Combating antimicrobial resistance in malaria, HIV and tuberculosis

Abstract

Antimicrobial resistance poses a significant threat to the sustainability of effective treatments against the three most prevalent infectious diseases: malaria, human immunodeficiency virus (HIV) infection and tuberculosis. Therefore, there is an urgent need to develop novel drugs and treatment protocols capable of reducing the emergence of resistance and combating it when it does occur. In this Review, we present an overview of the status and underlying molecular mechanisms of drug resistance in these three diseases. We also discuss current strategies to address resistance during the research and development of next-generation therapies. These strategies vary depending on the infectious agent and the array of resistance mechanisms involved. Furthermore, we explore the potential for cross-fertilization of knowledge and technology among these diseases to create innovative approaches for minimizing drug resistance and advancing the discovery and development of new anti-infective treatments. In conclusion, we advocate for the implementation of well-defined strategies to effectively mitigate and manage resistance in all interventions against infectious diseases.

Introduction

The rise of drug resistance presents a significant challenge in the treatment of infectious diseases. Although efforts to address this threat with strategies such as the continual development of new drugs and more effective combination regimens have achieved some success, it is widely acknowledged that the further proliferation of antimicrobial resistance could have catastrophic consequences for both global health and economies^1,2. Therefore, the ongoing development of these strategies is of paramount importance.

Prior to the onset of the COVID-19 pandemic, malaria, HIV infection and tuberculosis (TB) were responsible for the highest annual death tolls worldwide, and resistance to current therapies remains a persistent challenge for all three of these diseases.

Malaria, a parasitic infection caused by various Plasmodium spp., resulted in more than 600,000 deaths in 2021 (Box 1)³. The Plasmodium life cycle is intricate, involving the development of up to 10¹² parasites in a patient’s blood during the peak of infection. Resistance to artemisinin, the cornerstone of current antimalarial treatment (Table 1), was initially believed to have low potential for development due to the drug’s pleiotropic and rapid mode of action⁴. However, extensive usage of this antimalarial drug and its derivatives has led to resistance in parasites at early stages of the erythrocytic cycle, resulting in delayed parasite clearance. This partial resistance first emerged in Southeast Asia two decades ago⁵, and has more recently been observed in Africa⁶, with the constant threat of complete resistance to artemisinin still looming.

Box 1 |. Epidemiology of malaria, HIV infection and tuberculosis.

Malaria

With 247 million reported cases and 619,000 estimated deaths in 2021³, nearly half of the world population is at risk of malaria, but those disproportionally affected are in sub-Saharan Africa. Southeast Asia, the Eastern Mediterranean, the Western Pacific and the Americas are also considered regions at risk, although to a lower extent³. This inequality of distribution is both geographic and demographic: children younger than 5 years old are the most vulnerable group, followed by pregnant women and patients infected with human immunodeficiency virus (HIV)³. To date, five species of the protozoan pathogen Plasmodium are known to infect human and cause malaria: Plasmodium falciparum (up to 90% of reported cases), Plasmodium vivax (2% of cases), Plasmodium ovale, Plasmodium malariae and Plasmodium knowlesi.

HIV infection

Despite tremendous progress over the last two decades — mainly driven by the ambitious ‘90–90–90’ target (90% of infected persons aware of their diagnosis; 90% of diagnosed patients started on ART; 90% of treated patients showing viral suppression) set by the Joint United Nations Program on HIV/AIDS¹⁹⁴ — the World Health Organization (WHO) still estimated 1.5 million new infections and more than 37 million people living with HIV in 2020, linked to approximately 680,000 deaths from HIV-related causes¹⁹⁵. AIDS prevalence is unequally spread worldwide; Africa is significantly affected, followed to a much lesser extent by the Americas. HIV acquisition also greatly varies within the worldwide population and HIV prevalence is higher amongst men who have sex with men (23%), sex workers (11%), intravenous drug users (9%), transgender women (2%) and sex workers’ clients and the sex partners of these populations (20%)¹⁹⁶. HIV-1 is reported to currently infect approximately 40million individuals whereas HIV-2, a primate lentivirus that shares about 40–50% amino acid identity with HIV-1, is estimated to infect 1–2 million individuals, most of whom live in West Africa¹⁹⁷.

Tuberculosis

Mycobacterium tuberculosis (Mtb), the bacterium that causes tuberculosis (TB), is estimated to infect one-quarter of the world population¹⁹⁸. In 2021, 10.6million cases were estimated, resulting in 1.3 million deaths¹⁹⁹. Adult men accounted for more than 56% of the cases in 2019, whereas women represented about a third and children 11%. Recognized risk factors include living with HIV^76,200,201, undernutrition²⁰², indoor air pollution^202,203, type 2 diabetes mellitus²⁰⁴ and, to a lower extent, excessive alcohol use and smoking^203,205,206. TB is a disease predominantly affecting low and middle-income countries, mainly Southeast Asia, Africa and the Western Pacific. It also marginally impacts the Eastern Mediterranean, the Americas and Europe.

Table 1.

Current WHO-recommended therapies for malaria, HIV infection and tuberculosis

Population	First-line regimens	Alternative regimens
Malaria 12,184
Plasmodium falciparum uncomplicated malaria in adults and children	Any of the ACTs: ARTM–LUM, ARTS–AMQ, ARTS–MF, DHA–PIP, ARTS–SP, ARTS–PYN, 3-day course	Any of the ACTs in second or third-line regimen
Plasmodium vivax, Plasmodium ovale, Plasmodium malariae, Plasmodium knowlesi or P. falciparum uncomplicated malaria in adults, children, and lactating and pregnant women (second and third trimesters)	ARTM–LUM, ARTS–AMQ, ARTS–MF, DHA–PIP, ARTS–SP, 3-day course	Second or third-line regimen ACTs for all species; quinine, only if effective alternatives not available
P. falciparum uncomplicated malaria in pregnant women (first trimester)	ARTM–LUM, 3-day course	ARTS–AMQ, ARTS-MF and DHA–PIP when ARTM–LUM is not recommended or available.
P. vivax, P. ovale, P. malariae or P. knowlesi uncomplicated malaria in pregnant women (first trimester) in non-CQ resistance areas	ARTM–LUM in CQ resistance area or 3-day course of CQ in non-CQ resistance area	QN (in CQ resistance area), 7-day course
P. vivax and P. ovale relapses (except in G6PD deficiency, pregnant and lactating women, and infants aged <6 months)	PQ, 7-day or 14-day course; TF, single dose	In G6PD deficiency, PQ weekly for 8 weeks In pregnant or breastfeeding women, weekly chemoprophylaxis with CQ until delivery and breastfeeding completed
P. falciparum severe malaria	ARTS parenteral followed by full course of ACT	ARTM intramuscular or QN parenteral followed by full course of ACT
HIV infection 52,185
Adults and adolescents	DTG and two NRTIs (TDF + 3TC), fixed-dose combination	EFV and two NRTIs (TDF + 3TC), fixed-dose combination
Children (aged >4 weeks)	DTG and two NRTIs (ABC + 3TC), fixed-dose combination	ABC + 3TC + LPV/r or TAF + 3TC (or FTC) +DTG
Neonates	RAL and two NRTIs (RAL + AZT or ABC + 3TC), fixed-dose combination	NVP and two NRTIs (NVP + AZT + 3TC), fixed-dose combination
Tuberculosis 186
Drug-sensitive TB 187,188	2HRZE/4HR (2 months of INH + RIF +PZA + EMB, followed by 4 months of INH + RIF)	INH + RPT +PZA + MFX, 4-month course
Drug-resistant TB (MDR/RIF-resistant)	BPaL(M) regimen (BDQ + PMD + LZD (+ MFX, in patients susceptible to fluoroquinolones)), 6-month course	Oral regimen: 6 months of BDQ with 4 months of LFX/ MFX + ETH + EMB + INH (high dose) +PZA + CFZ, followed by 5 months of LFX/MFX + CFZ + EMB +PZA

ABC, abacavir; ACT, artemisinin combination therapy; AMQ, amodiaquine; ARTM, artemether; ARTS, artesunate; AZT, zidovudine; BDQ, bedaquiline; BPaL, bedaquiline–pretomanid–linezolid; CFZ, clofazimine; CQ, chloroquine; DHA, dihydroartemisinin; DTG, dolutegravir; EMB, ethambutol; ETH, ethionamide; EFV, efavirenz; G6PD, glucose-6-phosphate dehydrogenase; FTC, emtricitabine; HIV, human immunodeficiency virus; INH, isoniazid; LFX, levofloxacin; LP, lopinavir; LPV/r, ritonavir-boosted lopinavir; LZD, linezolid; LUM, lumefantrine; MDR, multidrug-resistant; MF, mefloquine; MFX, moxifloxacin; NNRTI, non-nucleoside reverse transcriptase inhibitor; NRTI, nucleoside reverse transcriptase inhibitor; NVP, nevirapine; PIP, piperaquine; PMD, pretomanid; PQ, primaquine; PYN, pyronaridine; PZA, pyrazinamide; QN, quinine; RAL, raltegravir; RIF, rifampicin; RPT, rifapentine; SP, sulfadoxine–pyrimethamine; 3TC, lamivudine; TAF, tenofovir alafenamide; TB, tuberculosis; TDF, tenofovir disoproxil fumarate; TF, tafenoquine; WHO, World Health Organization. TDF and TAF are prodrugs of tenofovir (now often abbreviated TFV). 3TC and FTC are cytosine analogues that are generally considered interchangeable.

Caused by the human immunodeficiency virus (HIV), AIDS has been responsible for more than 30 million deaths over the last four decades (Box 1). Extensive research has yielded a comprehensive under-standing of the HIV life cycle, facilitating the development of multiple classes of antiviral drugs (Table 1). These drugs can effectively control HIV replication for many years when combined into tailored, highly active antiretroviral therapy regimens, typically comprising two or three drugs from different classes. However, the elimination of HIV remains unattainable due to the existence of latent viral reservoirs. The emergence of resistant viral variants poses an ongoing challenge for patients receiving long-term antiretroviral regimens and for the overall effectiveness of antiretroviral therapies in newly infected patients.

TB, caused by the bacterium Mycobacterium tuberculosis (Mtb), ranks among the top ten causes of death globally (Box 1). Multiple classes of antibiotics with anti-TB activity have been developed (Table 1). However, owing in part to the complexity of the biology of Mtb, which can persist in a latent form, effective treatment is challenging even for drug-susceptible TB infections. Standard treatment for drug-susceptible TB involves four drugs and 6 months of therapy⁷. Furthermore, the situation worsens for drug-resistant forms of the disease; previous regimens were lengthy (up to 2 years), poorly tolerated, led to severe side effects and only cured 54% of patients with multidrug-resistant (MDR) TB⁷ and 34% of patients with extensively drug-resistant (XDR) TB⁸. These regimens are now being replaced by therapies based on bedaquiline, such as the 6-month bedaquiline–pretomanid–linezolid (BPaL) combination, which demonstrated 90% efficacy in clinical trials^7,9.

In this Review, we provide an overview of the characteristics (Box 2) and molecular mechanisms of drug resistance in malaria, HIV infection and TB, as well as current strategies to combat resistance for each of these diseases. Additionally, we explore the development of future strategies to address resistance, emphasizing the potential for cross-fertilization of knowledge between these therapeutic areas that share some common traits in the evolution of drug resistance. We centre our attention on the development of molecular resistance and its transmission within pathogen populations, and will not delve extensively into host-related resistance or dormancy and drug tolerance.

Box 2 |. Clinical resistance definitions and status.

Malaria

Clinical resistance to antimalarial drugs can occur when a drug loses potency and its efficacy is no longer sufficient to clear the total parasite load in patients. However, with endoperoxides such as artemisinin, no loss of potency is observed. Instead, their maximal effect linked to a given minimal parasiticidal concentration is lower. As a result, the World Health Organization (WHO) specifically defined artemisinin partial resistance as delayed parasite clearance following treatment with artemisinin-based monotherapy or with artemisinin combination therapy (ACT). Although the ACTs have remained efficacious so far, the recent emergence in Africa of isolates containing de novo mutations in the kelch13 gene linked to artemisinin partial resistance is of concern and calls for action.

Plasmodium parasites are qualified as multidrug-resistant (MDR) if they are resistant to more than two antimalarial compounds from different chemical classes. Resistant parasites for most current antimalarials have emerged^207,208 (Fig. 2), and MDR has led to clinical failures^209,210. Although lumefantrine and pyronaridine have remained exceptions so far, a longitudinal study performed between 2015 and 2021 in northern Uganda revealed a decreased susceptibility of the parasites to lumefantrine²¹¹.

HIV infection

Because human immunodeficiency virus (HIV) drugs are non-curative, HIV drug resistance is simply defined as the inability of drugs to block viral replication. Several types of resistance can occur: acquired resistance when mutations appear in treated patients, transmitted resistance when uninfected individuals become infected by a resistant virus and pretreatment resistance occurring in individuals having received preventive treatments²¹².

Resistance threatens the Joint United Nations AIDS ‘90–90–90’ goal, and in 2017 a 5-year Global Action Plan was aimed at preventing, monitoring and responding to resistance²¹³ and this was then extended to 2024²¹⁴. Meanwhile, three countries reported acquired drug resistance of up to 29% during 2014–2018. In patients lacking viral suppression, the prevalence of resistance to non-nucleoside reverse transcriptase inhibitors (NNRTIs) was 50–97%, and resistance to nucleoside reverse transcriptase inhibitors (NRTIs) was 2–91%. Dual class resistance to both targets correlated with failure of NNRTI-based first-line antiretroviral therapy²¹².

Concomitantly, 12 countries reported pretreatment resistance to the NNRTIs nevirapine and efavirenz. Additionally, during 2012–2018, nine sub-Saharan countries reported that more than half of infected infants carried a virus resistant to NRTIs. Recent independent studies also reported pretreatment^101,215, transmitted²¹⁶ or acquired resistance^217,218 globally. Pretreatment resistance to integrase strand transfer inhibitors (INSTIs) during 2014–2020 was extremely low (prevalence of 0.6%214), emphasizing the recently recommended transition to INSTI-based therapy^52,214. In the United States, the proportion of persons with transmitted resistance to one or more antiretrovirals exceeds 15%²¹⁹, whereas in many other regions, including Europe, Latin America and several countries in sub-Saharan Africa, transmitted resistance rates have generally been reported to be about 10%^214,220,221.

Tuberculosis

For anti-tuberculosis (anti-TB) drugs, the definition of the level of resistance depends on the number and nature of the drugs to which Mycobacterium tuberculosis (Mtb) is resistant. Resistance to all first-line anti-TB drugs is reported in virtually all countries²⁰⁰, and is usually due to inadequate TB management, including incorrect use of medications, improper treatment regimens and failure to complete the treatment course (caused by erratic supply, lack of access to treatment or patient non-compliance).

Different levels of resistance are defined by the WHO: MDR TB is caused by Mtb resistant to at least isoniazid and rifampicin, the cornerstone medicines for the treatment of TB; rifampicin-resistant disease requires similar clinical management to MDR TB; pre-XDR TB is MDR TB or rifampicin-resistant TB that is also resistant to any fluoroquinolone; and XDR TB is pre-XDR TB with resistance to at least bedaquiline or linezolid^222,223.

In 2021, the WHO reported nearly half a million cases of rifampicin-resistant TB or MDR TB. Overall, around 3% of new cases and 18% of previously treated cases were MDR TB²⁰⁰.

Molecular mechanisms of drug resistance

The development of molecular resistance to anti-infective agents can be divided into two separate events: the random occurrence of genetic polymorphism within the pathogen population (de novo resistance), which is favoured by high pathogen density in patients; and the subsequent propagation of mutants through drug selection pressure. These genetic alterations can disrupt the drug’s mode of action in several non-mutually exclusive ways (Fig. 1), most notably by altering the drug’s molecular target by increasing its expression or decreasing its binding affinity; by diminishing the concentration of the active compound or its access to the site of action, through altered transport, increased drug degradation, decreased metabolism or enzymatic inactivation; and by boosting the production of antagonizing metabolites.

Fig. 1 |. Comparison of drug resistance mechanisms across pathogens.

Two main steps in the development of widespread resistance: first, the occurrence of de novo resistance resulting from a genetic polymorphism (yellow star); second, the vertical transmission of this resistant state (represented in red) to subsequent microbial generations. The complexity of the resistance mechanism increases from viruses to eukaryotic parasites to bacteria, owing to increasing cellular complexity and/or the use of antimicrobial drugs in non-clinical settings. a, In viruses, an initial genetic event conferring drug resistance alters the molecular target and the resistant pathogen is vertically transmitted. b, In eukaryotic parasites, resistance can emerge from active drug efflux as well as target alteration. c, In bacteria, resistance emerges from active drug efflux, drug degradation or enzymatic inactivation, limited drug uptake or target alteration. In some bacteria, the propagation of resistance can occur via horizontal gene transfer as well as vertical transmission, although this is no longer seen for Mycobacterium tuberculosis (Mtb)¹⁰. CNV, copy number variation; SNP, single nucleotide polymorphism.

The complexity of resistance acquisition varies significantly based on the pathogen (Fig. 1). For instance, viruses, with their limited genomes, produce a relatively restricted number of proteins, offering limited mechanisms of resistance compared with eukaryotic parasites and prokaryotic bacteria. On the other hand, the essentiality of all viral proteins places substantial pressure on virus mutability. Bacteria acquire resistance through the presence of antibiotic resistance genes that naturally exist and can be horizontally exchanged, even among different bacterial genera. This mechanism appears to have played a role in the evolution of an ancestor of the Mtb complex but is not observed among Mtb strains today¹⁰. Besides resistance acquisition, the fitness cost of resistance is critical to the ability of resistant pathogens to spread. It is important to note that external factors, such as the intensive and/or unregulated use of antimicrobials in human communities or agricultural contexts, can significantly influence resistance acquisition and spread.

Malaria

The complexity of the Plasmodium life cycle (Fig. 2) provides scientists with various opportunities to intervene. Consequently, multiple antimalarial drugs have been developed and made available to patients¹¹. The front-line therapies recommended by the World Health Organization (WHO) vary based on the diagnostics and the patient’s situation¹² (Table 1).

Fig. 2 |. Plasmodium life cycle and WHO-recommended antimalarial drugs.

Malaria is transmitted when an infected female Anopheles mosquito injects sporozoites into the human host during a blood meal. Sporozoites are transported by the bloodstream to the liver and invade hepatocytes for a phase of intense asexual replication, producing up to tens of thousands of merozoites¹⁸⁹. In contrast to Plasmodium falciparum, Plasmodium vivax and Plasmodium ovale can form hypnozoites, latent dormant stages that cause disease relapses when reactivated¹⁸⁹. Merozoites are released into the bloodstream and invade erythrocytes, causing the onset of clinical symptoms (high fever, headache, nausea). Three successive development stages follow: rings, in the first hours following invasion; trophozoites, which are highly metabolically active and, consequently, a target of many antimalarial drugs; and schizonts, which are polynucleated and reflect their future differentiation into merozoites that will be released upon rupture of the infected erythrocyte to initiate a new replication cycle. A fraction of the merozoites commit to sexual differentiation and eventually develop into fully mature male and female gametocytes that are taken up by the female Anopheles while feeding. In the mosquito gut, gametocytes differentiate into male microgametes and female macrogametes and fuse to form a diploid zygote, further differentiating into a motile ookinete that migrates across the mosquito gut epithelium to form an oocyst. Within the oocyst, the parasites replicate through sporogony, producing thousands of haploid sporozoites, which, once released following oocyte bursting, migrate to the salivary glands of the mosquito, and perpetuate the parasite’s life cycle upon its next blood meal. WHO-recommended drugs are indicated according to their site of action. AMQ, amodiaquine; ARTM, artemether; ARTS, artesunate; CQ, chloroquine; DHA, dihydroartemisinin; LUM, lumefantrine; MF, mefloquine; PIP, piperaquine; PQ, primaquine; PYM, pyrimethamine; PYN, pyronaridine; QN, quinine; SFD, sulfadoxine; TF, tafenoquine.

Most antimalarial drugs target specific metabolic pathways in Plasmodium, and de novo mutations can easily compromise their efficacy and lead to resistance (Fig. 2). Genetic modifications that alter the binding affinity of the drug to its target can be exemplified by single nucleotide polymorphisms (SNPs): for example, in the pfdhfr gene coding for the target of pyrimethamine^13–16 and cycloguanil^14,17,18, in pfdhps coding for sulfadoxine’s target or in pfcytb coding for the target of atovaquone^19,20. Alternatively, SNPs in pfcrt, the gene encoding the plasmodial chloroquine transporter, lead to resistance to chloroquine^21–23 and piperaquine^24,25 by reducing the drug concentration at the site of action. Copy number variations (CNVs) of the pfmdr1 gene increase protein expression and drug transport, leading to resistance to chloroquine²⁶ and mefloquine^26,27. CNVs of Pfpm2/3 genes, resulting in overexpression of plasmepsin2/3, are associated with resistance to piperaquine²⁸, likely due to compensation for fitness cost rather than a causal decrease of parasite sensitivity to this antimalarial²⁹.

Other mechanisms of resistance can also have a role, such as partial resistance to artemisinin³⁰ caused by several mutations in the pfk13 gene³¹. The WHO published a list of SNPs considered validated markers of this ‘non-classical’ resistance³², which has been a growing concern from its emergence in Southeast Asia in the early 2000s³³ to its recent spreading to Africa³⁴. Because artemisinin combination therapies (ACTs) are the mainstay of treatment, if artemisinin partial resistance became more prevalent and widespread (Box 2) then the efficacy of the partner drugs could be endangered, potentially undermining ACT effectiveness.

HIV infection

Upon HIV infection, a fragile equilibrium between viral replication and the control of the infected individual’s immune system can persist for several years, although the immune system is gradually weakened as CD4+ T lymphocytes become invaded and destroyed (Fig. 3). HIV-1 is the most prevalent and pathogenic species responsible for the majority of the global pandemic³⁵.

Fig. 3 |. HIV intracellular replication cycle and most commonly used anti-HIV drugs.

The human immunodeficiency virus (HIV) life cycle in relation to CD4+ T cells follows several stages. In the attachment phase, HIV enters CD4+ T cells via binding of its gp120 envelope glycoprotein to the cell surface receptor CD4 and the co-receptor CCR5¹⁹⁰, triggering the fusion of the viral and T cell membranes. The viral capsid containing the RNA genome and reverse transcriptase (RT) enzymes is then transported to the nucleus via the nucleopore^191,192 where the RT alternates between the two single-stranded RNA (ssRNA) genome templates to synthesize a single-stranded DNA genome. This genome is then copied to generate double-stranded DNA (dsDNA), which is subsequently integrated into the host cell DNA where it is referred to as a provirus. Although most proviruses are transcribed by the host cell machinery, a few remain latent as a viral reservoir, resulting in the need for lifelong antiretroviral treatment. Once transcribed, the viral mRNA is translated into polyproteins that are cleaved by the HIV-1 protease enzyme³⁵. These new functional viral proteins and copies of the viral genome assemble at the cell membrane and the new virion obtains its lipid envelope by hijacking the host cell plasma membrane while budding out of it. The virion further matures into a new infectious virus, ready to enter and replicate within another host cell¹⁹³. ABC, abacavir; ATV/r, ritonavir-boosted atazanavir; BIC, bictegravir; CAB, cabotegravir; CI, capsid inhibitor; DOR, doravirine; DRV/r, ritonavir-boosted darunavir; DTG, dolutegravir; EFV, efavirenz; EIs, entry inhibitors; INSTIs, integrase strand transfer inhibitors; LEN, lenacapavir; LPV/r, ritonavir-boosted lopinavir; NNRTIs, non-nucleoside reverse transcriptase inhibitors; NRTIs, nucleoside reverse transcriptase inhibitors; PIs, protease inhibitors; RPV, rilpivirine; 3TC, lamivudine; TDF, tenofovir disoproxil fumarate.

The genetic variability of HIV-1, which is due to the high error rate of its reverse transcriptase (RT) coupled with frequent recombination events, is the basis for the development of drug resistance. The error rate of HIV-1 RT is approximately 1 in every 3 × 10⁴ nucleotides^36,37. Given that the HIV-1 genome contains about 104 nucleotides, an error is likely to occur during each round of virus replication^38,39.

As a result, HIV-1 drug resistance is primarily mediated by mutations in the targets of antiretroviral drug therapy. The selection of drug-resistant variants depends on sustained viral replication during therapy, the ease of acquiring specific drug-resistant mutations and the impact of these mutations on drug susceptibility and virus replication⁴⁰. Although mutations arise every day in viruses infecting untreated individuals, these variants rarely rise to detectable levels because in the absence of selective drug pressure they are less fit than drug-susceptible viruses⁴¹. However, these mutations can be selected relatively easily in the presence of antiretroviral monotherapies.

The evolution of these resistance mutations to overcome drug activity is referred to as the genetic barrier to resistance. HIV drug resistance can occur in patients with ongoing virus replication in the presence of incompletely suppressive therapy (typically due to incomplete adherence to treatment) or through direct infection by drug-resistant viruses at the point of transmission (Box 2).

Many drug-resistant mutations in HIV-1 are associated with reduced antiretroviral drug susceptibility^42,43, with their contribution varying widely in terms of range and impact. Some mutations directly reduce susceptibility to one or more drugs within a drug class, whereas others compensate for the reduced fitness associated with another drug-resistant mutation. For certain drugs, a single mutation can confer high-level resistance⁴⁴, although multiple mutations are often needed to render a drug ineffective⁴⁵. Essentially, there is no cross-resistance between drug classes⁴⁶.

Drug-resistant mutations are identified by in vitro passage experiments and genetic sequencing of virus isolates obtained from treated patients. The biological and clinical significance of these mutations is further evaluated in vitro by analysing their impact on the virological response to different antiretroviral therapy ART regimens.

The most frequently utilized antiretroviral drugs fall into four classes: nucleoside reverse transcriptase inhibitors (NRTIs), non-nucleoside reverse transcriptase inhibitors (NNRTIs), protease inhibitors and integrase strand transfer inhibitors (INSTIs) (Fig. 3). Additionally, there are also four US Food and Drug Administration (FDA)-approved inhibitors that prevent HIV entry into a cell, each operating via a distinct mechanism⁴⁷, and one capsid inhibitor⁴⁸. Entry inhibitors are less commonly used because they tend to be costlier, less effective and more prone to resistance development, and therefore they are not widely available in many countries.

NRTI resistance can be mediated by discriminatory mutations that reduce the affinity of RT for an NRTI and prevent its incorporation into the growing DNA chain⁴⁹, or by mutations that facilitate the phosphorolytic removal of NRTIs incorporated into HIV-1 DNA, allowing for the DNA chain synthesis to proceed⁵⁰.

If viruses carry one or two NNRTI resistance mutations, they might retain susceptibility to other NNRTIs, but viruses with more than two NNRTI resistance mutations are usually resistant to the entire class.

Protease inhibitor resistance arises from mutations that directly interfere with the drug binding to the protease or from compensatory mutations at protease cleavage sites, which are also frequently selected during treatment⁵¹. Some protease inhibitors, particularly those boosted by a cytochrome P450–3A4 inhibitor, have a high genetic barrier to resistance due to the need for multiple mutations.

Second-generation INSTIs, such as dolutegravir and bictegravir, are key components of first-line antiretroviral treatment^52–54. They bind to a well-defined pocket formed by the interface between two integrase enzymes and viral DNA. Although a single mutation can be sufficient to confer high-level resistance to the first-generation INSTIs, it typically takes three mutations to cause high-level resistance to the second-generation INSTIs.

Unlike malaria parasites and TB bacteria, HIV lacks drug transporters and resistance is mainly mediated by mutations in the viral genes encoding a limited number of well-known drug targets (Fig. 1). Host cell drug transporters and efflux pumps can, however, have a significant role in reducing drug concentrations in the cell⁵⁵.

Tuberculosis

Mtb can exist in two main stages: an active replicating stage and a non-proliferating persistent stage (Fig. 4). This duality contributes to its pathogenicity and the challenge of treating TB. Mutations in genes encoding drug targets or drug-activating enzymes are the primary mode of drug resistance, but genetic variants leading to the overexpression of efflux pumps or compensatory mechanisms also occur (Fig. 1). Drug resistance develops through the sequential acquisition of SNPs and insertions–deletions (indels)⁵⁶.

Fig. 4 |. Mycobacterium tuberculosis life cycle and the WHO-recommended anti-tuberculosis drugs.

As an airborne pathogen, Mycobacterium tuberculosis (Mtb) infects primarily the lungs, going down the lower respiratory tract and reaching the alveolar space, where it is internalized by alveolar macrophages and, to a lesser extent, by monocytes and dendritic cells⁷⁶. Mtb subsequently reaches the lung parenchyma, either via the infected macrophages or by directly infecting the alveolar epithelium. In parallel, infected monocytes and dendritic cells migrate to pulmonary lymph nodes for T cell priming, leading to the recruitment of various immune cells, including B cells and T cells, to the lung parenchyma. This accumulation of immune cells at the site of infection forms the granuloma, which serves to confine the infection as well as to present immune cells for further control of infection. Latent Mtb infection (the asymptomatic and non-transmittable stage) leads to two outcomes: the immune system either eliminates the infection, or only restrains it so that Mtb persists in a latent state. A subset of infected individuals develop active tuberculosis (TB), when Mtb continues to replicate until the granulomas cannot contain the infection anymore, and the infection enters the bloodstream and spreads to other organs, including the brain. The host becomes symptomatic (fever, fatigue, lack of appetite and weight loss; and in advanced disease, persistent cough and haemoptysis) and infectious upon reinfection of the respiratory tract. World Health Organization (WHO)-recommended drugs are: pretomanid (PMD), isoniazid (INH), rifampicin (RIF) or rifapentine, pyrazinamide (PZA) and ethambutol (EMB); carbapenems (CBPs), linezolid (LZD), D-cycloserine (DCS), terizidone (TRD), clofazimine (CFZ), fluoroquinolones (FLQs; levofloxacin, moxifloxacin, gatifloxacin), thioamides (THAs; ethionamide, prothionamide), bedaquiline (BDQ), delamanid (DLM) and thiacetazone (THZ).

Among major resistance mechanisms^56,57, those that directly affect drug targets include SNPs in the rpoB and inhA genes or the embCAB operon, which encode the targets of rifampicin⁵⁸, isoniazid⁵⁹ and ethambutol⁶⁰, respectively. Resistance mechanisms that abrogate prodrug activation can occur via mutations in the genes for metabolic enzymes, exemplified by SNPs or indels in katG, pncA and ddn, which are involved in isoniazid⁶¹, pyrazinamide^62,63 and delamanid/pretomanid⁶⁴ activation, respectively. Modulation of the expression of drug targets or metabolizing enzymes is exemplified by mutations in the inhA promoter that lead to upregulation of the gene^65,66. Finally, resistance mechanisms can lead to modified or overexpressed efflux pumps, such as mutations in the mmpR5 gene, which encodes a repressor of the MmpL5/MmpS5 efflux pump, leading to upregulation of the pump and cross-resistance to clofazimine and bedaquiline^67,68. Although direct causality has been confirmed by allelic exchange experiments only for some resistance-conferring mutations⁶⁹, whole-genome sequencing (WGS) and phenotypic susceptibility testing association studies, involving more than 38,000 isolates, have implicated several potentially causal mutations with high confidence⁷⁰. Other mutations are known as compensatory, meaning they do not directly confer resistance but mitigate the decrease in fitness caused by the ‘true’ resistance mutations, thus facilitating the spread of the mutants. This is the case for variants in genes such as rpoA and rpoC, which are found associated with rifampicin-resistant rpoB mutants^71,72. The rpoB gene encodes the direct target of rifampicin, the RNA polymerase β-subunit.

Current strategies to combat resistance

The prevailing paradigm suggests that drug resistance observed during in vitro drug discovery studies is likely to emerge in clinical settings. To combat antimicrobial resistance in the field and clinic, various strategies are employed such as the development of new drug candidates with reduced resistance risks, combining drugs into new treatment options and optimizing the use of existing drugs as first-line therapies. The latter approach is particularly used for the treatment of malaria^73–75 and TB^76,77, and can include treatment partitioning (tailoring treatments for specific populations), geographical rotation (rotating treatment regimens within defined regions), the use of higher drug doses and drug repurposing.

However, combining drugs stands out as the most effective approach to delay the onset of resistance to both new chemical entities and existing drugs. Compounds at a high risk of developing drug resistance should be either refined to reduce this risk or deprioritized, whereas those with well-characterized and acceptable resistance profiles can be considered for treatment, particularly when used in combination.

Malaria

Assessing antimalarial resistance in clinical settings plays a pivotal role in advancing new therapies. It enables drug developers to assess which combinations can effectively curb antimalarial resistance and guides priorities for early-stage research aimed at developing future drug candidates.

In Cambodia and Vietnam, where dihydroartemisinin–piperaquine and artesunate–mefloquine treatments resulted in high clinical failure rates, the decision was made not to introduce artemether–lumefantrine as a first-line treatment due to its efficacy falling below 90%. These findings motivated the initiation of the TACT-CV (Triple Artemisinin Therapy Cambodia–Vietnam) study (Clinical Trial ID: NCT03355664 (ref. 78)). This study involved the administration of artemether–lumefantrine to patients with malaria for 3 days, both alone and in combination with amodiaquine. The inclusion of amodiaquine increased the success rate from 90% to 96%⁷⁹.

Alternative clinical strategies to combat drug resistance involved extending the administration of ACTs from 3 days to 5 days, compensating for the decreased parasite clearance rate⁸⁰. Another approach is the rotation of ACTs to prevent resistance to artemisinin partner drugs from developing. Furthermore, adding single low-dose primaquine to ACTs effectively blocks the transmission and spread of both sensitive and resistant parasites⁸¹.

Recent evidence highlights the significance of combination therapies in containing resistance. Notably, recrudescent parasites were observed in three monotherapy clinical trials with the eEF2 inhibitor M5717⁸², as well as in two phase IIa studies with the dihydroorotate dehydrogenase (DHODH) inhibitor DSM265 and the PfATP4 inhibitor KAE609, respectively^83,84. A comparison of the genotypes and phenotypes of recrudescing parasites versus baseline parasites confirmed that mutations in drug targets were the cause of recrudescence. For instance, the inhibitor DSM265 led to the selection of mutations in the dhodh gene in 2 out of 24 patients infected by P. falciparum⁸³. These mutations resulted in a more than 10-fold decrease in DSM265 potency and informed new criteria for assessing resistance risk in drug discovery. In a volunteer infection study conducted with M5717, a protein translation inhibitor, recrudescent parasites in 8 out of 14 participants carried mutations in the pfeEF2 target gene which were similar to those selected in preclinical studies and led to a 5-fold to 40-fold loss of potency⁸².

Precisely predicting resistance emergence in humans based on preclinical data remains a challenging task. Therefore, in the absence of a clear method for accurately predicting the mutation frequency, geno-type and phenotype of clinical resistance, a standardized approach involving in vitro recrudescence of various parasite inocula (ranging from 10⁵ to 10⁹) under suboptimal drug pressure has been successfully implemented. This approach has been instrumental in identifying new drug resistance mechanisms^85,86 and in evaluating the frequency of resistance^87,88. It has also led to the establishment of a minimal inoculum for resistance (MIR) parameter, used alongside other criteria to assess the potential for resistance against a given drug candidate. Recent proposals suggest establishing a ‘triangle of resistance’ for each new antimalarial compound, which would link clinical efficacy with both in vitro phenotyping and genotyping⁸⁹. This approach now underpins all stages of the drug discovery and development pipeline, with ongoing refinement as new clinical data become available.

A significant paradigm shift in antimalarial research has occurred at three distinct levels. First, there is an early elimination of chemical series associated with unacceptable resistance risks. Second, the criteria for assessing resistance in selected compounds are applied. Last, resistance risks are quantified based on the ‘triangle of resistance’ in volunteer infection studies. With respect to early-stage compounds, a selection pressure of 3 × EC₉₀ (the concentration at which the compound inhibits 90% of parasite growth in vitro) is applied to 10⁶ P. falciparum parasites, typically employing the Dd2 strain known for its susceptibility to mutations. Currently, only compounds with a log₁₀MIR(Dd2) value >7 are pursued, provided their decrease in potency does not exceed 10-fold. Late-stage lead compounds exhibiting cross-resistance to laboratory strains or field-isolated parasites, or those losing more than 5-fold of their potency or showing more than a 2% recrudescence in a ring stage assay (indicating cross-resistance with artemisinin), are discontinued. To substantiate the candidate gene’s role in resistance, the complete genome of resistant clones is sequenced, and techniques such as gene editing with CRISPR–Cas9 or transgene expression are employed⁸⁹.

During the candidate selection stage, additional studies are con-ducted at various drug concentrations using inoculates ranging up to 10⁹ parasites (the practical in vitro culture limit) to obtain a higher resolution in the determination of the frequency of resistance. The drug resistance risks are then verified in vivo using the NOD SCID-γ (NSG) humanized mouse model. This preclinical risk analysis includes assessing pre-existing resistance by testing the compound on field isolates and analysing resistance markers against publicly available genomic data-bases. Once in the clinic, the risk of resistance is analysed in phase Ib volunteer infection studies, where recrudescent parasites are genotyped and phenotyped, followed by a comparison with baseline parasites. Importantly, in contrast to earlier optimization phases, no potential drug is halted at this point. Instead, this stage seeks a more precise evaluation of resistance risks to guide future development and com-bination decisions⁸⁹. These criteria have recently emphasized the need to discontinue the development of inhibitors such as the one targeting DHODH unless strategies are developed to significantly reduce resistance risks⁹⁰. Chemical series and candidate drugs demonstrating a low tendency to provoke resistance (logMIR > 10⁹) are strongly preferred and prioritized, and are often referred to as irresistible drugs⁸⁶.

HIV infection

The primary strategy for addressing HIV drug resistance has involved combining multiple drugs that have non-overlapping cross-resistance profiles. In individuals with drug-susceptible HIV-1 who have not received prior treatment, various drug combinations have proven effective in achieving prolonged virus suppression and, in most cases, immune reconstitution⁹¹. Once complete suppression of HIV-1 is achieved, it typically endures indefinitely if therapy is consistently maintained without interruption. Historically, three or more drugs displaying orthogonal resistance were combined to counteract the facile emergence of resistance⁹². However, more recently, drugs with a higher resistance barrier have been developed and employed in two-drug regimens, reducing drug exposure while maintaining an adequate barrier against resistance^93,94.

With the advances in structure-based drug design, the most recently approved enzyme inhibitors have achieved greater potency and a higher genetic barrier to resistance. This was accomplished, in part, by designing inhibitors that tightly bind to their targets while still fitting within the consensus volume of the substrate in the active site, known as the ‘substrate envelope’ of the target. This approach ensures that arising mutations cannot cause steric clashes with the inhibitor without simultaneously interfering with binding to the enzyme’s natural substrate. Another strategy involves designing inhibitors with conformational flexibility, enabling them to adapt to mutations that lead to minor changes in the binding pocket of the target⁹⁵.

Notable examples of inhibitors with increased potency and a high genetic barrier to resistance include the latest approved protease inhibitor, darunavir, and the second-generation INSTIs dolutegravir and bictegravir. In fact, a two-drug combination of dolutegravir and the NRTI lamivudine has proven as effective as any three-drug combination in achieving and maintaining long-term virological suppression in patients with wild-type viruses⁹⁶. Although darunavir offers a higher genetic barrier than the second-generation INSTIs, it is less well tolerated and less effective in clinical practice so is often reserved for salvage therapy^52–54. NRTIs are frequently used in two-drug combinations, with one common combination consisting of a cytosine analogue, either lamivudine or emtricitabine, paired with a tenofovir prodrug. This combination is particularly effective because the most common drug-resistant mutation obtained through cytosine analogue selection (M184V) increases HIV-1 susceptibility to tenofovir⁹⁷.

A successful strategy in combating HIV drug resistance has been the development of fixed-dose combinations that can be administered once daily. This approach significantly reduces the risk of exposing a patient to suboptimal treatment regimens⁹⁸. It involves formulating each drug to be administered once daily, with a pharmacological duration of coverage similar to the companion drugs. Optimized formulations of antiretroviral drugs can also enable bi-monthly intramuscular injections to improve compliance, as achieved for long-acting cabotegravir and rilpivirine formulations⁹⁹. Studying and modelling temporal and spatial aspects of drug exposure will likely contribute to a better understanding and potential mitigation of drug resistance⁴⁵.

Resistance to multiple classes of antivirals remains a challenge and inhibitors of novel targets will be important to develop in future. The capsid inhibitor lenacapavir was approved in 2022, specifically for individuals with viruses resistant to multiple classes of drugs.

Although the principles of managing HIV drug resistance are simi-lar across all populations, variations in access to diagnostic testing in low and middle-income countries have led to differing approaches in addressing acquired and transmitted HIV drug resistance. In high-income countries, genotypic resistance testing is recommended for all patients experiencing virological failure to assist in selecting a new treatment regimen. In most cases, causes of resistance to anti-HIV drugs are identified through the analysis of patients for whom current treatments have failed. WGS of samples from patients with HIV is used to detect mutations within the few drug target genes that could abrogate or significantly weaken the binding of antiviral drugs to their molecular targets¹⁰⁰.

Conversely, in low and middle-income countries, surveillance programmes are used to determine the frequency and genetic mechanisms of transmitted and acquired HIV drug resistance, refining national HIV treatment guidelines^101,102. These guidelines enable patients to access highly effective first-line, second-line and third-line treatment regimens.

Despite advancements in drug design and public health strategies, HIV drug resistance remains a significant challenge, particularly in patients who have received long-term treatment and in infants infected perinatally. Infants are often infected with viruses containing mutations associated with one or two drug classes, further limiting their treatment options, as several drugs are not recommended for use in infants^103,104.

In HIV drug research, the risk of emerging drug resistance against new compounds is evaluated in vitro, but no stop criteria have been defined^105,106. It is assumed that a new drug will inevitably lead to the emergence of resistance unless it is administered as part of a two-drug or three-drug regimen. Additionally, new drugs within an established class must be active against pre-existing resistance¹⁰⁷. However, in vitro resistance selection is not always straightforward and in vivo resistance outcomes do not always coincide with those obtained in vitro^108,109. The purpose of generating resistant viruses in vitro is to determine whether their mutation patterns resemble the natural polymorphism of the viruses observed in patients. This approach has proven to be a suitable surrogate for later clinical observations^110,111. Its limitations become evident in regions with high naturally occurring genetic polymorphism.

It is essential to demonstrate that new compounds are potent against a broad spectrum of viruses^112–115. Usually, this is achieved by testing against a set of up to 10 viruses with different genetic patterns and a wide range of sensitivities, which are selected in silico based on reported activities of 100 different mutated viruses isolated from patients. For instance, the validation of an allosteric HIV integrase inhibitor required demonstrating its full activity against multiple drug-resistant HIV strains¹¹⁶.

Tuberculosis

Combination therapy has been the cornerstone of TB treatment ever since a 1948 British Medical Research Council (BMRC) trial revealed the inadequacy of streptomycin monotherapy in reliably curing TB due to the development of resistance¹¹⁷. The current standard of care for drug-susceptible TB has evolved into a regimen consisting of the first-line drugs isoniazid, rifampicin, pyrazinamide and ethambutol. This regimen involves 6 months of treatment, with isoniazid and rifampicin administered during the first 2 months. More recently, the WHO has also endorsed a 4-month therapy where rifapentine and moxifloxacin replace rifampicin and ethambutol. This regimen has demonstrated non-inferiority to the standard of care in a substantial clinical trial¹¹⁸. The lengthy history of implementing a standard of care programmatically has contributed to the widely held belief that an effective TB therapy requires a minimum of four drugs, which contrasts with the approaches for malaria and HIV infection, where regimens typically involve two or three drugs.

Previously, treatment for MDR TB relied on quasi-individualized therapies rather than fixed combinations, involving the addition of up to six second-line anti-TB agents, some of which had questionable potency. However, this approach is evolving with the introduction of BPaL combined with moxifloxacin. This fixed regimen offers a simpler, shorter and oral treatment option, which should enhance treatment adherence. Poor adherence to TB medication remains a major contributor to negative treatment outcomes, particularly in resource-constrained settings¹¹⁹.

Although the in vitro frequency of resistance is studied during the research phase, it has not been an absolute criterion for anti-TB drug development. For example, two of the three new anti-TB agents introduced in the past four decades, delamanid and pretomanid, were registered despite their relatively high frequency of resistance¹²⁰. Both delamanid and pretomanid are nitroimidazoles, sharing the same activation pathway¹²¹ and molecular target; however, some Mtb iso-lates that exhibit high-level resistance to pretomanid are susceptible to delamanid, suggesting that delamanid could replace pretomanid if drug resistance emerges¹²².

Similarly to HIV, preclinical in vitro experiments on Mtb have not always predicted the dominant resistance mechanisms subsequently encountered in clinical settings. For example, early laboratory experiments on bedaquiline indicated that variants in the atpE gene, encoding the drug’s target ATP synthase subunit C, led to high-level resistance¹²³. However, data from trials and programmatic implementation revealed that such mutations are rare and the low-level bedaquiline resistance observed in clinical practice is primarily due to the upregulation of MmpL5 and MmpS5^124,125.

This lack of predictability underscores the importance of studying pre-existing and emerging resistance during phase II and phase III clinical trials. Consequently, recent TB trials have incorporated paired WGS analysis of Mtb isolates from treatment failures, along with their corresponding baseline samples. This approach allows emergent resistant variants to be identified, and for relapse (regrowth of the same strain) versus reinfection with a different strain to be differentiated¹²⁶. Additionally, in some trials, systematic WGS combined with phenotypic susceptibility testing has provided valuable information about pre-existing resistance rates for new compounds.

Most significantly, the introduction of new anti-TB drugs should be accompanied by robust susceptibility testing protocols, ideally employing rapid molecular tests, to ensure the effective management of patients under the new therapies and for drug resistance surveillance. Unfortunately, this was not the case for the new drugs bedaquiline, delamanid and pretomanid, nor for the repurposed drug linezolid¹²⁷. The lack of access to drug susceptibility testing remains a major challenge in TB control. Globally, only 70% of patients with bacteriologically confirmed TB undergo testing to assess bacterial sensitivity to rifampicin, despite the availability of highly sensitive and specific molecular tests⁸. The situation is even worse for second-line and new or repurposed drugs, for which phenotypic tests can take a month to complete and require complex centralized laboratory capacity. Genotypic approaches are a solution to this issue. In recent years, large consortia have been established to catalogue resistance-conferring mutations for all WHO-endorsed anti-TB drugs. For example, the Comprehensive Resistance Prediction for Tuberculosis: An International Consortium (CRyPTIC) amassed data from more than 12,200 Mtb clinical isolates, subjecting them to WGS and minimum inhibitory concentration determinations for 13 drugs¹²⁸. Also, the Seq&Treat Consortium and WHO initiatives extended this data set to include more than 38,200 clinical isolates⁷⁰. Smaller studies have utilized new approaches such as artificial intelligence, machine learning and convolution neural networks to manage the vast amount of genotypic and phenotypic data available for Mtb^129–131. Machine learning, in particular, has been employed to develop predictive models for identifying additional resistance, which can help guide antibiotic selection while awaiting susceptibility testing results¹³².

The role of mycobacterial fitness in the spread of drug-resistant Mtb has been extensively studied in recent decades. In vitro, mycobacterial fitness is determined by a complex interplay of factors, including growth impairments due to resistance-conferring mutations, strain genetic background and compensatory evolution. Studies on rifampicin resistance have provided examples of these phenomena. Competitive assays using laboratory-derived resistance mutants demonstrated that all rpoB mutations affected relative fitness, although the magnitude of this effect varied depending on the mutation and the genetic back-ground of the parent strain¹³³. Furthermore, comparisons of paired isolates from patients who developed rifampin resistance during treatment showed that some of the same mutations studied in laboratory strains conferred a growth advantage to clinical strains¹³³. This could be attributed to compensatory mutations that emerged in the clinical strains; namely, variants in rpoA and rpoC encoding other subunits of RNA polymerase enhanced the growth of rpoB mutants in vitro⁷². The extrapolation of these laboratory findings to clinical settings has been supported by modelling and epidemiological approaches. Recent studies suggest that the overall fitness of resistant strains might be comparable with that of sensitive strains, with transmission playing a significant role in the spread of drug resistance in certain settings^134,135.

Future strategies to combat drug resistance

Sharing insights from experiences of managing drug resistance in malaria, HIV infection and TB should lead to innovative strategies to reduce the risk of resistance. In addition, a future approach to drug optimization might involve a genetic adaptation step for existing drugs to engineer modified compounds that inhibit the same drug target. Also, leveraging known drug targets and adopting a structure-based drug design strategy to combat drug resistance has proven effective, as exemplified by the medicinal chemistry that resulted in the antimalarial candidate P218, which targets dihydrofolate reductase (DHFR)¹³⁶, and in several new NNRTIs and protease inhibitors against HIV^137,138.

Correlating clinical, phenotypic and genotypic outcomes: A common reason for the reduced therapeutic efficacy of malaria, HIV and TB drugs is the preselection of resistant pathogens before treatment or the de novo selection of resistant pathogens during treatment. Establishing a correlation between genetic mutations, phenotypic indicators (such as loss of in vitro potency) and clinical outcomes is crucial for validating drug resistance as the cause of reduced therapeutic efficacy.

In the case of TB, a comprehensive catalogue of 17,000 Mtb mutations, their association with phenotypic drug resistance and matched data from more than 38,000 isolates across 40 countries for 13 anti-TB drugs has been compiled⁷⁰. The malaria field has employed a similar approach by correlating mutations with loss of in vitro potency and treatment efficacy within the framework of the ‘triangle of resistance’⁸⁹.

Detecting unusual resistance traits that do not significantly affect in vitro potency but, instead, influence the drug’s speed of action also necessitates an approach guided by clinical outcomes. Designing appropriate in vivo studies can help characterize these non-canonical resistance traits. For example, the humanized mouse model of P. falciparum infection allows for genotyping recrudescent parasites post treatment, which mirrors the detection of drug resistance in human clinical trials. These advances, successfully applied in the malaria field, could now be extended to HIV based on the progress made with this model¹³⁹.

Raising the resistance barrier for new drugs

Raising the resistance barrier is a goal actively pursued in antimalarial⁸⁹ and anti-HIV research, yet it poses a greater challenge for anti-TB drugs given the array of strategies that Mtb has evolved to evade drug action.

In the context of malaria, Plasmodium parasites have developed partial resistance to artemisinin, driven by mutations in the kelch13 propeller gene that lead to less efficient parasite killing during short-pulse exposure in vitro¹⁴⁰ and a reduction in parasite clearance rates in patients³¹. Although the precise function of the Kelch13 protein remains unknown, it shares homology with the human proteins, KLHL12 and KLHL2, involved in ubiquitin-mediated protein degradation, and with human KEAP1, which has roles in adaptation to oxidative stress. The reduced clearance (speed of action in patients) of kelch13-mutated P. falciparum parasites promotes the emergence and transmission of partial resistance. Furthermore, the higher parasite load encountered by the partner drug once artemisinin is eliminated from the body can facilitate the development of resistance, potentially leading to treatment failure in combination therapy.

Given this experience with kelch13 mutant parasites, assessing drug resistance in other pathogens should not solely depend on surveilling phenotypic markers, such as the loss of in vitro potency, but should also consider the pathogen’s capacity to develop alternative strategies for evading drug pressure such as reducing the speed of action of the drug. Conducting in vitro recrudescence studies following different periods of drug incubation is crucial for evaluating the full ‘irresistibility’ profile of a drug candidate. In a broader perspective, the search for pharmacological and/or physicochemical commonalities between the drugs used for malaria, HIV infection and TB could potentially inform on their resistance-proof characteristics to guide the design of future anti-infectious agents. However, to our knowledge, such a study has not yet been conducted.

Furthermore, identifying mutation hot spots in pathogen genomes enhances our understanding of the risk of resistance associated with specific drugs and their mode of action, particularly genes localized in hypermutable genomic regions. In the malaria field, chemogenomic studies that link the specific chemistry of compounds with their mechanisms of action and resistance have led to the identification of genome alterations and a deeper comprehension of the malaria ‘resistome’, which can guide the discovery and development of new antimalarial drug candidates¹⁴¹.

It is important to note that hypermutability within the genome is not always detrimental, as illustrated by kataegis, a mutational process that has shown a favourable prognosis in the evolution of HER2+ tumours in patients with breast cancer¹⁴². Identifying similar ‘compensatory’ mutational events in other diseases could lead to improved clinical outcomes and effective management of drug resistance. In malaria research, naturally occurring genetic polymorphisms can result in mutations that confer a high, and hence unfavourable, fitness cost for drug-resistant parasites, preventing their proliferation and spread¹⁴³.

Polypharmacology is another compelling and robust paradigm for containing resistance by inhibiting different pathways or targets with a single drug, preventing pathogens or cancer cells from developing compensatory mechanisms to evade the drug’s action. In Plasmodium parasites, there are several possibilities for polypharmacology within broad classes of drug targets, as described by Arendse et al.¹⁴⁴. This approach appears more challenging in HIV infection due to the limited pool of sufficiently distinct drug targets. However, targeting common nucleoside binding sites or developing dual-specificity bifunctional compounds might offer a strategy to reduce the risk of rapid drug resistance development. The recent advancements in artificial intelligence and machine learning for drug screening and design could guide the generation of hits with a focus on targeted polypharmacological properties¹⁴⁵.

As discussed earlier, the in vitro frequency of resistance has not consistently been a priority for anti-TB drug development. Pro-drugs, in particular, pose a lower barrier to resistance, as both their activation pathways and their molecular targets are susceptible to mutations. Notable examples include isoniazid, pyrazinamide, delamanid and pretomanid, all of which exhibit relatively high frequency of resistance. In the future, greater attention should be devoted to this disadvantage associated with pro-drugs.

Systematic dissection of mutations in target genes

To gauge the potential for pathogens to mutate and evade drug pres-sure, as well as to detect pre-existing resistance mutations in drug tar-gets before drugs are deployed, the wealth of pathogen whole-genome sequences should be exploited. In malaria research, the Pf3k genome database, housing more than 2,600 whole-genome sequences of field-collected parasites, allows researchers to validate whether a mutation newly identified under in vitro resistance selection conditions already exists with high occurrence in natural settings. If such mutations arise in the absence of drug pressure, they are likely attributed to naturally occurring genetic polymorphisms, with associated fitness costs low enough to sustain them in the parasite population.

Taking a cue from malaria research, preclinical research for HIV infection and TB could consider the approach of inducing in vitro resistance through suboptimal drug pressure, which enhances under-standing of the druggable genome and the correlation with drug resistance risk patterns¹⁴¹. Applying this approach to TB research could improve knowledge of the most promising drug targets, as recently demonstrated by Bosch et al. who revealed drug vulnerability pathways involving tRNA synthetases, replicative DNA polymerases and DNA gyrases using a genome-wide gene regulation approach involving CRISPR interference¹⁴⁶.

Alternatively, valuable insights can be drawn from the HIV field, where compounds are screened specifically for their potential to counter observed mutations. Genomic analysis of malaria parasites revealed a high diversity of SNPs in genes of known drug targets such as aminoacyl-tRNA synthetases, which were validated as markers of resistance. Similarly in TB, 31 potential resistance mutations for isoniazid were recently identified in katG, the target gene of this drug¹⁴⁷. Investigating gene polymorphisms associated with resistance can inform the development of compounds expected to mitigate resistance in patients¹⁴⁸.

Uncovering collateral drug sensitivity

In Mtb, one encouraging aspect of drug resistance is the potential for the decreased sensitivity of the mutated pathogen to a particular drug to increase its sensitivity to an unrelated drug¹⁴⁹. This phenomenon is known as collateral sensitivity and is also observed with certain anti-malarial drug candidates, such as inhibitors of P. falciparum acetyl-CoA synthase 10 (ACS10), to which mutant parasites became resistant while exhibiting higher sensitivity to other inhibitors of this enzyme compared with wild-type parasites¹⁵⁰. In addition, compounds that are selective for chloroquine-resistant parasites have been used to select de novo compensatory mutations in vitro that reverse parasite resistance to chloroquine¹⁵¹. Combining such compounds with older antimalarials prone to resistance could rejuvenate efficacy of the older drugs and reinstate their use in clinical practice.

A range of profiling methods, including drug susceptibility assays, genomic analysis and evolutionary analyses, can be employed to reveal collateral susceptibility and select compounds against drug-resistant strains, limiting the emergence and spread of these strains. The practice of cycling compounds in therapies based on the principle of collateral sensitivity to deter resistance development is a concept that was initially proposed for antibiotics and holds the potential to be applied to various therapeutic areas grappling with drug resistance emergence and spread¹⁵².

Exploring multidrug combinations beyond dual therapies

The emergence of drug resistance and lack of efficacy were the main reasons for the development of multidrug combinations that became the standard of care for HIV and Mtb treatments. This strategy is now replicated by the antimalarial community, who recently opted for augmenting an existing antimalarial double ACT with a third complementary drug to create a triple ACT¹⁵³. This is expected to offer a potential solution to control the spread of the partial artemisinin resistance that emerged two decades ago in Southeast Asia¹⁵⁴, and more recently in Africa³⁴, and to extend the partner drug’s longevity. The Tracking Resistance to Artemisinin Collaboration II (TRAC II) study (Clinical Trial ID: NCT02453308), spanning 18 sites in 8 countries, compared the clinical outcomes of patients treated with artemether–lumefantrine alone and in combination with amodiaquine. Triple ACTs demonstrated good efficacy, safety and tolerability, presenting valuable options in an increasingly artemisinin-resistant environment¹⁵³.

Artemisinin partner drugs were shown to have a role in reducing transmission when gametocyte clearance was found to be faster in patients receiving artemether–lumefantrine compared with dihydroartemisinin–piperaquine¹⁵⁵. Such a reduction of transmission by artemether–lumefantrine is even greater with the addition of low-dose primaquine (0.25 mg/kg), as demonstrated by its negative impact on gametocyte carriage and the consequent prevention of transmission to mosquitoes¹⁵⁶. It is expected that this transmission blocking strategy will drastically diminish the transmission of resistance and that triple non-artemisinin combinations could be deployed in the near future.

In TB, the current drug development pipeline contains at least ten main chemical classes of drug that could be used in dual, triple, quadruple or even quintuple combinations. It is anticipated that the use of a panel of preclinical tools to assess pharmacokinetic and pharmacodynamic parameters associated with adaptive trial design could lead to the selection of a limited yet powerful set of anti-TB drug combination regimens that could improve efficacy and shorten the duration of treatments¹⁵⁷.

Modelling the emergence and spread of resistance

Advances in computational modelling have increased the accuracy of modelling the risk of resistance emergence and evolution, leading to improved predictions of the success of new treatments in infectious diseases. In HIV research, modelling has been utilized to estimate the increased risk of resistance when pre-exposure prophylaxis is administered orally. In malaria, modelling the development and spread of resistance based on the fitness cost of mutant strains emerging under suboptimal drug pressure in vitro is not common. In contrast, modelling has shown that the emergence and spread of artemisinin resistance are facilitated by pre-existing resistance to the partner drug in ACTs¹⁵⁸. A similar modelling approach could assess the risk of resistance development in chemoprevention, another critical component of the malaria eradication agenda. The frequency of resistance of Mtb exposed to anti-TB drugs is now complemented by detailed analysis and modelling of the resistance development rate in different epidemiological settings¹⁵⁹. A notable artificial intelligence-based application has been developed with images of Mtb infections^160,161, offering potential applications in malaria research for counting asexual blood stage parasites in blood smears.

Improving formulations for less frequent drug dosing

In HIV infection, in addition to fixed-dose combinations of potent antiretroviral drugs with high genetic resistance barriers, the development of long-acting drugs that require less frequent administration has been an important trend¹⁶². Remarkable among these are the nano-formulations of cabotegravir and rilpivirine, administered intramuscularly every 2 months¹⁶³, and lenacapavir, administered subcutaneously every 6 months¹⁶⁴. Additionally, the clinical development of long-acting antiretroviral-containing implants for HIV treatment and prevention is in its initial stages¹⁶⁵.

The FDA has sanctioned the injectable nanoformulations of cabotegravir and rilpivirine for patients who are virologically suppressed without a history of treatment failure, although this treatment is not approved for patients with active HIV-1 replication. Bi-monthly cabotegravir has also been authorized for pre-exposure prophylaxis in individuals at high risk of contracting HIV-1^166,167.

Although lenacapavir can be administered orally or subcutaneously, its approval is limited to treating individuals with viruses resistant to multiple drug classes. Consequently, it is currently combined with non-long-acting drugs. However, preliminary data from a pilot clinical trial indicated that a single subcutaneous administration of lenacapavir in conjunction with two non-cross-reactive long-acting broadly neutralizing antibodies effectively maintained virological suppression for 6 months¹⁶⁸. The viruses present in the trial participants were assessed to ensure susceptibility to both neutralizing antibodies and, consequently, clinical efficacy measured as HIV-1 suppression for at least 26 weeks. Phase II/III studies are currently underway to explore the use of lenacapavir for pre-exposure prophylaxis¹⁶⁹.

Long-acting formulations of the anti-TB drug rifabutin have also been developed, which could enhance patient adherence during active disease treatment and simplify prophylaxis for latent infections¹⁷⁰.

Moreover, long-acting injectables are being developed for malaria chemoprevention. The objective is to provide a combination product administered intramuscularly from a single vial, with at least 3 months of coverage above the minimum inhibitory concentration for the asexual blood stage. This sustained pharmacological coverage with non-cross-resistant drugs aims to protect individuals from symptomatic infection while reducing resistance selection and transmission.

Depleting infection reservoirs

A multidisciplinary strategy designed to deplete the latent HIV-1 reservoir in patients through immune activators, neutralizing antibodies and therapeutic vaccines is part of a scientific agenda aimed at achieving a ‘functional cure’ (that is, a long-term control of the virus without treatment) for individuals infected with HIV-1 who attain prolonged virological suppression with standard antiretroviral agents¹⁷¹. Such a strategy might be applicable to malaria and TB and offer an effective synergy between drugs that eliminate pathogens and new therapeutics that boost the immune response against these pathogens.

Although bed nets and antimalarial drugs remain the most efficient preventive tools for curbing malaria infection and reducing the disease burden in endemic regions with high transmission, the recent development of next-generation vaccines such as R21 and monoclonal antibodies such as L9LS¹⁷² offers new possibilities for infection prevention and reservoir reduction. These advances work in tandem with curative treatments, which continue to decrease the number of circulating parasites. Another strategy for depleting infection reservoirs and blocking transmission, both in drug-sensitive and drug-resistant malaria cases, is to utilize drugs that, after a mosquito blood meal, clear vector-stage parasitaemia (sporontocidal drugs) or eliminate the mosquito (endectocidal drugs).

Developing host-targeted therapies

Host-directed medicines can complement pathogen-targeted therapies. Exploring drugs that target the host–pathogen interface is one strategy expected to contribute to managing drug resistance in infectious diseases, provided that the toxicity risks are carefully man-aged. Human receptors, enzymes or metabolites that are essential for pathogen infection, proliferation and spread present potential targets for developing antimicrobial drugs with minimal risk of resistance developing. For example, in HIV infection, the entry of the virus into T cells relies on its interaction with host cellular receptors such as CCR5 and CXCR4 (Fig. 3). CCR5 and CXCR4 antagonists have been developed that prevent the virus from binding to these receptors and inhibit infection^173,174. These agents are being tested in clinical trials (in phases I–III for CCR5 antagonists and in phase IB/IIA for CXCR4 antagonists)^173,174.

In malaria and Leishmania parasites, human protein kinases that have crucial roles in cell signalling and transcriptional regulation are potential targets for innovative drug candidates to thwart infection^175,176. However, selecting a kinase that can be inhibited with a broad enough safety window to minimize impact on the host relative to the pathogen remains a challenge. For TB, preclinical studies have identified promising candidate drugs such as granulocyte–macrophage colony-stimulating factor (GM-CSF)¹⁷⁷, which increases macrophage phagocytosis and survival to enhance mycobacterial killing, and tumour necrosis factor (TNF) inhibitors, which mitigate inflammatory host responses¹⁷⁸.

Recently, a phase II trial demonstrated the benefits of combining the malaria standard of care therapy dihydroartemisinin–piperaquine with the anticancer drug imatinib, which targets multiple host tyrosine kinases. This combination enhanced the speed and efficiency of antimalarial treatment without detectable adverse events¹⁷⁹. Similarly, the JAK1/2 inhibitor ruxolitinib was administered together with artemether–lumefantrine to study its potential in modulating the parasite-induced immune response in patients with malaria¹⁸⁰.

Concluding remarks

Recent strategies for combating drug resistance in malaria have involved prioritizing and selecting compounds that are least susceptible to resistance, coupled with a comprehensive assessment of their resistance risk (Fig. 5). Combination products have then been developed to counteract clinical resistance.

Fig. 5 |. Comparison of resistance management strategies for combating drug resistance.

Resistance management strategies during the discovery and development stages of drugs to treat malaria, human immunodeficiency virus (HIV) infection and tuberculosis (TB). MDR, multidrug-resistant; MIR, minimum inoculum for resistance; WGS, whole-genome sequencing.

In contrast, anti-HIV drug development is hindered by a limited genome size and only a handful of well-defined drug targets. Consequently, resistance management relies on re-engineering drugs to restore their potency against mutated, resistant viruses as well as on the development of drugs with novel mechanisms of action¹⁸¹. To mitigate clinical resistance, two or more compounds with orthogonal modes of action are combined (Fig. 5). This combination approach has been extended to other viruses. For example, in hepatitis C therapy, inhibiting the viral protein NS5A with daclatasvir significantly reduces the viral load, but resistance quickly developed in vitro and in patients. This challenge was addressed by combining daclatasvir with the viral polymerase inhibitor sofosbuvir¹⁸².

In TB drug research, and for antibacterials more generally, the search for new resistance-proof antibiotics has been hampered by several hurdles, including a disengagement of pharmaceutical companies from the field, the absence of ideal preclinical models and, especially, the lengthy development of the disease, which leads to long cycles of hypothesis testing for classical medicinal chemistry strategies. Now, the pipeline of anti-TB drugs includes 17 new compounds in phase I or II trials, given renewed investment by governmental, philanthropic, academic and pharmaceutical organizations, often working together as large consortia¹⁸³. A pragmatic approach to avoid resistance has been to combine up to three or four drugs as new treatment regimens (Fig. 5).

Each infectious disease field stands to gain valuable insights from the others. The antimalarial field, with its historical success rooted in phenotypic screening, is transitioning to a target-based screening paradigm. Here, it can draw from the HIV experience to systematically explore a broad spectrum of potential mutations in the target genes to prioritize targets and optimize new compounds accordingly. In malaria, considering fitness costs earlier in the drug discovery process, as is done in TB research, could provide better anticipation of potential drug resistance in the field. Moreover, the concept of ‘irresistible’ compounds in malaria could be a valuable lesson for TB. Furthermore, advances from cancer research — such as the study of epigenetic events, the development of single-cell genome sequencing, the selection of combined therapies and machine learning-driven pattern recognition to study the impact of drugs on cell morphology — could be applied to support the study of drug resistance development and spread in malaria, HIV infection and TB.

Beyond malaria, HIV infection and TB, the diverse approaches outlined in this Review could inform resistance management strategies for other infectious diseases. Neglected infectious diseases such as human African trypanosomiasis, Chagas disease or leishmaniasis are likely to encounter drug resistance issues, particularly as they approach elimination.

Finally, as underscored by the recent SARS-CoV-2 pandemic during which resistant variants posed a constant threat to the efficacy of interventions, well-defined strategies to mitigate and manage resistance are critical for all infectious disease interventions. This is especially pertinent for malaria, HIV infection and TB. Proactively developing such strategies and transferring lessons across diseases will be pivotal in the ongoing battle against antimicrobial resistance.

Glossary

Fitness cost of resistance

The capacity, or lack thereof, for a mutated drug-resistant pathogen to outcompete susceptible pathogens in the absence of the drug

Frequency of resistance

The frequency at which a detectable mutant cell emerges in a pathogen population in the presence of a drug

Genetic barrier to resistance

The number of mutations required in a molecular target gene to confer a meaningful loss of susceptibility to a drug

Irresistible drugs

In the context of antimalarial drug development, when all attempts to provoke resistance in vitro with a drug fail; that is, drugs with a low frequency of resistance and/or a high genetic barrier to resistance

Partial resistance (Also known as reduced parasite clearance)

In the context of antimalarial drug resistance, particularly to artemisinin-based drugs, a delay in the clearance of malaria parasites due to a decrease of the parasite clearance rate after a properly administered treatment

Phenotypic susceptibility testing association studies

Analyses of common mutations to determine their association with a specific phenotype or clinical outcome