and to be transparent about those that cannot: this takes the form of a spreadsheet template of guidelines for comprehensive reporting, intended to accompany publications. In addition, a survey of recent developments in benchmarking is provided as well as specific guidance for commonly encountered difficulties.

BEERS2: RNA-Seq simulation through high fidelity in silico modeling.

Simulation of RNA-seq reads is critical in the assessment, comparison, benchmarking and development of bioinformatics tools. Yet the field of RNA-seq simulators has progressed little in the last decade. To address this need we have developed BEERS2, which combines a flexible and highly configurable design with detailed simulation of the entire library preparation and sequencing pipeline. BEERS2 takes input transcripts (typically fully length messenger RNA transcripts with polyA tails) from either customizable input or from CAMPAREE simulated RNA samples. It produces realistic reads of these transcripts as FASTQ, SAM or BAM formats with the SAM or BAM formats containing the true alignment to the reference genome. It also produces true transcript-level quantification values. BEERS2 combines a flexible and highly configurable design with detailed simulation of the entire library preparation and sequencing pipeline and is designed to include the effects of polyA selection and RiboZero for ribosomal depletion, hexamer priming sequence biases, GC-content biases in polymerase chain reaction (PCR) amplification, barcode read errors and errors during PCR amplification. These characteristics combine to make BEERS2 the most complete simulation of RNA-seq to date. Finally, we demonstrate the use of BEERS2 by measuring the effect of several settings on the popular Salmon pseudoalignment algorithm.

Differential Impact In Vivo of Pf4-ΔCre-Mediated and Gp1ba-ΔCre-Mediated Depletion of Cyclooxygenase-1 in Platelets in Mice.

BACKGROUND: Low-dose aspirin is widely used for the secondary prevention of cardiovascular disease. The beneficial effects of low-dose aspirin are attributable to its inhibition of platelet Cox (cyclooxygenase)-1-derived thromboxane A2. Until recently, the use of the Pf4 (platelet factor 4) Cre has been the only genetic approach to generating megakaryocyte/platelet ablation of Cox-1 in mice. However, Pf4-ΔCre displays ectopic expression outside the megakaryocyte/platelet lineage, especially during inflammation. The use of the Gp1ba (glycoprotein 1bα) Cre promises a more specific, targeted approach. METHODS: To evaluate the role of Cox-1 in platelets, we crossed Pf4-ΔCre or Gp1ba-ΔCre mice with Cox-1flox/flox mice to generate platelet Cox-1-/- mice on normolipidemic and hyperlipidemic (Ldlr-/-; low-density lipoprotein receptor) backgrounds. RESULTS: Ex vivo platelet aggregation induced by arachidonic acid or adenosine diphosphate in platelet-rich plasma was inhibited to a similar extent in Pf4-ΔCre Cox-1-/-/Ldlr-/- and Gp1ba-ΔCre Cox-1-/-/Ldlr-/- mice. In a mouse model of tail injury, Pf4-ΔCre-mediated and Gp1ba-ΔCre-mediated deletions of Cox-1 were similarly efficient in suppressing platelet prostanoid biosynthesis. Experimental thrombogenesis and attendant blood loss were similar in both models. However, the impact on atherogenesis was divergent, being accelerated in the Pf4-ΔCre mice while restrained in the Gp1ba-ΔCres. In the former, accelerated atherogenesis was associated with greater suppression of PGI2 biosynthesis, a reduction in the lipopolysaccharide-evoked capacity to produce PGE2 (prostaglandin E) and PGD2 (prostanglandin D), activation of the inflammasome, elevated plasma levels of IL-1ß (interleukin), reduced plasma levels of HDL-C (high-density lipoprotein receptor-cholesterol), and a reduction in the capacity for reverse cholesterol transport. By contrast, in the latter, plasma HDL-C and α-tocopherol were elevated, and MIP-1α (macrophage inflammatory protein-1α) and MCP-1 (monocyte chemoattractant protein 1) were reduced. CONCLUSIONS: Both approaches to Cox-1 deletion similarly restrain thrombogenesis, but a differential impact on Cox-1-dependent prostanoid formation by the vasculature may contribute to an inflammatory phenotype and accelerated atherogenesis in Pf4-ΔCre mice.

Modulation of the Immune Response to Severe Acute Respiratory Syndrome Coronavirus 2 Vaccination by Nonsteroidal Anti-Inflammatory Drugs.

Evidence is scarce to guide the use of nonsteroidal anti-inflammatory drugs (NSAIDs) to mitigate severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) vaccine-related adverse effects, given the possibility of blunting the desired immune response. In this pilot study, we deeply phenotyped a small number of volunteers who did or did not take NSAIDs concomitant with SARS-CoV-2 immunizations to seek initial information on the immune response. A SARS-CoV-2 vaccine-specific receptor binding domain (RBD) IgG antibody response and efficacy in the evoked neutralization titers were evident irrespective of concomitant NSAID consumption. Given the sample size, only a large and consistent signal of immunomodulation would have been detectable, and this was not apparent. However, the information gathered may inform the design of a definitive clinical trial. Here we report a series of divergent omics signals that invites additional hypotheses testing. SIGNIFICANCE STATEMENT: The impact of nonsteroidal anti-inflammatory drugs (NSAIDs) on the immune response elicited by repeat severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) immunizations was profiled by immunophenotypic, proteomic, and metabolomic approaches in a clinical pilot study of small sample size. A SARS-CoV-2 vaccine-specific immune response was evident irrespective of concomitant NSAID consumption. The information gathered may inform the design of a definitive clinical trial.

Comparative evaluation of full-length isoform quantification from RNA-Seq.

BACKGROUND: Full-length isoform quantification from RNA-Seq is a key goal in transcriptomics analyses and has been an area of active development since the beginning. The fundamental difficulty stems from the fact that RNA transcripts are long, while RNA-Seq reads are short. RESULTS: Here we use simulated benchmarking data that reflects many properties of real data, including polymorphisms, intron signal and non-uniform coverage, allowing for systematic comparative analyses of isoform quantification accuracy and its impact on differential expression analysis. Genome, transcriptome and pseudo alignment-based methods are included; and a simple approach is included as a baseline control. CONCLUSIONS: Salmon, kallisto, RSEM, and Cufflinks exhibit the highest accuracy on idealized data, while on more realistic data they do not perform dramatically better than the simple approach. We determine the structural parameters with the greatest impact on quantification accuracy to be length and sequence compression complexity and not so much the number of isoforms. The effect of incomplete annotation on performance is also investigated. Overall, the tested methods show sufficient divergence from the truth to suggest that full-length isoform quantification and isoform level DE should still be employed selectively.

CAMPAREE: a robust and configurable RNA expression simulator.

BACKGROUND: The accurate interpretation of RNA-Seq data presents a moving target as scientists continue to introduce new experimental techniques and analysis algorithms. Simulated datasets are an invaluable tool to accurately assess the performance of RNA-Seq analysis methods. However, existing RNA-Seq simulators focus on modeling the technical biases and artifacts of sequencing, rather than on simulating the original RNA samples. A first step in simulating RNA-Seq is to simulate RNA. RESULTS: To fill this need, we developed the Configurable And Modular Program Allowing RNA Expression Emulation (CAMPAREE), a simulator using empirical data to simulate diploid RNA samples at the level of individual molecules. We demonstrated CAMPAREE's use for generating idealized coverage plots from real data, and for adding the ability to generate allele-specific data to existing RNA-Seq simulators that do not natively support this feature. CONCLUSIONS: Separating input sample modeling from library preparation/sequencing offers added flexibility for both users and developers to mix-and-match different sample and sequencing simulators to suit their specific needs. Furthermore, the ability to maintain sample and sequencing simulators independently provides greater agility to incorporate new biological findings about transcriptomics and new developments in sequencing technologies. Additionally, by simulating at the level of individual molecules, CAMPAREE has the potential to model molecules transcribed from the same genes as a heterogeneous population of transcripts with different states of degradation and processing (splicing, editing, etc.). CAMPAREE was developed in Python, is open source, and freely available at https://github.com/itmat/CAMPAREE .

Correction: A specific sequence in the genome of respiratory syncytial virus regulates the generation of copy-back defective viral genomes.

[This corrects the article DOI: 10.1371/journal.ppat.1007707.].

A specific sequence in the genome of respiratory syncytial virus regulates the generation of copy-back defective viral genomes.

Defective viral genomes of the copy-back type (cbDVGs) are the primary initiators of the antiviral immune response during infection with respiratory syncytial virus (RSV) both in vitro and in vivo. However, the mechanism governing cbDVG generation remains unknown, thereby limiting our ability to manipulate cbDVG content in order to modulate the host response to infection. Here we report a specific genomic signal that mediates the generation of a subset of RSV cbDVG species. Using a customized bioinformatics tool, we identified regions in the RSV genome frequently used to generate cbDVGs during infection. We then created a minigenome system to validate the function of one of these sequences and to determine if specific nucleotides were essential for cbDVG generation at that position. Further, we created a recombinant virus unable to produce a subset of cbDVGs due to mutations introduced in this sequence. The identified sequence was also found as a site for cbDVG generation during natural RSV infections, and common cbDVGs originated at this sequence were found among samples from various infected patients. These data demonstrate that sequences encoded in the viral genome determine the location of cbDVG formation and, therefore, the generation of cbDVGs is not a stochastic process. These findings open the possibility of genetically manipulating cbDVG formation to modulate infection outcome.

RNA-seq analysis of galaninergic neurons from ventrolateral preoptic nucleus identifies expression changes between sleep and wake.

BACKGROUND: Previous studies show that galanin neurons in ventrolateral preoptic nucleus (VLPO-Gal) are essential for sleep regulation. Here, we explored the transcriptional regulation of the VLPO-Gal neurons in sleep by comparing their transcriptional responses between sleeping mice and those kept awake, sacrificed at the same diurnal time. RESULTS: RNA-sequencing (RNA-seq) analysis was performed on eGFP(+) galanin neurons isolated using laser captured microdissection (LCM) from VLPO. Expression of Gal was assessed in our LCM eGFP(+) neurons via real time qPCR and showed marked enrichment when compared to LCM eGFP(-) cells and to bulk VLPO samples. Gene set enrichment analysis utilizing data from a recent single-cell RNA-seq study of the preoptic area demonstrated that our VLPO-Gal samples were highly enriched with galanin-expressing inhibitory neurons, but not galanin-expressing excitatory neurons. A total of 263 genes were differentially expressed between sleep and wake in VLPO-Gal neurons. When comparing differentially expressed genes in VLPO-Gal neurons to differentially expressed genes in a wake-active neuronal region (the medial prefrontal cortex), evidence indicates that both systemic and cell-specific mechanisms contribute to the transcriptional regulation in VLPO-Gal neurons. In both wake-active and sleep-active neurons, ER stress pathways are activated by wake and cold-inducible RNA-binding proteins are activated by sleep. In contrast, expression of DNA repair genes is increased in VLPO-Gal during wakefulness, but increased in wake-active cells during sleep. CONCLUSION: Our study identified transcriptomic responses of the galanin neurons in the ventrolateral preoptic nucleus during sleep and sleep deprivation. Data indicate that VLPO contains mainly sleep-active inhibitory galaninergic neurons. The VLPO galanin neurons show responses to sleep and wake similar to wake-active regions, indicating these responses, such as ER stress and cold-inducible RNA-binding proteins, are systemic affecting all neuronal populations. Region-specific differences in sleep/wake responses were also identified, in particular DNA repair. Our study expands knowledge about the transcriptional response of a distinct group of neurons essential for sleep.

Simulation-based comprehensive benchmarking of RNA-seq aligners.

Alignment is the first step in most RNA-seq analysis pipelines, and the accuracy of downstream analyses depends heavily on it. Unlike most steps in the pipeline, alignment is particularly amenable to benchmarking with simulated data. We performed a comprehensive benchmarking of 14 common splice-aware aligners for base, read, and exon junction-level accuracy and compared default with optimized parameters. We found that performance varied by genome complexity, and accuracy and popularity were poorly correlated. The most widely cited tool underperforms for most metrics, particularly when using default settings.

Outlier detection for improved differential splicing quantification from RNA-Seq experiments with replicates.

Motivation: A key component in many RNA-Seq-based studies is contrasting multiple replicates from different experimental conditions. In this setup, replicates play a key role as they allow to capture underlying biological variability inherent to the compared conditions, as well as experimental variability. However, what constitutes a 'bad' replicate is not necessarily well defined. Consequently, researchers might discard valuable data or downstream analysis may be hampered by failed experiments. Results: Here we develop a probability model to weigh a given RNA-Seq sample as a representative of an experimental condition when performing alternative splicing analysis. We demonstrate that this model detects outlier samples which are consistently and significantly different compared with other samples from the same condition. Moreover, we show that instead of discarding such samples the proposed weighting scheme can be used to downweight samples and specific splicing variations suspected as outliers, gaining statistical power. These weights can then be used for differential splicing (DS) analysis, where the resulting algorithm offers a generalization of the MAJIQ algorithm. Using both synthetic and real-life data, we perform an extensive evaluation of the improved MAJIQ algorithm in different scenarios involving perturbed samples, mislabeled samples, same condition groups, and different levels of coverage, showing it compares favorably to other tools. Overall, this work offers an outlier detection algorithm that can be combined with any splicing pipeline, a generalized and improved version of MAJIQ for DS detection, and evaluation metrics with matching code and data for DS algorithms. Availability and implementation: Software and data are accessible via majiq.biociphers.org/norton_et_al_2017/. Contact: yosephb@upenn.edu. Supplementary information: Supplementary data are available at Bioinformatics online.

Lipocalin-Like Prostaglandin D Synthase but Not Hemopoietic Prostaglandin D Synthase Deletion Causes Hypertension and Accelerates Thrombogenesis in Mice.

Prostaglandin (PG) D2 is formed by two distinct PGD synthases (PGDS): lipocalin-type PGDS (L-PGDS), which acts as a PGD2-producing enzyme and as extracellular lipophilic transporter, and hematopoietic PGDS (H-PGDS), a σ glutathione-S-transferase. PGD2 plays an important role in the maintenance of vascular function; however, the relative contribution of L-PGDS- and H-PGDS-dependent formation of PGD2 in this setting is unknown. To gain insight into the function played by these distinct PGDS, we assessed systemic blood pressure (BP) and thrombogenesis in L-Pgds and H-Pgds knockout (KO) mice. Deletion of L-Pgds depresses urinary PGD2 metabolite (PGDM) by ∼35%, whereas deletion of H-Pgds does so by ∼90%. Deletion of L-Pgds, but not H-Pgds, elevates BP and accelerates the thrombogenic occlusive response to a photochemical injury to the carotid artery. HQL-79, a H-PGDS inhibitor, further depresses PGDM in L-Pgds KO mice, but has no effect on BP or on the thrombogenic response. Gene expression profiling reveals that pathways relevant to vascular function are dysregulated in the aorta of L-Pgds KOs. These results indicate that the functional impact of L-Pgds deletion on vascular homeostasis may result from an autocrine effect of L-PGDS-dependent PGD2 on the vasculature and/or the L-PGDS function as lipophilic carrier protein.

Histone posttranslational modifications predict specific alternative exon subtypes in mammalian brain.

A compelling body of literature, based on next generation chromatin immunoprecipitation and RNA sequencing of reward brain regions indicates that the regulation of the epigenetic landscape likely underlies chronic drug abuse and addiction. It is now critical to develop highly innovative computational strategies to reveal the relevant regulatory transcriptional mechanisms that may underlie neuropsychiatric disease. We have analyzed chromatin regulation of alternative splicing, which is implicated in cocaine exposure in mice. Recent literature has described chromatin-regulated alternative splicing, suggesting a novel function for drug-induced neuroepigenetic remodeling. However, the extent of the genome-wide association between particular histone modifications and alternative splicing remains unexplored. To address this, we have developed novel computational approaches to model the association between alternative splicing and histone posttranslational modifications in the nucleus accumbens (NAc), a brain reward region. Using classical statistical methods and machine learning to combine ChIP-Seq and RNA-Seq data, we found that specific histone modifications are strongly associated with various aspects of differential splicing. H3K36me3 and H3K4me1 have the strongest association with splicing indicating they play a significant role in alternative splicing in brain reward tissue.

Propagation of adipogenic signals through an epigenomic transition state.

The transcriptional mechanisms by which temporary exposure to developmental signals instigates adipocyte differentiation are unknown. During early adipogenesis, we find transient enrichment of the glucocorticoid receptor (GR), CCAAT/enhancer-binding protein beta (CEBPbeta), p300, mediator subunit 1, and histone H3 acetylation near genes involved in cell proliferation, development, and differentiation, including the gene encoding the master regulator of adipocyte differentiation, peroxisome proliferator-activated receptor gamma2 (PPARgamma2). Occupancy and enhancer function are triggered by adipogenic signals, and diminish upon their removal. GR, which is important for adipogenesis but need not be active in the mature adipocyte, functions transiently with other enhancer proteins to propagate a new program of gene expression that includes induction of PPARgamma2, thereby providing a memory of the earlier adipogenic signal. Thus, the conversion of preadipocyte to adipocyte involves the formation of an epigenomic transition state that is not observed in cells at the beginning or end of the differentiation process.

A comparison of Illumina and Ion Torrent sequencing platforms in the context of differential gene expression.

BACKGROUND: Though Illumina has largely dominated the RNA-Seq field, the simultaneous availability of Ion Torrent has left scientists wondering which platform is most effective for differential gene expression (DGE) analysis. Previous investigations of this question have typically used reference samples derived from cell lines and brain tissue, and do not involve biological variability. While these comparisons might inform studies of tissue-specific expression, marked by large-scale transcriptional differences, this is not the common use case. RESULTS: Here we employ a standard treatment/control experimental design, which enables us to evaluate these platforms in the context of the expression differences common in differential gene expression experiments. Specifically, we assessed the hepatic inflammatory response of mice by assaying liver RNA from control and IL-1ß treated animals with both the Illumina HiSeq and the Ion Torrent Proton sequencing platforms. We found the greatest difference between the platforms at the level of read alignment, a moderate level of concordance at the level of DGE analysis, and nearly identical results at the level of differentially affected pathways. Interestingly, we also observed a strong interaction between sequencing platform and choice of aligner. By aligning both real and simulated Illumina and Ion Torrent data with the twelve most commonly-cited aligners in the literature, we observed that different aligner and platform combinations were better suited to probing different genomic features; for example, disentangling the source of expression in gene-pseudogene pairs. CONCLUSIONS: Taken together, our results indicate that while Illumina and Ion Torrent have similar capacities to detect changes in biology from a treatment/control experiment, these platforms may be tailored to interrogate different transcriptional phenomena through careful selection of alignment software.

Myeloid cell microsomal prostaglandin E synthase-1 fosters atherogenesis in mice.

Microsomal prostaglandin E synthase-1 (mPGES-1) in myeloid and vascular cells differentially regulates the response to vascular injury, reflecting distinct effects of mPGES-1-derived PGE2 in these cell types on discrete cellular components of the vasculature. The cell selective roles of mPGES-1 in atherogenesis are unknown. Mice lacking mPGES-1 conditionally in myeloid cells (Mac-mPGES-1-KOs), vascular smooth muscle cells (VSMC-mPGES-1-KOs), or endothelial cells (EC-mPGES-1-KOs) were crossed into hyperlipidemic low-density lipoprotein receptor-deficient animals. En face aortic lesion analysis revealed markedly reduced atherogenesis in Mac-mPGES-1-KOs, which was concomitant with a reduction in oxidative stress, reflective of reduced macrophage infiltration, less lesional expression of inducible nitric oxide synthase (iNOS), and lower aortic expression of NADPH oxidases and proinflammatory cytokines. Reduced oxidative stress was reflected systemically by a decline in urinary 8,12-iso-iPF2α-VI. In contrast to exaggeration of the response to vascular injury, deletion of mPGES-1 in VSMCs, ECs, or both had no detectable phenotypic impact on atherogenesis. Macrophage foam cell formation and cholesterol efflux, together with plasma cholesterol and triglycerides, were unchanged as a function of genotype. In conclusion, myeloid cell mPGES-1 promotes atherogenesis in hyperlipidemic mice, coincident with iNOS-mediated oxidative stress. By contrast, mPGES-1 in vascular cells does not detectably influence atherogenesis in mice. This strengthens the therapeutic rationale for targeting macrophage mPGES-1 in inflammatory cardiovascular diseases.

Systematic evaluation of spliced alignment programs for RNA-seq data.

High-throughput RNA sequencing is an increasingly accessible method for studying gene structure and activity on a genome-wide scale. A critical step in RNA-seq data analysis is the alignment of partial transcript reads to a reference genome sequence. To assess the performance of current mapping software, we invited developers of RNA-seq aligners to process four large human and mouse RNA-seq data sets. In total, we compared 26 mapping protocols based on 11 programs and pipelines and found major performance differences between methods on numerous benchmarks, including alignment yield, basewise accuracy, mismatch and gap placement, exon junction discovery and suitability of alignments for transcript reconstruction. We observed concordant results on real and simulated RNA-seq data, confirming the relevance of the metrics employed. Future developments in RNA-seq alignment methods would benefit from improved placement of multimapped reads, balanced utilization of existing gene annotation and a reduced false discovery rate for splice junctions.

Benchmark analysis of algorithms for determining and quantifying full-length mRNA splice forms from RNA-seq data.

MOTIVATION: Because of the advantages of RNA sequencing (RNA-Seq) over microarrays, it is gaining widespread popularity for highly parallel gene expression analysis. For example, RNA-Seq is expected to be able to provide accurate identification and quantification of full-length splice forms. A number of informatics packages have been developed for this purpose, but short reads make it a difficult problem in principle. Sequencing error and polymorphisms add further complications. It has become necessary to perform studies to determine which algorithms perform best and which if any algorithms perform adequately. However, there is a dearth of independent and unbiased benchmarking studies. Here we take an approach using both simulated and experimental benchmark data to evaluate their accuracy. RESULTS: We conclude that most methods are inaccurate even using idealized data, and that no method is highly accurate once multiple splice forms, polymorphisms, intron signal, sequencing errors, alignment errors, annotation errors and other complicating factors are present. These results point to the pressing need for further algorithm development. AVAILABILITY AND IMPLEMENTATION: Simulated datasets and other supporting information can be found at http://bioinf.itmat.upenn.edu/BEERS/bp2. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Deep sequencing the circadian and diurnal transcriptome of Drosophila brain.

Eukaryotic circadian clocks include transcriptional/translational feedback loops that drive 24-h rhythms of transcription. These transcriptional rhythms underlie oscillations of protein abundance, thereby mediating circadian rhythms of behavior, physiology, and metabolism. Numerous studies over the last decade have used microarrays to profile circadian transcriptional rhythms in various organisms and tissues. Here we use RNA sequencing (RNA-seq) to profile the circadian transcriptome of Drosophila melanogaster brain from wild-type and period-null clock-defective animals. We identify several hundred transcripts whose abundance oscillates with 24-h periods in either constant darkness or 12 h light/dark diurnal cycles, including several noncoding RNAs (ncRNAs) that were not identified in previous microarray studies. Of particular interest are U snoRNA host genes (Uhgs), a family of diurnal cycling noncoding RNAs that encode the precursors of more than 50 box-C/D small nucleolar RNAs, key regulators of ribosomal biogenesis. Transcriptional profiling at the level of individual exons reveals alternative splice isoforms for many genes whose relative abundances are regulated by either period or circadian time, although the effect of circadian time is muted in comparison to that of period. Interestingly, period loss of function significantly alters the frequency of RNA editing at several editing sites, suggesting an unexpected link between a key circadian gene and RNA editing. We also identify tens of thousands of novel splicing events beyond those previously annotated by the modENCODE Consortium, including several that affect key circadian genes. These studies demonstrate extensive circadian control of ncRNA expression, reveal the extent of clock control of alternative splicing and RNA editing, and provide a novel, genome-wide map of splicing in Drosophila brain.

Myeloid Cell COX-2 deletion reduces mammary tumor growth through enhanced cytotoxic T-lymphocyte function.

Cyclooxygenase-2 (COX-2) expression is associated with poor prognosis across a range of human cancers, including breast cancer. The contribution of tumor cell-derived COX-2 to tumorigenesis has been examined in numerous studies; however, the role of stromal-derived COX-2 is ill-defined. Here, we examined how COX-2 in myeloid cells, an immune cell subset that includes macrophages, influences mammary tumor progression. In mice engineered to selectively lack myeloid cell COX-2 [myeloid-COX-2 knockout (KO) mice], spontaneous neu oncogene-induced tumor onset was delayed, tumor burden reduced, and tumor growth slowed compared with wild-type (WT). Similarly, growth of neu-transformed mammary tumor cells as orthotopic tumors in immune competent syngeneic myeloid-COX-2 KO host mice was reduced compared with WT. By flow cytometric analysis, orthotopic myeloid-COX-2 KO tumors had lower tumor-associated macrophage (TAM) infiltration consistent with impaired colony stimulating factor-1-dependent chemotaxis by COX-2 deficient macrophages in vitro. Further, in both spontaneous and orthotopic tumors, COX-2-deficient TAM displayed lower immunosuppressive M2 markers and this was coincident with less suppression of CD8(+) cytotoxic T lymphocytes (CTLs) in myeloid-COX-2 KO tumors. These studies suggest that reduced tumor growth in myeloid-COX-2 KO mice resulted from disruption of M2-like TAM function, thereby enhancing T-cell survival and immune surveillance. Antibody-mediated depletion of CD8(+), but not CD4(+) cells, restored tumor growth in myeloid-COX-2 KO to WT levels, indicating that CD8(+) CTLs are dominant antitumor effectors in myeloid-COX-2 KO mice. Our studies suggest that inhibition of myeloid cell COX-2 can potentiate CTL-mediated tumor cytotoxicity and may provide a novel therapeutic approach in breast cancer therapy.