A scoping review of genetic studies of treatment-resistant schizophrenia

Abstract

Background

This scoping review synthesized the current evidence on the genetic architecture of treatment-resistant schizophrenia (TRS). TRS is typically characterized by an inadequate response to antipsychotic treatment in individuals with schizophrenia. Approximately 30% of individuals with schizophrenia develop treatment resistance, which is associated with greater disability, poorer prognosis, and increased mortality compared to treatment-responsive schizophrenia. Numerous studies have explored various genetic aspects of TRS; therefore, a scoping review was needed to summarize findings and consistent patterns, clarify the current level of understanding, and highlight remaining knowledge gaps.

Methods

A systematic search was conducted in PubMed up to March 2025. This scoping review followed PRISMA-ScR guidelines. Studies were included if they reported on genetic factors of TRS and its related constructs (clozapine resistance). Data on treatment resistance, study design, population characteristics, and genetic findings were extracted and synthesized.

Results

A total of 102 studies were included. Definitions of TRS varied across studies, with most using proxies such as clozapine use, antipsychotic polypharmacy, or they used presence of symptoms despite antipsychotic treatment. Most studies compared TRS with treatment-responsive schizophrenia and predominately included participants of European genetic ancestry. Genetic findings spanned common variants (e.g., identified from genome-wide association studies or in cumulative measures such as polygenic risk scores [PRS]), rare variants, and copy number variants and functional genomics such as gene expression and epigenetic markers. Common variants dominated the literature but explained only a small proportion of TRS liability. Higher schizophrenia PRS, specific rare variants, and copy number variants were associated with TRS, while TRS-specific PRS remain in development. GWAS largely focused on schizophrenia broadly, with substantial genetic overlap between TRS and schizophrenia. Transcriptomic and epigenomic data provide additional but limited insights, often confounded by drug exposure.

Conclusion

Heterogeneous TRS definitions and limited ancestry diversity constrain progress, and robust TRS-specific genetic markers remain scarce. Harmonized criteria and larger, diverse cohorts are needed. Integrating genetic, epigenetic, and clinical data could improve early risk identification and guide precision treatment strategies.

1. Introduction

Schizophrenia is a psychiatric disorder affecting ∼1% of the global population (Solmi et al., 2023; Saha et al., 2005). This illness is characterized by positive symptoms (e.g., hallucinations, delusions), negative symptoms (e.g., social withdrawal, blunted affect), and deficits in cognition (e.g., impairments in attention, memory, and executive function). These symptoms can substantially impair daily functioning and often require lifelong management with antipsychotic medications, psychotherapy, and support services. Approximately 30% of individuals with schizophrenia do not respond to ≥2 antipsychotic medication despite multiple trials at adequate dose and duration (Kane et al., 1988; Conley and Kelly, 2001). This subset of individuals is classified as having treatment-resistant schizophrenia (TRS). TRS is associated with significant clinical and societal burden, including more frequent hospitalizations (Kennedy et al., 2014), increased risk of suicide and mortality (Wimberley et al., 2017; Wimberley et al., 2016), greater functional impairment (Millgate et al., 2022), and reduced quality of life (Kennedy et al., 2014) compared to treatment-responsive schizophrenia. Clozapine is currently the only antipsychotic specifically approved for TRS, and while it is effective at reducing hospitalizations and suicide rates (Land et al., 2017; Masuda et al., 2019; Stroup et al., 2016), it is also associated with strong adverse effects and requires rigorous blood monitoring (Tiihonen et al., 2017).

Schizophrenia is highly heritable, with pedigree-based estimates from family and twin studies of approximately 70%–80% (Sullivan et al., 2003). Of this genetic liability, approximately 25% is explained by common genetic variation in European populations, with 287 common risk loci identified to date, implicating synaptic, neuronal, and immune pathways (Liu et al., 2017; The Network and Pathway Analysis Subgroup of the Psychiatric Genomics Consortium, 2015; Trubetskoy et al., 2022). The genetics of treatment response in schizophrenia are likely related, but not identical, to the genetics of schizophrenia risk itself, as this has been observed in other psychiatric disorders such as depression (Xiong et al., 2025). Genetics research on TRS has employed a range of approaches, including common and rare variants, and copy number variants (CNV), and functional genomic investigations. Given the heterogeneity of schizophrenia and the diversity of methods and phenotypes used in TRS research, a synthesis of the genetic findings is needed to clarify what is known about the genetic architecture of TRS and to identify key knowledge gaps. A better understanding of the genetic architecture that contributes to poor treatment response in schizophrenia remains an important unmet need and could facilitate earlier identification of individuals at risk for TRS, support earlier or optimized interventions (e.g., clozapine initiation), and aid in the identification of novel pharmacological targets that may not be apparent when examining schizophrenia as a single, homogeneous entity.

We conducted a scoping review to summarize and assess the current literature on the genetic architecture of TRS and related outcomes, including treatment response and drug reactions, particularly to clozapine. More specifically, we aimed to 1) Examine how TRS has been operationally defined across genetic studies; 2) Synthesize the genetic findings for TRS; 3) Identify potential genetic factors associated with TRS; and 4) Summarize the current state of knowledge on TRS genetics and highlight unexplored areas. Conducting a scoping review enabled a broad exploration of the genetic underpinnings of TRS by incorporating studies focusing directly on TRS as well as those assessing treatment outcomes with clozapine. By consolidating these findings, our goal is to provide an overview of the current landscape of TRS genetics and contribute to an improved mechanistic understanding of treatment resistance within schizophrenia.

2. Materials and methods

2.1. Search strategy

The study protocol was registered on the Open Science Framework (registration doi: 10.17605/OSF.IO/938BR). Articles were identified through PubMed up to and including March 2025. PubMed was specifically chosen due to its comprehensive coverage of genetic and psychiatric research. The following search query was used: (“genetic” [All Fields] OR “genetical” [All Fields] OR “genetically” [All Fields] OR “genetics” [MeSH Subheading] OR “genetics” [All Fields] OR “genetics” [MeSH Terms]) AND (“schizophrenia, treatment resistant” [MeSH Terms] OR (“schizophrenia” [All Fields] AND “treatment resistant” [All Fields]) OR “treatment-resistant schizophrenia” [All Fields] OR (“treatment” [All Fields] AND “resistant” [All Fields] AND “schizophrenia” [All Fields]) OR “treatment resistant schizophrenia” [All Fields]). In addition to the database search, the reference lists of included articles were screened for additional relevant publications that were not captured in the initial search. The scoping review was conducted in accordance with the Preferred Reporting Items for Systematic Reviews: Extension for Scoping Reviews (PRISMA-ScR) (Tricco et al., 2018).

2.2. Study selection and data extraction

Article screening was performed using Covidence and involved initial title and abstract screening, full-text review, and data extraction. Screening was conducted by two independent reviewers, and any discrepancies were resolved by a third independent reviewer.

Articles were included if they reported on the underlying genetic components or architecture of TRS, included participants aged 18+, and were published in English language. Review articles, grey literature (e.g., preprints), theses or dissertations, case reports, and non-article publications were excluded. For each included study, data were extracted on study design, TRS definition, population characteristics, and genetic findings.

3. Results

A total of 543 articles were identified through PubMed (Figure 1). After removing duplicates and applying the screening protocol, 102 articles were included in the scoping review (Supplementary Table S1). Most of the studies were from Japan (n = 15), followed by United Kingdom (n = 11), and United States of America (n = 11, Figure 2). In terms of population structure (Figure 3), study participants were predominantly of European descent (n = 47 studies, 46%), followed by East Asian (24%), Middle Eastern (8%), Latin American (5%), South Asian (4%), and African (1%). Additionally, 12% of studies included multi-ancestry cohorts; however, even within these populations, European ancestry was the majority comprising 45%–90% of participants.

FIGURE 1.

PRISMA Diagram for the scoping review of the genetic architecture of treatment-resistant schizophrenia.

FIGURE 2.

Global distribution of included studies. The world map (left) illustrates the geographic distribution of studies included in the review, with color intensity indicating the number of studies per country. The accompanying bar plot (right) displays the corresponding frequencies by country.

FIGURE 3.

Cumulative ancestry-specific sample counts over time. The plot shows cumulative sample counts from included studies by ancestry between 2015 and 2025, with the x-axis representing publication year and the y-axis representing cumulative sample size.

The comparator groups varied slightly: the majority (n = 48, 69%) included individuals without TRS (non-TRS or treatment-responsive cases) as the comparator, and the remaining 21 studies (31%) specified healthy controls as a comparator, of which 10 also included non-TRS cases as a second comparator.

3.1. Defining treatment-resistant schizophrenia (TRS)

There was considerable variability in how TRS was defined across studies. Seven broad definition categories were identified: 1) Failure to respond to two or more antipsychotics (n = 27); 2) Continued worsening of symptoms despite antipsychotic treatment (n = 21); 3) Clozapine use (n = 30); 4) Polypharmacy use (n = 6); 5) Worsening symptoms and clozapine use (n = 5); 6) Worsening symptoms and polypharmacy use (n = 4); or 7) Worsening symptoms and a failure to respond to antipsychotics (n = 9). Approximately 30% (n = 29/102) incorporated clinical rating scales (e.g., Positive and Negative Syndrome Scale, Global Assessment of Functioning) to support their identification of TRS.

A subset of studies (n = 5) examined clozapine-resistance, often referred to as “ultra-resistant” (Mouaffak et al., 2011a) or “super-refractory” (Krzystanek et al., 2021) schizophrenia. Unlike TRS as defined above, ultra-resistant TRS was characterized by non-response to clozapine as defined by adequate duration (≥6 weeks to 6 months) and therapeutic dosing of clozapine (250–600 mg/day). Several studies also incorporated clozapine plasma concentration thresholds (≥350 ng/mL) or clozapine-to-norclozapine metabolic ratios as part of the definition, in addition to persistent positive symptoms assessed with clinical rating scales.

3.2. Genetic insights

The genetic findings related to TRS were grouped into three domains (Table 1): common variant studies (n = 75) (Li and Meltzer, 2014; Miyazawa et al., 2022; Naveen et al., 2020; Ji et al., 2008a; Hong et al., 2003; Teo et al., 2012; Tsai et al., 2001; Mouaffak et al., 2011b; Taheri et al., 2023; Jia et al., 2011; Takao et al., 2006; Escamilla et al., 2018; Hotta et al., 2011; Hwang et al., 2005; Hong et al., 2000; Bozina et al., 2008; Hajj et al., 2019; Willcocks et al., 2021; Kondo et al., 2003; Bosia et al., 2015; Piatkov et al., 2017; Lim et al., 2023; Ota et al., 2012; Arranz et al., 1998; Rajagopal et al., 2018; Taylor et al., 2016; Terzić et al., 2016; Terzić et al., 2015a; Zhang et al., 2013; Legge et al., 2017; Pinheiro et al., 2017; Xu et al., 2015; Konte et al., 2021; Frank et al., 2015; Werner et al., 2020; Terzić et al., 2015b; Souza et al., 2010; Kogure et al., 2021; Pae et al., 2006; Del Casale et al., 2026; Aytac et al., 2021; Kohlrausch et al., 2008; Eap et al., 2004; NCBI, 2026a; Rajkumar et al., 2012; Nakata et al., 2021; Taylor et al., 2017; Ruderfer et al., 2016; O’Connell et al., 2023; Zazueta et al., 2022; NCBI, 2026b; Patil et al., 2021; Inada et al., 2003; Ji et al., 2008b; Gasse et al., 2019; Facal and Costas, 2025; Talarico et al., 2022; Bani-Fatemi et al., 2019; Ozdemir et al., 2025; Cheng et al., 2023; Griffiths et al., 2023; Bilic et al., 2014; Aytac et al., 2022; Oishi et al., 2018) including genome-wide association studies (GWAS), candidate genes, polygenic risk scores (PRS); rare variant and CNVs (n = 10) (Legge et al., 2017; Ruderfer et al., 2016; Farrell et al., 2023; Morimoto et al., 2021; Räsänen et al., 2025); and functional genomics studies (n = 17) (Ozdemir et al., 2025; Räsänen et al., 2025; Funahashi et al., 2023; Menus et al., 2020; Moretti et al., 2018; Nakaz et al., 2017; Pérez-Rodríguez et al., 2023; You et al., 2020), which characterized how genetic variation influences cellular and molecular processes, including gene expression and epigenetic mechanisms. Common genetic variant studies were defined as those assessing variants with a minor allele frequency (MAF) of ≥5% of the population, whereas rare variants were defined as those with MAF <1%; however, these categories were not distinct.

TABLE 1.

Summary of genetic findings related to treatment-resistant schizophrenia (TRS).

Category	Summary
Common variant - candidate genes	Studies examined common genetic variants related to neurotransmitter signaling and antipsychotic drug metabolism in relation to TRS. Variants in dopaminergic, serotonergic, and glutamatergic pathway genes were associated with altered receptor function, neurotransmission, and antipsychotic response, including clozapine efficacy, while some suggested sex-specific effects. Variation in clozapine metabolism genes, particularly CYP enzymes, as well as safety-associated variants such as ACKR1, may contribute to variability in drug levels, treatment response, adverse effects, and the need for individualized prescribing
Common variant - genome-wide association studies	GWAS examined broad TRS-related phenotypes, including TRS case-control status, symptom severity in clozapine-treated individuals, clozapine plasma concentrations, and clozapine-induced neutropenia, with most conducted in ancestry-specific cohorts and few using multi-ancestry samples. The largest GWAS found TRS and non-TRS schizophrenia were highly genetically correlated but identified no TRS-specific loci. GWAS of clozapine pharmacokinetics and neutropenia more consistently identified significant loci, particularly in CYP and UGT metabolism genes and in ancestry-relevant neutropenia-associated loci, highlighting the importance of pharmacogenomics and ancestry in clozapine-related outcomes
Common variant - polygenic (risk) scores (PRS)	Studies show a higher schizophrenia PRS is modestly associated with TRS and influences clozapine use and response, with elevations in schizophrenia PRS may be linked to better clozapine outcomes, while the highest PRS levels tend to occur in individuals with the most severe and refractory illness. PRS for other traits show little association with TRS, though glutamatergic PRS may modify the impact of childhood adversity on cognitive outcomes in TRS.
Rare variants and copy number variants	Studies showed that TRS is associated with a higher burden of rare genetic variants than non-TRS. TRS individuals carry more rare damaging variants and CNVs, including schizophrenia- and neurodevelopment-associated CNVs. Variants affecting drug metabolism and transport genes may also influence antipsychotic response, suggesting a genetic contribution to treatment resistance
Functional genomics - gene expression	TRS showed distinct gene expression patterns compared with non-TRS and healthy controls, indicating biological differences as well as variability in study design and treatment exposure. Dysregulation primarily involved genes related to stress response, synaptic function, neurodevelopment, immune signaling, and regulatory non-coding RNAs. Several transcriptomic changes appeared influenced by antipsychotic treatment, particularly clozapine, affecting miRNAs, lncRNAs, and immune markers. Overall, these findings suggested that TRS comprised altered molecular pathways linked to synaptic regulation, neurodevelopment, and treatment response, although some effects may reflect underlying schizophrenia biology rather than treatment resistance alone
Functional genomics - epigenetic	Studies primarily reported DNA methylation changes in TRS. While some studies found no significant differences, others reported methylation patterns that distinguish TRS or predict clozapine exposure. Twin studies further suggested that clozapine itself can modify methylation. Overall, it remains unclear whether these changes reflect TRS biology or medication effects

3.2.1. Common genetic variant studies

3.2.1.1. Candidate genes

Fifty-eight studies examined associations between specific common genetic variants and TRS, with a focus on neurotransmitter systems (dopaminergic or serotonergic) or antipsychotic drug metabolism.

Variants in dopaminergic genes, including DRD1, DRD2, DRD3, DRD4, DAT, and DDC, were frequently linked to altered receptor function and dopaminergic signaling, contributing to diminished efficacy of dopamine-targeting treatments, including clozapine. The DRD3 rs6280 (Ser9Gly) polymorphism was associated with clozapine resistance in TRS, while several DRD2 SNPs were associated with variability in antipsychotic response (Casale et al., 2023). Variants in COMT, a key the dopamine-metabolizing gene that influences enzymatic activity, were linked to TRS. In females with TRS, low-activity COMT genotypes were more frequent, whereas high-activity COMT haplotypes were more common in treatment responders, suggesting a protective effect. In contrast, males with TRS showed a higher frequency of the Met allele at rs4680, particularly the Met/Met genotype, indicating that although reduced COMT activity is associated with TRS in both sexes, the specific genotype patterns differ between males and females (Sagud et al., 2018).

Genes in the serotonergic pathway, including HTR3A, HTR3B, HTR2A, HTR4, 5-HT1A, SLC6A4 (serotonin transporter), SERT-PR, SERT-in2, and 5-HT6 were frequently associated with TRS. Variants in these genes were found to disrupt the serotonin-dopamine balance critical for optimal antipsychotic efficacy and clozapine response. Similarly, glutamatergic genes such as GRIN1, GRIN2A, GRIN2B, SLC1A2, SLC6A9, GRIA1, GRM2, and GAD1, were identified in TRS in relation to cognitive deficits and dysregulation of excitatory and inhibitory neurotransmission.

Common variants in cytochrome P450 (CYP) enzymes such as CYP1A2, CYP3A4, CYP3A5, and POR (cytochrome P450 oxidoreductase), which play central roles in clozapine biotransformation (Demirbugen et al., 2023), were shown to influence plasma drug concentrations and therapeutic efficacy, contributing to inter- and intra-individual variability in treatment response in TRS (Demirbugen et al., 2023). Individuals who carried multiple rare damaging variants or clinically significant CNVs in CYP1A2 exhibited subtherapeutic clozapine responses, altered clozapine pharmacokinetics, and required higher clozapine doses (Kappel et al., 2024). Additionally, variants in CYP2C18, CYP2C19, and of UGT2 gene family were shown to influence the clozapine-to-norclozapine metabolic ratio, a clinically relevant marker for enzymatic activity and individualized dose adjustment (Okhuijsen-Pfeifer et al., 2022).

Beyond metabolism and efficacy, rs2814778 emerged as a safety-related pharmacogenetic variant. Individuals of African ancestry carrying this variant were associated with clozapine-induced neutropenia (odds ratio [OR] = 20.4, p = 3.44 × 10⁻⁷) and explained more variance in neutrophil counts than genetic ancestry alone, making it a key marker for ancestry-informed safe prescribing (Legge et al., 2019). In addition, genetic predispositions interacted with lifestyle factors (e.g., smoking and caffeine intake), influencing variability in clozapine metabolism and response.

3.2.1.2. Genome-wide association studies (GWAS)

Eight GWAS investigated broad aspects of TRS such as case-control status, symptom severity among clozapine-treated individuals, clozapine plasma concentrations, or clozapine-induced neutropenia. Two GWAS used multi-ancestry cohorts, and the remaining were conducted in ancestry-specific samples (four European, one African, and one Turkish cohort).

The largest included GWAS analyzed 10,501 TRS cases and 24,542 healthy controls, alongside a second GWAS of 20,325 non-TRS cases and 30,122 healthy controls (Pardiñas et al., 2022). Interaction analyses comparing effect sizes between TRS and non-TRS revealed a high genetic correlation (r = 0.966) and replicated known schizophrenia loci, but no TRS-specific associations were detected; SNP-based heritability for TRS was estimated at 1%–4% (Pardiñas et al., 2022). A smaller separate GWAS (478 TRS vs. 808 non-TRS cases) reported a suggestive association at rs79229764 within LINC00523 (p = 1.8 × 10⁻⁷) (Lenk et al., 2024).

Five GWAS concentrated on clozapine treatment in TRS. Focusing on symptom severity among clozapine-treated schizophrenia individuals, genome-wide analyses of 684 TRS cases from five cohorts reported a suggestive association between symptom severity and a variant in NFIB (p = 3.78 × 10⁻⁷) (Okhuijsen-Pfeifer et al., 2022). In relation to clozapine metabolism, a GWAS using 10,353 pharmacokinetic assays from 2,989 TRS cases (with >90% probability of European ancestry) identified four genome-wide significant loci in CYP and UGT, drug metabolism genes associated with clozapine and norclozapine plasma levels (Pardiñas et al., 2019). A subsequent cross-ancestry GWAS using 16,068 pharmacokinetic assays from 4,495 TRS cases revealed ancestry-specific differences in clozapine metabolism and identified eight associated loci, most of which had stronger effects in non-European populations including Southwest Asian, North African, Sub-Saharan African (Pardiñas et al., 2023).

Genome-wide studies on clozapine-induced neutropenia identified key genetic associations, while not reflecting TRS biology per se, provide clinically valuable insight into susceptibility of this adverse effect. One study of 66 neutropenia cases (absolute neutrophil count [ANC] ≤1,500/mm³) and 5,583 clozapine treated controls (maintained ANC ≥ 2,000/mm³ for at least 1 year of clozapine treatment) of European ancestry identified genetic variation in hepatic transporter genes SLCO1B3 and SLCO1B7 (OR = 4.32, p = 1.79 × 10⁻⁸), while also replicating the previously reported HLA-DQB1 variant (OR = 15.6, p = 0.015, positive predictive value = 35.1%) (Konte et al., 2021). Another GWAS of lowest ANC during clozapine treatment in 552 individuals of African ancestry identified two loci associated with low neutrophil count, with the strongest association observed at rs2814778 in ACKR1 (p = 4.21 × 10⁻²¹) (Legge et al., 2019).

Lastly, one GWAS of 84 individuals (31 TRS, 53 non-TRS) tested 1,178,234 SNPs using a main effect model for TRS status and an interaction model with childhood trauma. However, no associations reached genome-wide significance in either model (Koga et al., 2017).

3.2.1.3. Polygenic risk scores (PRS)

Eleven studies examined schizophrenia PRS in relation to TRS given the hypothesis that higher genetic burden for schizophrenia may increase the risk of treatment resistance. In a large study of 10,501 TRS cases and 20,325 non-TRS cases, schizophrenia PRS explained ∼2% of the variance in TRS liability, indicating a modest but significant association (Pardiñas et al., 2022). Several studies also assessed the relationship between schizophrenia PRS and clozapine treatment outcomes. Higher schizophrenia PRS was associated with greater clozapine dose requirements in large UK cohorts such as CLOZUK2 and CLOZUK3 (approximately 4,000 individuals), as well as in a Norwegian cohort of 417 individuals (Kappel et al., 2023). Among TRS individuals receiving clozapine, those in the highest schizophrenia PRS tertile were 1.94 times more likely to exhibit low symptom severity than those in the lowest tertile (p = 6.84 × 10⁻⁴), indicating a better response to clozapine in this subgroup (Okhuijsen-Pfeifer et al., 2022). In contrast, another study found that although individuals treated with clozapine had higher schizophrenia PRS (p = 0.02) than those who had never received clozapine, the highest schizophrenia PRS were observed among clozapine non-responders or extreme-TRS cases, particularly those with poor premorbid functioning, early illness onset, and severe illness progression (Okhuijsen-Pfeifer et al., 2022). The findings reflect that moderately elevated schizophrenia PRS may be associated with better clozapine response within TRS cohorts, whereas the highest PRS values tend to cluster in individuals with the most severe, and treatment-refractory illness.

Three studies have investigated PRS for non-schizophrenia traits in relation to TRS. A large population-based Swedish study (n = 2,997 TRS; n = 3,675 non-TRS) conducted by members of our group examined PRS for bipolar disorder, major depressive disorder, autism spectrum disorder, cognitive ability, and educational attainment, but found no significant associations with TRS (Kowalec et al., 2021). Similarly, another study assessed PRS for BMI, smoking behaviour, caffeine intake, and a range of psychiatric, cognitive, and personality traits in relation to clozapine dose among individuals with TRS (n = 4,459), and likewise reported no significant association (Kappel et al., 2023). The third study derived a glutamatergic PRS from variants in glutamate signaling genes in association with childhood adversity and IQ in TRS (n = 51) and non-TRS (n = 154). Although individuals with TRS had a slightly higher mean glutamatergic PRS compared to non-TRS, this difference was not statistically significant (Mohamed Saini et al., 2023). However, glutamatergic PRS significantly moderated the impact of childhood adversity on cognitive and structural outcomes in TRS, amplifying the negative association with IQ (β = −125.05, p = 0.0002), whereas no such moderation was observed in non-TRS.

3.2.2. Rare genetic variants and copy number variants (CNV)

A study of 112 individuals identified as having severe TRS (defined by a minimum of 5 years of continuous hospitalization) reported significantly greater enrichment of rare damaging variants in intolerant genes (loss-of-function variants OR = 1.91, missense variants OR = 2.90) compared to 218 non-TRS cases and 4,929 healthy controls (Zoghbi et al., 2021). Nearly half (48.2%) of the individuals with severe TRS carried at least one rare damaging variant compared to 29.8% and 25.4% in non-TRS individuals and healthy controls, respectively.

Multiple studies reported a higher burden of CNVs among individuals with TRS. One such study found a significant association between total genome-wide copy number duplication burden and treatment resistance (OR = 1.04; p < 0.01) between TRS and non-TRS cases (Martin and Mowry, 2016). A large study comparing TRS (n = 509) and non-TRS (n = 21,094) cases found that rare CNVs were significantly more prevalent in TRS–approximately 4% in TRS cases compared to 2.18% in non-TRS cases. In TRS cases, 9.2% had at least one CNV linked to neuropsychiatric risk and 4.7% carried CNVs associated with neurodevelopmental risk, although these CNVs are also enriched in schizophrenia in general, making their TRS-specificity uncertain. The most frequently reported CNVs included duplications at 15q11.2–q13.1 and 16p11.2, and deletions at 22q11.21 (Kushima et al., 2017). In addition, somatic CNVs were found at higher frequencies in TRS compared to non-TRS individuals (p = 0.03), affecting transporter gene ABCB11 associated with antipsychotic metabolism (Maury et al., 2023). One final study of TRS cases reported finding rare damaging variants, or clinically significant CNVs, in CYP1A2, which altered clozapine pharmacokinetics (Kappel et al., 2024).

3.2.3. Functional genomic research

3.2.3.1. Gene expression

Across many studies, TRS was associated with distinct gene expression profiles in peripheral blood and post-mortem brain tissue compared to non-TRS and healthy controls. The direction and magnitude of effects varied likely reflecting heterogeneity in TRS definitions and study designs. These transcriptional alterations were modulated by regulatory elements, including microRNA (miRNAs) and long non-coding RNAs (lncRNAs), particularly in the context of antipsychotic exposure with clozapine.

Transcriptomic studies highlighted dysregulation of stress-related genes (e.g., SIRT1, TRIM28) and synaptic regulators (e.g., CNR1, UFD1L, AKT1, DICER1) in TRS compared to non-TRS cases (Pérez-Rodríguez et al., 2023). In monozygotic twins discordant for clozapine response, protocadherin gene alterations in iPSC-derived neurons were identified, implicating synaptic remodeling in clozapine efficacy (Nakaz et al., 2017). Several schizophrenia-related miRNAs were upregulated in TRS relative to non-TRS, with miR-181b-5p, miR-195-5p, and miR-301a-3p consistently dysregulated and associated with treatment response via silencing of antipsychotic receptor genes (Alacam et al., 2016). Additional candidates include clozapine-regulated miR-675-3p, linked to neuronal function via plasma exosomes, as well as miR-218-5p and miR-1262, which have been reported as early blood-based biomarkers for TRS, and miR-143-3p, induced by NRG1/mTOR signaling, as a putative mediator of olanzapine response (Sun et al., 2022).

BDNF, a critical neurodevelopmental regulatory gene, showed region-specific expression changes in brain and variable serum levels based on antipsychotic treatment type (Badrlou et al., 2021). Similarly, elevated expression of BDNF and associated lncRNAs (e.g., BDNF-AS, MIR137HG, MIAT, PNKY) were reported in individuals with TRS compared to healthy controls, which more so highlight schizophrenia biology and not necessarily that of treatment resistance (Badrlou et al., 2021). Another important neurodevelopmental and synaptic function, the NRG–ErbB signaling pathway, showed increased transcript expression in TRS; particularly, higher P70S6K mRNA levels (p _corrected = 0.018) were negatively correlated with illness duration (Mostaid et al., 2017). The last study found peripheral TNF-α mRNA level reductions in TRS (p < 0.001) compared to healthy controls. The levels increased with clozapine treatment; however, as the comparison was made with healthy controls, the contribution of underlying schizophrenia biology to these findings cannot be completely excluded (Mostaid et al., 2018; Kluge et al., 2009).

3.2.3.2. Epigenetics

DNA methylation was a focus for multiple studies. One genome-wide analysis found no significant methylation sites between TRS (n = 43) and non-TRS (n = 66) individuals (De Luca et al., 2023). In contrast, another study using a two-phase (discovery and validation) design identified six probes that distinguished TRS (n = 48) from non-TRS (n = 48) individuals, with a methylation risk model achieving 88.3% accuracy (area under the curve = 0.95) (Lu et al., 2023).

An epigenome-wide study (EWAS) of >431,000 methylation sites identified seven differentially methylated positions associated with clozapine exposure in TRS (p < 9 × 10⁻⁸). These sites showed a mean methylation difference of 1.47% (SD = 0.24%), with hypermethylation observed only in TRS individuals (p = 0.0156) (Hannon et al., 2021) compared to non-TRS individuals. Another EWAS comparing TRS (n = 67) with non-TRS individuals not treated with clozapine (n = 314) generated an epigenetic score associated with clozapine use (R² = 0.022, p = 3.85 × 10⁻³) (Ki et al., 2024). Lastly, a study of monozygotic twins with TRS but discordant for clozapine response reported methylation differences between responders and non-responders. Clozapine responders showed a higher proportion of differentially methylated genes related to neuronal and synaptic functions (35.7%) compared to non-responders (6.7%). Increased methylation in the MECP2 promoter, alongside reduced expression following clozapine treatment in responders, was also reported (Kikuchi et al., 2021). Even within a monozygotic twin pair, where genetic background is controlled, the findings showed substantial clozapine-related methylation changes, underscoring the challenge of isolating TRS-specific epigenetic changes. Together, these observations also highlight the need for study designs with pre-treatment baselines to clarify whether such methylation differences represent underlying TRS mechanisms or are primarily treatment-induced.

4. Discussion

This scoping review summarized the current literature on the genetic architecture underlying TRS, examining how TRS has been identified and defined across studies and to integrate findings across multiple genomic domains. The included articles spanned GWAS, candidate genes, PRS, rare variants and CNVs, gene expression, and epigenetic research. Together, these findings suggest that TRS has a measurable, but complex and heterogeneous, genetic component and that key knowledge gaps must be addressed to advance mechanistic understanding and potential clinical translation.

4.1. Definition of TRS

Establishing the genetic architecture of any complex trait requires having a clear and consistent definition. The Treatment Response and Resistance in Psychosis working group developed consensus guidelines outlining minimum and optimal criteria for defining TRS across several clinical areas including current symptoms, adequacy of treatment, symptom domains, and time course (Howes et al., 2017). In practice, however, applying even the minimum consensus criteria in large-scale genetic studies may be time- and resource-intensive. This was reflected in the scoping review, where the definition and identification of TRS ranged from detailed clinical criteria to proxy measures such as clozapine use or polypharmacy. Nevertheless, across studies, most definitions of TRS converged on three criteria–failure to respond to antipsychotic trials (though the adequacy of such treatment was not always evaluated), persistence or worsening of symptoms over time, and the use of clozapine or polypharmacy. These shared criterions provide a starting point for harmonizing TRS definitions in future genetic research.

4.2. Definition of TRS and the potential for misclassification

An important limitation is the potential for treatment resistance misclassification due to substantial heterogeneity in how TRS was defined across studies. Clozapine use should be interpreted as an imperfect proxy for TRS, as its initiation is influenced by geographic and temporal variation in clinical practice; differences in access, monitoring requirements, local guidelines, and prescribing preferences may affect who receives clozapine and when, potentially introducing misclassification of the treatment-resistant phenotype. Using clozapine as a proxy for TRS may also exclude individuals with delayed access or barriers to clozapine treatment (leading to false negatives), while symptom persistence despite antipsychotic treatment could reflect partial response rather than true resistance (leading to false positives). This differential misclassification has important implications for power, effect size estimation, and the interpretation of null findings.

4.3. Selection of comparator groups

Selecting the appropriate comparison groups is equally important for interpreting genetic and biological findings. Comparing TRS to treatment-responsive schizophrenia can help identify mechanisms specifically related to non-response and treatment efficacy. Conversely, studies using only healthy controls contribute to our understanding of schizophrenia genetics more broadly, but not treatment resistance biology. Across studies, nearly 70% included individuals without TRS (non-TRS or treatment-responsive cases) as the comparator group.

Evidence from other psychiatric disorders support the distinction between genetic liability for illness risk and treatment response, with only partial overlap in conditions such as major depressive disorder (Fabbri et al., 2021). Future research should be deliberate and transparent in the selection of comparison groups and, where possible, include both treatment-responsive schizophrenia and healthy controls to clarify which differences are specific to treatment resistance.

4.4. Insights

Evidence from this scoping review indicates that the genetic architecture of TRS is only beginning to be delineated. More than 50% of the included studies focused on common genetic variants, yet findings explained a relatively small proportion of genetic liability for TRS. Most GWAS to date have concentrated on schizophrenia in general, rather than specific subsets, such as TRS. Consequently, TRS-specific PRS are still in early phases of development and existing work largely extrapolates from schizophrenia case-control PRS, which offers a rough approximation of treatment resistance risk. TRS and schizophrenia share a high genetic correlation as many of the genes and variants associated with TRS largely overlap with those for schizophrenia (Trubetskoy et al., 2022). In addition, clozapine-related outcomes, such as drug response or adverse effects, likely represent a secondary rather than direct genetic correlate of TRS. Resultant effects, such as neutropenia, are better understood as moderators of clozapine tolerability and treatment continuation. Given that clozapine is central to TRS management, genetic liability to neutropenia may still be highly clinically relevant but should be interpreted regarding safe clozapine use, rather than mechanisms of treatment resistance.

The included studies also pointed to a role for rare, high-impact genetic variation, but this area remains nascent. Rare variant analyses in TRS reported finding significantly greater burden of damaging variants in genes intolerant to loss-of-function or deleterious missense mutations, suggesting that the genome of severe and persistent refractory schizophrenia may carry highly penetrant variants, yet replication studies are needed. Most existing rare variant data come from exome-focused designs since systematic whole-genome sequencing efforts that integrate rare intronic and regulatory variants are largely absent. This represents a clear space for future research.

Transcriptomic and epigenomic data added further, yet still fragmented, insight into the genetic architecture of TRS. Differential gene expression profiles in peripheral and brain-derived tissues have been reported in TRS when compared with healthy controls, but specificity relative to treatment-responsive schizophrenia remains unclear. Similarly, TRS-associated differentially methylated loci often overlapped with methylation changes observed following clozapine initiation, raising the question of whether these epigenetic patterns index treatment resistance itself, cumulative drug effects, or a combination of both, highlighting the complexity of interpreting epigenetic signatures in this context.

Genetic pleiotropy with TRS comorbidities is an emerging area of investigation. Although schizophrenia exhibits genetic overlap with other conditions, current research assessing non-schizophrenia PRS in relation to TRS outcomes is sparse and largely non-significant. For instance, although polygenic liability for smoking initiation has been linked to schizophrenia risk, evidence for a robust or specific association with TRS remains limited (Peterson et al., 2021).

At the extreme end of the spectrum, ultra-TRS, where individuals fail to respond to clozapine and multiple augmentation strategies, has been understudied. Genetic, transcriptomic, and epigenomic data in this subgroup remains limited. This may reflect the rarity of ultra-TRS and the challenges associated with studying these individuals. Consequently, it remains unclear whether ultra-TRS reflects a quantitative extreme of the same liability spectrum or a qualitatively distinct subgroup with unique biological mechanisms. Addressing this question will require dedicated, well-powered studies of ultra-TRS, ideally incorporating longitudinal designs and multi-omic profiling.

A further limitation of the existing literature is the generally small sample sizes, with many studies including fewer than 200 participants. Null findings in underpowered studies may reflect false negatives rather than a true absence of association, while positive findings from small studies, particularly those lacking replication, should be interpreted cautiously due to an increased risk of false positives. Publication bias also cannot be excluded, as small candidate gene studies reporting positive findings may be more likely to be published than those reporting null results.

4.5. Ancestry and sex considerations

Most included articles were conducted in populations of predominantly European ancestry, with few modestly sized cohorts representing African and Asian populations. This likely reflects both the lower prevalence of TRS overall, which complicates recruitment of adequately powered multi-ancestry cohorts, and the methodological demands of TRS studies, including detailed longitudinal phenotyping, that concentrate research in high-income settings with established infrastructure. This imbalance has important analytical implications. Differences in population structure require careful control for population stratification within and across studies, typically through ancestry-informed design and adjustment to reduce the risk of confounding. Additionally, ancestry-specific differences in allele frequencies may affect the power to detect associations, particularly for low-frequency variants, and differences in linkage disequilibrium patterns can influence both locus discovery and the transferability of findings, including fine-mapping and polygenic prediction, across populations. Together, these limitations reduce the generalizability of current TRS genetic findings and hinder identification of ancestry-specific risk or protective factors. Broadening ancestral diversity in TRS research will be essential for developing more inclusive etiological models and for supporting equitable implementation of precision medicine.

Sex-specific characteristics in TRS are also underexplored in genetic research approaches. A small number of candidate gene studies reported sex-stratified associations, suggesting that genetic contributions to TRS may differ between males and females. Considering there are established sex-differences in prevalence, clinical presentation, and symptom severity within schizophrenia (Ferrer-Quintero et al., 2021; Li et al., 2016), integrating sex-stratified analyses in future TRS genetics studies may help clarify distinct clinical presentations of treatment-resistant phenotypes and differentiate TRS from treatment-responsive schizophrenia.

4.6. Limitations

This scoping review has limitations. First, article inclusion was restricted to English-language, which may have led to the omission of relevant studies published in other languages. Second, although the literature search was limited to PubMed, which includes both MEDLINE-indexed and non-MEDLINE biomedical literature, this may have reduced coverage of studies indexed exclusively in other databases. Third, substantial heterogeneity in TRS definitions, study designs, and outcome measures limited the extent to which findings could be quantitatively synthesized, though this variability reflects real-world clinical practice and provides insight into how TRS is operationalized across research. Lastly, as a scoping review, formal assessment of the methodological quality or risk of bias of individual studies was not conducted, and small, underpowered studies with positive findings may be overrepresented in the published literature.

4.7. Future directions

The available evidence indicates that TRS has a genetic architecture that both overlaps with and differs from that of schizophrenia risk. Rare variants and CNVs may be particularly relevant for a subset of individuals with severe or early-onset TRS, whereas common variant burden (as captured by schizophrenia PRS) appears to index genetic liability and illness severity rather than TRS-specific risk.

Despite increased attention in research, important questions remain. It is not yet clear whether TRS is genetically distinct from schizophrenia more broadly, or to what extent genetic markers reflect treatment resistance itself versus the effects of long-term antipsychotic exposure, illness severity and chronicity, or environmental stressors (e.g., trauma, substance use). Longitudinal evidence of gene expression and epigenetic variation in relation to treatment outcomes and symptom trajectories are also poorly understood.

Future research should prioritize large, ancestrally diverse TRS cohorts with standardized definitions aligned as closely as possible with consensus criteria (Howes et al., 2017). This could involve adopting stage-based or tiered TRS phenotypes (e.g., failure of two adequate antipsychotic trials, clozapine eligibility, clozapine non-response), clearly distinguishing treatment resistance from intolerance or non-adherence and applying post hoc phenotype mapping to harmonize heterogeneous cohorts for genetic analyses. Explicitly incorporating longitudinal designs with pre-treatment baselines will be essential for separating TRS trait markers from state effects and from molecular signatures driven by treatment, especially clozapine. Multi-omics integration approaches such as Mendelian randomisation, colocalisation analyses, transcriptome-wide association studies (TWAS), and gene-by-environment interaction models will be critical for distinguishing causal mechanisms from correlated signals. Whole-genome sequencing may further improve detection of rare variants associated with treatment resistance, potentially revealing novel biological pathways and therapeutic targets (Singh et al., 2024). Incorporating sex-stratified analyses and systematically capturing environmental exposures will also help clarify how genetic risk interacts with clinical and contextual factors to shape treatment outcomes for individuals with TRS.

4.8. Conclusion

This scoping review provides, to our knowledge, the first integrated overview of the genetics architecture of TRS across common variants, rare variants and CNVs, gene expression, and epigenetic studies. The findings highlight substantial complexity and heterogeneity in both TRS definitions and genetic results, with no single genetic profile emerging as definitive for TRS. Instead, TRS appears to arise from multiple, partly overlapping biological mechanisms, some shared with schizophrenia more broadly and others potentially specific to treatment resistance or clozapine response.

Future research incorporating trait-based phenotyping, longitudinal follow-up, and multi-omic approaches in diverse populations will be essential for refining the genetic architecture of TRS. A deeper understanding of TRS biology remains needed to meaningfully improve early identification of individuals at risk, optimize treatment strategies, and advance precision psychiatry for individuals living with schizophrenia.

Funding Statement

The author(s) declared that financial support was received for this work and/or its publication. This study was funded by the National Institute of Mental Health (NIMH; MH123724) and the University of Manitoba.

Author contributions

UD: Methodology, Visualization, Conceptualization, Data curation, Investigation, Writing – review and editing, Formal Analysis, Writing – original draft. HR: Investigation, Writing – original draft, Formal Analysis, Data curation, Methodology, Writing – review and editing. ME: Writing – review and editing, Data curation. MS: Data curation, Writing – review and editing. AR: Data curation, Writing – review and editing. MH: Data curation, Writing – review and editing. YL: Writing – review and editing, Supervision. KK: Supervision, Methodology, Conceptualization, Data curation, Funding acquisition, Resources, Writing – review and editing.

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

The handling editor REP declared a past co-authorship with the author LY.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fgene.2026.1802472/full#supplementary-material