Cognitive and clinical predictors of a long-term course in obsessive compulsive disorder: A machine learning approach in a prospective cohort study

BACKGROUND

Obsessive compulsive disorder (OCD) is a disabling illness with a chronic course, yet data on long-term outcomes are scarce. This study aimed to examine the long-term course of OCD in patients treated with different approaches (drugs, psychotherapy, and psychosurgery) and to identify predictors of clinical outcome by machine learning.

METHOD

We included outpatients with OCD treated at our referral unit. Demographic and neuropsychological data were collected at baseline using standardized instruments. Clinical data were collected at baseline, 12 weeks after starting pharmacological treatment prescribed at study inclusion, and after follow-up.

RESULTS

Of the 60 outpatients included, with follow-up data available for 5-17 years (mean = 10.6 years), 40 (67.7 %) were considered non-responders to adequate treatment at the end of the study. The best machine learning model achieved a correlation of 0.63 for predicting the long-term Yale-Brown Obsessive Compulsive Scale (Y-BOCS) score by adding clinical response (to the first pharmacological treatment) to the baseline clinical and neuropsychological characteristics.

LIMITATIONS

Our main limitations were the sample size, modest in the context of traditional ML studies, and the sample composition, more representative of rather severe OCD cases than of patients from the general community.

CONCLUSIONS

Many patients with OCD showed persistent and disabling symptoms at the end of follow-up despite comprehensive treatment that could include medication, psychotherapy, and psychosurgery. Machine learning algorithms can predict the long-term course of OCD using clinical and cognitive information to optimize treatment options.

eQTL colocalization analysis highlights novel susceptibility genes in Autism Spectrum Disorders (ASD)

Autism Spectrum Disorders (ASD) are a group of neurodevelopmental disorders (NDDs) characterized by difficulties in social interaction and communication, repetitive behavior, and restricted interests. ASD has proven to have a strong genetic component. However, defining causal genes is still one of the main challenges in GWAS, since the vast majority (>90%) of detected signals lie within the non-coding genome. Expression quantitative trait locus (eQTL) colocalization analysis determines whether a specific variant is responsible for both a local eQTL and GWAS association and has helped leverage data and rendering gene discovery for a wide array of diseases. Here we further mine the largest ASD GWAS performed to date (18,381 cases and 27,969 controls) altogether with GWAS summary statistics from the main PGC studies (Schizophrenia, MD (Major Depression) and ADHD (Attention Deficit/Hyperactivity Disorder)), by using eQTpLot, a newly developed tool that illustrates the colocalization of GWAS and eQTL signals in a locus, and the enrichment of and correlation between the candidate gene eQTLs and trait-significant variants. This analysis points up 8 genes with a significant eQTL colocalization signal in ASD (CRHR1, KANSL1, MANBA, MAPT, MMP12, NKX2-2, PTPRE and WNT3) and one gene (SRPK2) with a marginally significant colocalization signal (r = 0.69, p < 1 × 10-6), and specifically highlights the potentially causal role of MAPT (r = 0.76, p < 1 × 10-6), NKX2-2 (r = 0.71, p-value = 2.26-02) and PTPRE (r = 0.97, p-value = 2.63-04) when restricting the analysis to brain tissue.

Psychiatric polygenic risk as a predictor of COVID-19 risk and severity: insight into the genetic overlap between schizophrenia and COVID-19

Despite the high contagion and mortality rates that have accompanied the coronavirus disease-19 (COVID-19) pandemic, the clinical presentation of the syndrome varies greatly from one individual to another. Potential host factors that accompany greater risk from COVID-19 have been sought and schizophrenia (SCZ) patients seem to present more severe COVID-19 than control counterparts, with certain gene expression similarities between psychiatric and COVID-19 patients reported. We used summary statistics from the last SCZ, bipolar disorder (BD), and depression (DEP) meta-analyses available on the Psychiatric Genomics Consortium webpage to calculate polygenic risk scores (PRSs) for a target sample of 11,977 COVID-19 cases and 5943 subjects with unknown COVID-19 status. Linkage disequilibrium score (LDSC) regression analysis was performed when positive associations were obtained from the PRS analysis. The SCZ PRS was a significant predictor in the case/control, symptomatic/asymptomatic, and hospitalization/no hospitalization analyses in the total and female samples; and of symptomatic/asymptomatic status in men. No significant associations were found for the BD or DEP PRS or in the LDSC regression analysis. SNP-based genetic risk for SCZ, but not for BD or DEP, may be associated with higher risk of SARS-CoV-2 infection and COVID-19 severity, especially among women; however, predictive accuracy barely exceeded chance level. We believe that the inclusion of sexual loci and rare variations in the analysis of genomic overlap between SCZ and COVID-19 will help to elucidate the genetic commonalities between these conditions.

The genomics of visuospatial neurocognition in obsessive-compulsive disorder: A preliminary GWAS

BACKGROUND

The study of Obsessive-Compulsive Disorder (OCD) genomics has primarily been tackled by Genome-wide association studies (GWAS), which have encountered troubles in identifying replicable single nucleotide polymorphisms (SNPs). Endophenotypes have emerged as a promising avenue of study in trying to elucidate the genomic bases of complex traits such as OCD.

METHODS

We analyzed the association of SNPs across the whole genome with the construction of visuospatial information and executive performance through four neurocognitive variables assessed by the Rey-Osterrieth Complex Figure Test (ROCFT) in a sample of 133 OCD probands. Analyses were performed at SNP- and gene-level.

RESULTS

No SNP reached genome-wide significance, although there was one SNP almost reaching significant association with copy organization (rs60360940; P = 9.98E-08). Suggestive signals were found for the four variables at both SNP- (P < 1E-05) and gene-levels (P < 1E-04). Most of the suggestive signals pointed to genes and genomic regions previously associated with neurological function and neuropsychological traits.

LIMITATIONS

Our main limitations were the sample size, which was limited to identify associated signals at a genome-wide level, and the composition of the sample, more representative of rather severe OCD cases than a population-based OCD sample with a broad severity spectrum.

CONCLUSIONS

Our results suggest that studying neurocognitive variables in GWAS would be more informative on the genetic basis of OCD than the classical case/control GWAS, facilitating the genetic characterization of OCD and its different clinical profiles, the development of individualized treatment approaches, and the improvement of prognosis and treatment response.

The non-coding genome in Autism Spectrum Disorders

Autism Spectrum Disorders (ASD) are a group of neurodevelopmental disorders (NDDs) characterized by difficulties in social interaction and communication, repetitive behavior, and restricted interests. While ASD have been proven to have a strong genetic component, current research largely focuses on coding regions of the genome. However, non-coding DNA, which makes up for ∼99% of the human genome, has recently been recognized as an important contributor to the high heritability of ASD, and novel sequencing technologies have been a milestone in opening up new directions for the study of the gene regulatory networks embedded within the non-coding regions. Here, we summarize current progress on the contribution of non-coding alterations to the pathogenesis of ASD and provide an overview of existing methods allowing for the study of their functional relevance, discussing potential ways of unraveling ASD's "missing heritability".

Development and evaluations of the ancestry informative markers of the VISAGE Enhanced Tool for Appearance and Ancestry

The VISAGE Enhanced Tool for Appearance and Ancestry (ET) has been designed to combine markers for the prediction of bio-geographical ancestry plus a range of externally visible characteristics into a single massively parallel sequencing (MPS) assay. We describe the development of the ancestry panel markers used in ET, and the enhanced analyses they provide compared to previous MPS-based forensic ancestry assays. As well as established autosomal single nucleotide polymorphisms (SNPs) that differentiate sub-Saharan African, European, East Asian, South Asian, Native American, and Oceanian populations, ET includes autosomal SNPs able to efficiently differentiate populations from Middle East regions. The ability of the ET autosomal ancestry SNPs to distinguish Middle East populations from other continentally defined population groups is such that characteristic patterns for this region can be discerned in genetic cluster analysis using STRUCTURE. Joint cluster membership estimates showing individual co-ancestry that signals North African or East African origins were detected, or cluster patterns were seen that indicate origins from central and Eastern regions of the Middle East. In addition to an augmented panel of autosomal SNPs, ET includes panels of 85 Y-SNPs, 16 X-SNPs and 21 autosomal Microhaplotypes. The Y- and X-SNPs provide a distinct method for obtaining extra detail about co-ancestry patterns identified in males with admixed backgrounds. This study used the 1000 Genomes admixed African and admixed American sample sets to fully explore these enhancements to the analysis of individual co-ancestry. Samples from urban and rural Brazil with contrasting distributions of African, European, and Native American co-ancestry were also studied to gauge the efficiency of combining Y- and X-SNP data for this purpose. The small panel of Microhaplotypes incorporated in ET were selected because they showed the highest levels of haplotype diversity amongst the seven population groups we sought to differentiate. Microhaplotype data was not formally combined with single-site SNP genotypes to analyse ancestry. However, the haplotype sequence reads obtained with ET from these loci creates an effective system for de-convoluting two-contributor mixed DNA. We made simple mixture experiments to demonstrate that when the contributors have different ancestries and the mixture ratios are imbalanced (i.e., not 1:1 mixtures) the ET Microhaplotype panel is an informative system to infer ancestry when this differs between the contributors.

Neuropsychological performance and predictors of pharmacological treatment response in obsessive compulsive disorder

BACKGROUND

Obsessive Compulsive Disorder (OCD) is characterized by the presence of executive dysfunctions. As organizational strategies may play an important role as a possible endophenotype of the disorder, we decided to investigate non-verbal memory and organizational abilities in OCD. We also investigated how organization and non-verbal memory differ between responder and non-responder patients to pharmacological treatment, to test whether cognitive functions can predict the response to pharmacological treatment.

METHODS

In Study 1, executive and clinical functioning measures were applied to 162 OCD and 95 controls. In Study 2, clinical, intelligence and executive functioning measures were applied to 72 OCD responders and 63 OCD non-responder patients.

RESULTS

OCD patients and controls from Study 1 differed in copy organization (p < 0.01) and delayed recall (p = 0.048). In Study 2, the OCD responders displayed better copy organization (p = 0.013) and lower depressive, anxious and OCD symptoms (p < 0.01 in the three cases). Scores in the following instruments were found to predict the response to pharmacological treatment: HDRS, Y-BOCS, Raven progressive matrices, and Direct digit subtest from the Wechsler's scale (p < 0.01 in all four cases).

LIMITATIONS

In Study 1, the imbalance of the sample can be considered a limitation, whilst in Study 2, some of the levels of pharmacological resistance were not represented.

CONCLUSIONS

In this study, non-verbal memory and organization was affected in OCD. Responder patients also displayed better executive functioning and fluid intelligence. Organizational ability is a predictor of pharmacological response to SSRI monotherapy in a predictive model controlling for anxious symptoms.

How is COVID-19 affecting patients with obsessive-compulsive disorder? A longitudinal study on the initial phase of the pandemic in a Spanish cohort

BACKGROUND

Although the consequences of the COVID-19 pandemic on emotional health are evident, little is known about its impact on patients with obsessive-compulsive disorder (OCD).

METHODS

One hundred and twenty-seven patients with OCD who attended a specialist OCD Clinic in Barcelona, Spain, were assessed by phone from April 27 to May 25, 2020, during the early phase of the pandemic, using the Yale-Brown Obsessive-Compulsive Scale (Y-BOCS) and a structured interview that collected clinical and sociodemographic information. Results were compared with those for 237 healthy controls from the same geographic area who completed an online survey.

RESULTS

Although 65.3% of the patients with OCD described a worsening of their symptoms, only 31.4% had Y-BOCS scores that increased >25%. The risk of getting infected by SARS-CoV2 was reported as a new obsession by 44.8%, but this only became the main obsessive concern in approximately 10% of the patients. Suicide-related thoughts were more frequent among the OCD cohort than among healthy controls. The presence of prepandemic depression, higher Y-BOCS scores, contamination/washing symptoms, and lower perceived social support all predicted a significantly increased risk of OCD worsening.

CONCLUSIONS

Most patients with OCD appear to be capable of coping with the emotional stress of the COVID-19 outbreak and its consequences during the initial phase of the pandemic. Nevertheless, the current crisis constitutes a risk factor for a significant worsening of symptoms and suicidal ideation. Action is needed to ensure effective and individualized follow-up care for patients with OCD in the COVID-19 era.

Population structure and relatedness estimates in a Mexican sample

Population stratification (PS) is a confounding factor in genome-wide association studies (GWASs) and also an interesting process itself. Latin American populations have mixed genetic ancestry, which may account for PS. We have analyzed the relatedness, by means of the identity-by-descent (IBD) estimations, in a sample of 1805 individuals and 1.006.703 autosomal mutations from a case-control study of colorectal cancer in Mexico. When using the recommended protocol for quality control assessment, 402 should have been removed due to relatedness. Our purpose was to analyze this value in the context of an admixed population. For that aim, we reanalyzed the sample using two software designed for admixed populations, obtaining estimates of 110 and 70 related individuals to remove. The results showed that the first estimation of relatedness was an effect of the higher Native American contribution in part of the data samples, being a confounding factor for IBD estimations. We conclude in the importance of considering PS and genetic ancestry in order to avoid spurious results, not only in GWAS but also in relatedness analysis.

Exploring the biological role of postzygotic and germinal de novo mutations in ASD

De novo mutations (DNMs), including germinal and postzygotic mutations (PZMs), are a strong source of causality for Autism Spectrum Disorder (ASD). However, the biological processes involved behind them remain unexplored. Our aim was to detect DNMs (germinal and PZMs) in a Spanish ASD cohort (360 trios) and to explore their role across different biological hierarchies (gene, biological pathway, cell and brain areas) using bioinformatic approaches. For the majority of the analysis, a combined ASD cohort (N = 2171 trios) was created using previously published data by the Autism Sequencing Consortium (ASC). New plausible candidate genes for ASD such as FMR1 and NFIA were found. In addition, genes harboring PZMs were significantly enriched for miR-137 targets in comparison with germinal DNMs that were enriched in GO terms related to synaptic transmission. The expression pattern of genes with PZMs was restricted to early mid-fetal cortex. In contrast, the analysis of genes with germinal DNMs revealed a spatio-temporal window from early to mid-fetal development stages, with expression in the amygdala, cerebellum, cortex and striatum. These results provide evidence of the pathogenic role of PZMs and suggest the existence of distinct mechanisms between PZMs and germinal DNMs that are influencing ASD risk.

Obesity-Related Genetic Determinants of Heart Failure Prognosis

PURPOSE

Recent advances in genomics offer a smart option for predicting future risk of disease and prognosis. The objective of this study was to examine the prognostic value in heart failure (HF) patients, of a series of single nucleotide polymorphisms (SNPs).

METHODS

A selection of 192 SNPs found to be related with obesity, body mass index, circulating lipids or cardiovascular diseases were genotyped in 191 patients with HF. Anthropometrical and clinical variables were collected for each patient, and death and readmission by HF were registered as the primary endpoint.

RESULTS

A total of 53 events were registered during a follow-up period of 438 (263-1077) days (median (IQR)). Eight SNPs strongly related to obesity and HF prognosis were selected as possible prognostic variables. From these, rs10189761 and rs737337 variants were independently associated with HF prognosis (HR 2.295 (1.287-4.089, 95% CI); p = 0.005), whereas rs10423928, rs1800437, rs737337 and rs9351814 were related with bad prognosis only in obese patients (HR 2.142 (1.438-3.192, 95% CI); p = 0.00018). Combined scores of the genomic variants were highly predictive of poor prognosis.

CONCLUSIONS

SNPs rs10189761 and rs737337 were identified, for the first time, as independent predictors of major clinical outcomes in patients with HF. The data suggests an additive predictive value of these SNPs for a HF prognosis. In particular for obese patients, SNPs rs10423928, rs1800437, rs737337 and rs9351814 were related with a bad prognosis. Combined scores weighting the risk of each genomic variant could effect interesting new tools to stratify the prognostic risk of HF patients.

PIMA: A population informative multiplex for the Americas

We describe an ancestry-informative autosomal SNP multiplex designed to be a small-scale, flexible panel that can complement uniparental markers in assessing the American variability (i.e. pre-Colombian) found in contemporary indigenous American populations. This study centered on choosing SNPs with the specific characteristics of: 1) extreme allele frequency differences between indigenous Americans and the African, European and East Asian population groups that contribute to present-day population variation in the Americas; 2) high informativeness-for-assignment I_n values; and 3) well-spaced genomic distribution and chromosomal separation from existing small-scale forensic ancestry marker sets. The resulting capillary electrophoresis SNaPshot single base extension test was named: PIMA (Population Informative Multiplex for the Americas), comprising 26 autosomal SNPs, a single X-chromosome SNP plus the amelogenin sex marker adapted for SNaPshot. PIMA complements the established 34plex forensic ancestry panel to provide a powerful and simple tool for the analysis of American populations, including those with admixed histories, commonly encountered in America. Comparing the results obtained with the combined marker panels of PIMA and 34plex to SNP data from a much larger ancestry panel allowed us to gauge their relative efficiency. PIMA+34plex gives equivalent power to the 314-SNP 'LACE' genomic ancestry control panel, while requiring a much smaller genotyping effort. The ancestry profiles and genetic structure of 22 populations spread across the American continent were estimated using PIMA+34plex data, and those estimates were contrasted with information provided by uniparental markers (mtDNA and Y-chromosome loci) for a small set of admixed individuals from Venezuela. Our results indicate that an American genetic component is efficiently detected in contemporary American populations using a small set of ancestry informative SNPs, and these co-ancestry estimates are consistent with the known history and demography of the Americas. The small scale and high population differentiation power of PIMA, particularly when combined with 34plex, provides a practical and powerful tool for genetic studies of American populations as well as forensic DNA analyses.

Genetic susceptibility in pharmacodynamic and pharmacokinetic pathways underlying drug-induced arrhythmia and sudden unexplained deaths

Drug-induced arrhythmia is an adverse drug reaction that can be potentially fatal since it is mostly related to drug-induced QT prolongation, a known risk factor for Torsade de Pointes and sudden cardiac death (SCD). Several risk factors have been described in association to these drug-induced events, such as preexistent cardiac disease and genetic variation. Our objective was to study the genetic susceptibility in pharmacodynamic and pharmacokinetic pathways underlying suspected drug-induced arrhythmias and sudden unexplained deaths in 32 patients. The genetic component in the pharmacodynamic pathway was studied by analysing 96 genes associated with higher risk of SCD through massive parallel sequencing. Pharmacokinetic-mediated genetic susceptibility was investigated by studying the genes encoding cytochrome P450 enzymes using medium-throughput genotyping. Pharmacodynamic analysis showed three probably pathogenic variants and 45 variants of uncertain significance in 28 patients, several of them previously described in relation to mild or late onset cardiomyopathies. These results suggest that genetic variants in cardiomyopathy genes, in addition to those related with channelopathies, could be relevant to drug-induced cardiotoxicity and contribute to the arrhythmogenic phenotype. Pharmacokinetic analysis showed three patients that could have an altered metabolism of the drugs they received involving CYP2C19 and/or CYP2D6, probably contributing to the arrhythmogenic phenotype. The study of genetic variants in both pharmacodynamic and pharmacokinetic pathways may be a useful strategy to understand the multifactorial mechanism of drug-induced events in both clinical practice and forensic field. However, it is necessary to comprehensively study and evaluate the contribution of the genetic susceptibility to drug-induced cardiotoxicity.

The Role of Colorectal Cancer Risk Chromosomal Regions in Colombian Admixed Populations

Background: Several colorectal cancer susceptibility disease loci have been discovered through genome-wide association studies. However most of the variants were originally identified in Caucasian populations. Aim: To analyze the role of 20 known risk SNPs for colorectal cancer. Methods: Given that linkage disequilibrium is highly dependent on population demographic history and admixture background, we studied 20 risk SNPs in a pooled sample of 955 cases and 968 controls from admixed populations in Colombia. Results: The replication was reached for 11 out of 20 nominally associated SNPs; with allelic odds ratios (OR) ranging from 1.14 to 1.41, indicating a minimal increase in risk individually, however coinheritance of those SNPs resulted in an overall OR = 5.4 (95% CI: 3.052-9.731, P = 1.16E−08). Most of the variants followed a recessive model consistent with significant homozygous ORs distributed between 1.3 and 1.65. Among the most associated markers we found: rs4939827 (18q21.1, P = 7.35E−6), rs10411210 (19q13.11, P = 0.001), rs10795668 (10p14, P = 0.0024), rs4444235 (14q.2.2, P = 0.005), rs961253 (20p12.3, P = 0.006), rs16892766 (8q23.3, P = 0.011) and rs1050547 (8q24.21, P = 0.017). Conclusion: Our findings in Colombia have addressed the admixture and how this has influenced the risk associated with the known/unknown colorectal cancer regions, providing a comprehensive vision about several CRC-susceptibility SNPs identified in European populations, which also resulted, associated with an increased risk to CRC in the Colombian population, even though frequency and genetic structure differences accounted for those nonreplicated SNPs.

Two Different Genetic Markers for High and Low Myopia

In myopia patients, Rh and acid phosphatase were typed in two groups: group 1 consisted of 214 patients with low myopia (−6 D or less); group 2 of 124 patients with high myopia (more than −6 D). Statistical analysis of the markers showed a good Hardy-Weinberg equilibrium for both groups. In the Rh system there was a significant difference between group 1 and the control population (p < 0.05), but not between group 2 and control (p > 0.1). In the case of ACP there was a significant difference between group 2 and the control population (p < 0.05), but not between group 1 and control (p > 0.25). We conclude that the observed association between myopia and Rh system (chromosome 1) involves low myopia, while the association between myopia and acid phosphatase (chromosome 2) involves high myopia. Further DNA researche will lead to more specific results.

A pathway-based association study reveals variants from Wnt signalling genes contributing to asthma susceptibility

BACKGROUND

Genetic susceptibility to asthma is currently linked to a handful of genes which have a limited ability to predict the overall disease risk, suggesting the existence of many other genes involved in disease development. Accumulated evidence from association studies in genes related by biological pathways could reveal novel asthma genes.

OBJECTIVE

To reveal novel asthma susceptibility genes by means of a pathway-based association study.

METHODS

Based on summary data from a previous a genomewide association study (GWAS) of asthma, we first identified significant biological pathways using a gene-set enrichment analysis. We then mapped all tested single nucleotide polymorphisms (SNPs) on the genes contributing to significant pathways and prioritized those with a disproportionate number of nominal significant associations for further studies. For those prioritized genes, association studies were performed for selected SNPs in independent case-control samples (n = 1765) using logistic regression models, and results were meta-analysed with those from the GWAS.

RESULTS

Two biological processes were significantly enriched: the cytokine-cytokine receptor interaction (P = 0.002) and the Wnt signalling (P = 0.012). From the 417 genes interacting in these two pathways, 10 showed an excess of nominal associations, including a known asthma susceptibility locus (encoding SMAD family member 3) and other novel candidate genes. From the latter, association studies of 14 selected SNPs evidenced replication in a locus near the frizzled class receptor 6 (FZD6) gene (P = 9.90 × 10^-4 ), which had a consistent direction of effects with the GWAS findings (meta-analysed odds ratio = 1.49; P = 5.87 × 10^-6 ) and was in high linkage disequilibrium with expression quantitative trait loci in lung tissues.

CONCLUSIONS AND CLINICAL RELEVANCE

This study revealed the importance of two biological pathways in asthma pathogenesis and identified a novel susceptibility locus near Wnt signalling genes.

Forensic SNP genotyping with SNaPshot: Technical considerations for the development and optimization of multiplexed SNP assays

This review explores the key factors that influence the optimization, routine use, and profile interpretation of the SNaPshot single-base extension (SBE) system applied to forensic single-nucleotide polymorphism (SNP) genotyping. Despite being a mainly complimentary DNA genotyping technique to routine STR profiling, use of SNaPshot is an important part of the development of SNP sets for a wide range of forensic applications with these markers, from genotyping highly degraded DNA with very short amplicons to the introduction of SNPs to ascertain the ancestry and physical characteristics of an unidentified contact trace donor. However, this technology, as resourceful as it is, displays several features that depart from the usual STR genotyping far enough to demand a certain degree of expertise from the forensic analyst before tackling the complex casework on which SNaPshot application provides an advantage. In order to provide the basis for developing such expertise, we cover in this paper the most challenging aspects of the SNaPshot technology, focusing on the steps taken to design primer sets, optimize the PCR and single-base extension chemistries, and the important features of the peak patterns observed in typical forensic SNP profiles using SNaPshot. With that purpose in mind, we provide guidelines and troubleshooting for multiplex-SNaPshot-oriented primer design and the resulting capillary electrophoresis (CE) profile interpretation (covering the most commonly observed artifacts and expected departures from the ideal conditions).

Stem cell-like transcriptional reprogramming mediates metastatic resistance to mTOR inhibition

Inhibitors of the mechanistic target of rapamycin (mTOR) are currently used to treat advanced metastatic breast cancer. However, whether an aggressive phenotype is sustained through adaptation or resistance to mTOR inhibition remains unknown. Here, complementary studies in human tumors, cancer models and cell lines reveal transcriptional reprogramming that supports metastasis in response to mTOR inhibition. This cancer feature is driven by EVI1 and SOX9. EVI1 functionally cooperates with and positively regulates SOX9, and promotes the transcriptional upregulation of key mTOR pathway components (REHB and RAPTOR) and of lung metastasis mediators (FSCN1 and SPARC). The expression of EVI1 and SOX9 is associated with stem cell-like and metastasis signatures, and their depletion impairs the metastatic potential of breast cancer cells. These results establish the mechanistic link between resistance to mTOR inhibition and cancer metastatic potential, thus enhancing our understanding of mTOR targeting failure.