Work

Author response: Diverse evolutionary pathways challenge the use of collateral sensitivity as a strategy to suppress resistance

Year: 2023

Type: peer-review

Abstract: Full text Figures and data Side by side Abstract Editor's evaluation eLife digest Introduction Results Discussion Methods Appendix 1 Data availability References Decision letter Author response Article and author information Metrics Abstract Drug resistance remains a major obstacle to malaria control and eradication efforts, necessitating the development of novel therapeutic strategies to treat this disease. Drug combinations based on collateral sensitivity, wherein resistance to one drug causes increased sensitivity to the partner drug, have been proposed as an evolutionary strategy to suppress the emergence of resistance in pathogen populations. In this study, we explore collateral sensitivity between compounds targeting the Plasmodium dihydroorotate dehydrogenase (DHODH). We profiled the cross-resistance and collateral sensitivity phenotypes of several DHODH mutant lines to a diverse panel of DHODH inhibitors. We focus on one compound, TCMDC-125334, which was active against all mutant lines tested, including the DHODH C276Y line, which arose in selections with the clinical candidate DSM265. In six selections with TCMDC-125334, the most common mechanism of resistance to this compound was copy number variation of the dhodh locus, although we did identify one mutation, DHODH I263S, which conferred resistance to TCMDC-125334 but not DSM265. We found that selection of the DHODH C276Y mutant with TCMDC-125334 yielded additional genetic changes in the dhodh locus. These double mutant parasites exhibited decreased sensitivity to TCMDC-125334 and were highly resistant to DSM265. Finally, we tested whether collateral sensitivity could be exploited to suppress the emergence of resistance in the context of combination treatment by exposing wildtype parasites to both DSM265 and TCMDC-125334 simultaneously. This selected for parasites with a DHODH V532A mutation which were cross-resistant to both compounds and were as fit as the wildtype parent in vitro. The emergence of these cross-resistant, evolutionarily fit parasites highlights the mutational flexibility of the DHODH enzyme. Editor's evaluation This study addresses an important question in the field of antimicrobial chemotherapy: whether combinations of enzyme inhibitors that select for mutations that confer resistance to one inhibitor and at the same time increased sensitization to the other inhibitor, can provide a path towards mitigating resistance risks. The authors investigated one such combination of inhibitors of Plasmodium falciparum DHODH (dihydroorotate dehydrogenase), finding that despite "collateral sensitivity", it was still possible to select parasites with resistance to both inhibitors without any change in parasite fitness. Additional cross-susceptibility and structural modelling strengthen this study, which is performed to a high technical standard and presents a convincing body of data. https://doi.org/10.7554/eLife.85023.sa0 Decision letter Reviews on Sciety eLife's review process eLife digest Malaria affects around 240 million people around the world every year. The microscopic parasite responsible for the disease are carried by certain mosquitoes and gets transmitted to humans through bites. These parasites are increasingly acquiring genetic mutations that make anti-malaria medication less effective, creating an urgent need for alternative treatment approaches. Several new malaria drugs being explored in preclinical research work by binding to an enzyme known as DHODH and preventing it from performing its usual role in the parasite. Previous work found that, in some cases, malaria parasites that evolved resistance to one type of DHODH inhibitor (by acquiring mutations in their DHODH enzyme) then became more vulnerable to another kind. It may be possible to leverage this ‘collateral sensitivity’ by designing treatments which combine two DHODH inhibitors and therefore make it harder for the parasites to evolve resistance. To investigate this possibility, Mandt et al. first tested several DHODH inhibitors to find the one that was most potent against drug-resistant parasites. In subsequent experiments, they combined TCMDC-125334, the best candidate that emerged from these tests, with a DHODH inhibitor that works well against vulnerable parasites. However, the parasites still rapidly evolved resistance. Further work identified a new DHODH mutation that allowed the parasites to evade both drugs simultaneously. Together, these findings suggest that the DHODH enzyme may not be the best target for new malaria drugs because many it can acquire many possible mutations that confer resistance. Such results may inform other studies that aim to harness collateral sensitivity to fight against a range of harmful agents. Introduction Antimicrobial resistance threatens our ability to combat multiple infectious agents, and is widely recognized as one of the greatest public health threats of the 21st century (Jee et al., 2018). Globally, drug resistance is a recurring obstacle in efforts to control many endemic infectious diseases, and malaria is no exception. The emergence of resistance in Plasmodium parasite populations and subsequent treatment failure has been documented for all antimalarial drugs in clinical use, including frontline artemisinin combination therapies (Haldar et al., 2018). Thus, there is an ongoing need to develop a next generation of drugs targeting distinct vulnerabilities of Plasmodium parasites. Additionally, understanding the pathways and susceptibility to drug resistance is important to assess next-generation antimalarials, ideally while they are still early in the development process. Inhibitors targeting the Plasmodium dihydroorotate dehydrogenase (DHODH) offer an illustrative example. DHODH is an electron transport chain protein that catalyzes the oxidation of dihydroorotate in orotic acid, which is the rate-limiting step of pyrimidine biosynthesis. In Plasmodium, this reaction is dependent on flavin mononucleotide and ubiquinone cofactors and is coupled to mitochondrial respiration. Because Plasmodium lacks pyrimidine scavenging pathways, this enzyme is essential for parasite growth and survival (Phillips and Rathod, 2010). Multiple high-throughput screens have identified DHODH as a drug target in Plasmodium falciparum (Baldwin et al., 2005; Pavadai et al., 2016; Patel et al., 2008; Ross et al., 2018; Phillips et al., 2008), and the triazolopyrimidine-based DHODH inhibitor DSM265 was developed as a clinical antimalarial candidate (Phillips et al., 2015). While the clinical DHODH inhibitor candidate DSM265 showed promising activity in pre-clinical and clinical studies (Llanos-Cuentas et al., 2018; McCarthy et al., 2017; Murphy et al., 2018; Phillips et al., 2015; Sulyok et al., 2017), we previously demonstrated that resistance to DSM265 emerges rapidly both in vitro and in a humanized mouse model of P. falciparum infection (Mandt et al., 2019). Resistance was primarily conferred by point mutations in the dhodh locus. In Phase 2 clinical trials with DSM265, 2 out of 24 patients failed treatment due to resistance, with recrudescent parasites harboring mutations in dhodh including DHODH C276Y, C276F, and G181S (Llanos-Cuentas et al., 2018). The C276Y and C276F mutations were also observed in in vitro resistance selections, as was DHODH G181C (Phillips et al., 2015; Mandt et al., 2019; White et al., 2019). The point mutations C276F and G181D were observed in our in vivo selection model (Mandt et al., 2019). A common strategy to counter the evolution of resistance is to utilize a combination of two or more drugs with distinct antimicrobial mechanisms. The basic rationale for this strategy is that even if there is some probability of an organism acquiring resistance to each drug separately, the probability that two resistance mutations would occur simultaneously is much lower. Combination therapy is the standard of treatment for malaria (Hastings, 2011), tuberculosis (Kerantzas and Jacobs, 2017), and HIV/AIDS (Cihlar and Fordyce, 2016), and is increasingly being recommended for the treatment of other bacterial infections (Schmid et al., 2019; Coates et al., 2020; Bodie et al., 2019). However, multidrug resistance has emerged in many major human pathogens, complicating patient treatment and threatening public health efforts to control the spread of disease (Hamilton et al., 2019; Manson et al., 2017; Dheda et al., 2017; Gregson et al., 2017). Large-scale bacterial evolution studies have also shown that certain combinations of inhibitors can actually accelerate the emergence of resistance compared to treatment with a single inhibitor (Hegreness et al., 2008; Dean et al., 2020). Thus, there is a need to re-evaluate the current treatment approach, and to strategically design combination therapies based on their ability to slow or suppress the emergence of resistance. One possible strategy takes advantage of collateral sensitivity, in which resistance to one drug causes increased sensitivity to another (Baym et al., 2016). The phenomenon of collateral sensitivity, or ‘negative cross-resistance’, has been observed across various antibiotic classes in bacteria, as well as anti-fungals, and even cancer therapies (Imamovic and Sommer, 2013; Rosenkilde et al., 2019; Maltas and Wood, 2019; Rank et al., 1975; Leroux et al., 2000; Dhawan et al., 2017; Lorendeau et al., 2017). Laboratory studies evaluating the evolution of resistance to antibiotics have demonstrated that when resistance to one inhibitor leads to collateral sensitivity to another, the two inhibitors can be used in a sequential cycling strategy to maintain sensitive bacterial populations (Imamovic and Sommer, 2013; Kim et al., 2014), or simultaneously in combination to slow or suppress the evolution of resistance (Gonzales et al., 2015; Munck et al., 2014; Rodriguez de Evgrafov et al., 2015). Collateral sensitivity to various classes of inhibitors targeting Plasmodium parasites has also been observed. For example, isoforms of the chloroquine resistance transporter (PfCRT) that transport chloroquine out of the parasite’s digestive vacuole can cause increased sensitivity to a variety of compounds (Lukens et al., 2014; Johnson et al., 2004; Sisowath et al., 2009; Richards et al., 2016; Cooper et al., 2002; Evans and Havlik, 1993; Small-Saunders et al., 2022). Collateral sensitivity has also been observed between compounds targeting different subunits of the Plasmodium proteasome Pf20S (Kirkman et al., 2018; Stokes et al., 2019), between different chemotypes targeting the P-Type Cation-Transporter ATPase 4 (PfATP4) (Flannery et al., 2015), and between compounds targeting the Qo and Qi sites of cytochrome b (Lukens et al., 2015). We have also previously shown that collateral sensitivity occurs during the evolution of resistance to Plasmodium DHODH inhibitors. Mutations in dhodh conferring resistance to one inhibitor can alter the enzyme such that it becomes more sensitive to inhibition by other, structurally distinct compounds (Lukens et al., 2014; Ross et al., 2014). The existence of collateral sensitivity among a wide range of antimalarial inhibitors suggests a promising opportunity to exploit this phenomenon to block the emergence of resistant parasites (Lukens et al., 2014). In an effort to expand this work, we screened antimalarial libraries from GSK, and successfully identified several compounds with increased activities against DHODH mutants relative to wildtype parasites (Ross et al., 2018). In this study, we test the hypothesis that combining two DHODH inhibitors based on their collateral sensitivity profiles would suppress the emergence of resistant parasites. We focused on the compound TCMDC-125334, a DHODH inhibitor that exhibited increased activity against all mutant lines tested, including those that arose during selection with the clinical candidate DSM265. We used in vitro resistance selections to characterize the pathways to resistance for TCMDC-125334 and then tested whether parasites develop resistance in the context of combined treatment with DSM265 and TCMDC-125334. Results TCMDC-125334 demonstrates activity against all DHODH mutant parasite lines tested Previously, we screened GSK chemical libraries from the Tres Cantos Antimalarial Set for activity against either wildtype or DHODH mutant parasites (Supplementary file 1a; Ross et al., 2018). In an extended cross-resistance analysis, we characterized the activity of 17 compounds against three additional DHODH mutant lines, DHODH C276Y, DHODH F227L, and DHODH F227L/L531F (Figure 1A, Figure 1—figure supplement 1, Figure 1—source data 1). We visualized the log-transformed fold change in EC50 of each mutant relative to its parental line on a heatmap (Figure 1B, Figure 1—figure supplement 1, Figure 1—source data 1). Hierarchical clustering of this data reveals broad patterns of cross-resistance and collateral sensitivity. The DHODH C276Y, F227I, F227I/L527I, F227L, and F227L/L31F mutant lines have similar dose-response phenotypes. The DHODH E182D and I263F mutant lines also share similar cross-resistance profiles, as we previously reported (Ross et al., 2018). In contrast, the DHODH L531F mutant line has a unique sensitivity phenotype (Figure 1B). This analysis also suggests some structure-activity relationships. Five related triazolopyrimidine-based compounds (DSM265, DSM267, TCMDC-124402, TCMDC-124417, TCMDC-123826) cluster together, as all mutant lines show cross-resistance against these molecules (Figure 1B, Figure 1—figure supplement 1, Figure 1—source data 1). TCMDC-123823, TCMDC-123545, and TCMDC-123566 also cluster together based on their activity profile, and have a similar chemical structure (Figure 1B, Figure 1—source data 1). We had previously reported that the compound TCMDC-125334 was active against the DHODH E182E, I263F, F227I, and F227I/L527I lines (Ross et al., 2018). In our extended cross-resistance analysis, we showed that TCMDC-125334 is also active against the DHODH C276Y, F227L, and F227L/L531F lines. The eight mutant lines had varying degrees of sensitivity to this compound, from DHODH I263F which is 1.3-fold more sensitive than the wildtype parent, to DHODH C276Y, which is 12-fold sensitive. In contrast, all mutant lines were at least 10-fold resistant to DSM265 (Figure 1C). Figure 1 with 1 supplement see all Download asset Open asset Identifying DHODH inhibitors with broad activity against mutant lines. We tested several DSM265-resistant parasite lines against a set of 17 compounds identified from a previous screen of GSK libraries (Ross et al., 2018). (A) Crystal structure of DHODH bound to DSM265 (PDB ID: 4RX0). Mutations included in our cross-resistance analysis are highlighted in pink. (B) To visualize patterns of cross-resistance and collateral sensitivity, we calculated the fold change of each mutant line over its wildtype parent, and plotted the log10-transformed values in a heatmap. Hierarchical clustering by Euclidean distance based on both parasite line and compound was performed using MultiExperimentViewer v4.9. Shades of red indicate that a parasite line is resistant to the indicated compound while blue indicates that it is sensitive. The DHODH C276Y, F227L, and F227L/L531F are newly reported, while other mutant lines were reported in our previous cross-resistance analysis (Ross et al., 2018). The wildtype parent for the DHODH F227I, L531F, F227I/L527I, and I263F lines is Dd2. The wildtype parent for the DHODH E182D line is 3D7. The wildtype parent for the DHODH C276Y, F227L, and F227L/L531F lines is 3D7 A10. (C) One compound, TCMDC-125334 stands out in this analysis as being active against all mutant lines tested. Shown is a bar graph of the log10 fold change in EC50 relative to wildtype for all eight mutant lines tested against DSM265 and TCMDC-125334. DSM265 data was previously reported (Mandt et al., 2019). Each dose-response assay was performed with triplicate technical replicates, and average EC50’s were obtained from 4 to 5 independent biological replicates. See also Figure 1—source data 1. Figure 1—source data 1 Individual bioreplicates of EC50 values (nM) obtained from dose-response assays. https://cdn.elifesciences.org/articles/85023/elife-85023-fig1-data1-v2.xlsx Download elife-85023-fig1-data1-v2.xlsx While none of the mutations in dhodh conferred resistance to TCMDC-125334, we previously reported that this compound appears to exclusively target the DHODH enzyme. Expression of the Saccharomyces cerevisiae cytosolic type I DHODH (ScDHODH) bypasses the mitochondrial DHODH enzyme, rendering blood-stage parasites insensitive to inhibitors that target electron transport chain proteins DHODH and cytochrome bc1. However, addition of proguanil has a potentiating effect for compounds that target cytochrome bc1, but not DHODH (Painter et al., 2007). We previously confirmed that expression of ScDHODH ablates the activity of TCMDC-125334, and that treating the ScDHODH line with TCMDC-125334 and proguanil in combination does not impact this activity (Ross et al., 2018; Dong et al., 2011). Resistance to TCMDC-125334 can be conferred by copy number variation as well as the novel point mutation DHODH I263S Since TCMDC-125334 showed activity against all DHODH mutant lines tested, we wanted to characterize the evolution of resistance to this compound to determine its independent ability to select resistant mutants. We performed in vitro selections with three independent populations of 108 Pf3D7 A10 parasites (Cowell et al., 2018). Cultures were treated with 830 nM TCMDC-125334 (approximately the EC99) until no living parasites were visible by thin smear microscopy, then allowed to recover in compound-free media. After two rounds of treatment, populations displayed moderate (~2-fold) resistance to TCMDC-125334 (Figure 2A and B). Selected parasites were cloned by limiting dilution. Clonal lines isolated from flask 1 had two- or threefold copy number variation (CNV) encompassing the dhodh locus, which corresponded to a two- or three-fold resistance phenotype to TCMDC-125334, as well as to the structurally distinct DHODH inhibitors DSM265, Genz669178, and IDI6273. Of the clones isolated from flask 2, four did not have a resistance phenotype, and had no genetic changes in the dhodh locus (Figure 2C–F, Table 1, Supplementary file 1b, Figure 2—figure supplement 1, Table 1—source data 1, Table 1—source data 2, Table 1—source data 3). One clone, T-F2-C1, was 3-fold resistant to TCMDC-125334 and Genz669178, and 13-fold resistant to IDI-6273 (Figure 2C–F, Table 1, Table 1—source data 1). Whole-genome sequencing (WGS) revealed that clone T-F2-C1 had a point mutation resulting in a DHODH I263S amino acid change (Table 1—source data 3). Notably, this DHODH I263S parasite line remained sensitive to DSM265 (Figure 2D, Table 1, Table 1—source data 1). (Note that Figure 2 and Table 1 show two representative clones from each flask. See Supplementary file 1b for a list of all isolated parasite clones.) Figure 2 with 3 supplements see all Download asset Open asset In vitro resistance to TCMDC125334 can be mediated by copy number variation as well as the novel point mutation DHODH I263S. (A) Protocol for in vitro selection with TCMDC-125334. Parasite populations in three independent 25 mL culture flasks were exposed to 830 nM TCMDC-125334, then allowed to recover in the absence of compound. Resistant parasites emerged after two rounds of treatment within the indicated timeframe. (B) In vitro dose-response curve of bulk populations recovered after the second pulse of compound treatment. Error bars show standard deviation of three technical replicates. (C–E) Resistant populations were cloned by limiting dilutions. Shown are the average EC50 and standard deviation for three biological replicates of dose-response phenotype of two representative clones from each flask for TCMDC-125334 (C), DSM265 (D), Genz669178 (E), and IDI-6273 (F). Each dose-response assay was performed with triplicate technical replicates, and average EC50’s were obtained from 3 to 4 independent biological replicates. Compound structures for TCMDC-125334, DSM265, Genz669178, and IDI-6273 are displayed below their corresponding graphs. The parasite lines are each designated with a unique identifier; for example, T-F1-C1 was selected with TCDC-125334 (T) came from flask 1 (F1), and is designated as ‘C1’ for ‘Clone 1’. DHODH genotypes for representative clones are as follows: T-F1-C1: 2x dhodh copy number variation (CNV); T-F1-C2: 3x dhodh CNV; T-F2-C1: DHODH I263F; T-F2-C1: WT. Statistical significance was determined by a Kruskal-Wallis test, with post hoc multiple comparisons (Dunn’s) of each clone to 3D7 A10. *p≤0.05; **p≤0.01. See also Table 1, Supplementary file 1b, and Table 1—source data 1. Table 1 DHODH genotype and corresponding dose-response phenotype for representative in vitro TCMDC-125334 selected lines. Clone IDDHODH mutation(s)*DHODH CNVTCMDC-125334 EC50 (nM)DSM265EC50 (nM)IDI-6273 EC50 (nM)Genz669178 EC50 (nM)3D7 A10WT1140±42.64.5±1.6502±2096.9±2.2T-F1-C1WT2280±81.38.2±2.6993±22016.9±6.70T-F1-C2WT3426±146**12.5±2.36*1597±212*27.0±10.0*T-F2-C1I263S1432±130.8**2.5±0.747057±671**26.7±15.5*T-F2-C2WT1137±25.14.3±0.94572±91.57.79±3.42Overall p-value The parasite lines are each designated with a unique identifier; for example, T-F1-C1 was selected with TCDC-125334 (T) came from flask 1 (F1), and is designated as ‘C1’ for ‘Clone 1’. Data is shown as EC50 standard Statistical significance was determined by a Kruskal-Wallis test, with post hoc multiple comparisons (Dunn’s) of each clone to 3D7 A10. *p≤0.05; **p≤0.01. are reported for each Each dose-response assay was performed with triplicate technical replicates, and average EC50’s were obtained from 3 to 4 independent biological replicates. identified by Table 1—source data 1 Individual bioreplicates of EC50 values (nM) obtained from dose-response assays. Download Table 1—source data 2 number variation (CNV) analysis based on copy were calculated across the of the for all were calculated using the Download Table 1—source data 3 identified from in vitro selections by Each a single mutation that is in the clone but not in the parent line arose during the drug selection Download In a second independent selection we treated three additional parasite populations with 1 TCMDC-125334. bulk populations recovered after either one or two rounds of treatment displayed moderate (~2-fold) resistance to multiple DHODH and exhibited at the dhodh locus as by (Figure 2—figure supplement 1, Figure 2—figure supplement of the populations treated with 1 were to a second of treatment at 1 TCMDC-125334 for then for These populations developed a resistance phenotype (Figure 2—figure supplement and showed increased copy number at the dhodh locus (Figure 2—figure supplement while no mutations were in the dhodh locus by (Figure 2—figure supplement data 1). a we performed a set of selections with nM DSM265 were exposed to compound for previously parasites resistant to DSM265 emerged rapidly after one of treatment, after in compound-free (Mandt et al., 2019). This was that the of resistance for DSM265 is et al., and we 108 parasites in our selection of the bulk populations revealed two point mutations in C276Y and (Figure 2—figure supplement Supplementary file Figure 2—figure supplement data 1). treatment of DSM265-resistant parasites with TCMDC125334 for additional mutations in dhodh the of DHODH C276Y parasites to TCMDC-125334, we also wanted to test would if we treated DHODH C276Y parasites with this compound. In our previous work, selections with two distinct DHODH and yielded resistant parasites with a DHODH E182D mutation, which were to the DHODH inhibitor we treated the DHODH E182D parasite line with parasites to the wildtype protein with a different The line had a sensitivity phenotype similar to wildtype parasites (Lukens et al., 2014). We that treating DHODH C276Y parasites with TCMDC-125334 would cause a to and offer a strategy to sensitivity to DSM265 (Figure Figure 3 with 1 supplement see all Download asset Open asset parasites can collateral sensitivity by acquiring additional genetic changes that confer resistance to wildtype inhibitors. (A) for sequential (B) Protocol for in vitro DHODH C276Y mutant parasites in three independent 25 mL culture flasks were exposed to nM TCMDC-125334, then allowed to The time on compound and time to are After two rounds of treatment and resistant parasites are observed in all three (C) In vitro dose-response curve of selected bulk populations. Error bars show standard deviation of technical replicates. Resistant populations were cloned by limiting dilution. Shown is the average EC50 of two representative clones from each flask for TCMDC-125334 (C), and DSM265 (D), with chemical structures average EC50 with bars standard Individual biological replicates are also Each dose-response assay was performed with triplicate technical replicates, and average EC50’s were obtained from 3 to independent biological replicates. The parasite lines are each designated with a unique identifier; for example, was selected first with DSM265 (D), then from TCMDC-125334 came from flask 1 (F1), and is designated as ‘C1’ for ‘Clone 1’. DHODH genotypes for representative clones are as follows: DHODH dhodh copy number variation (CNV); DHODH dhodh CNV; DHODH dhodh CNV; DHODH dhodh CNV; DHODH DHODH Statistical significance was determined by a Kruskal-Wallis test, with post hoc multiple comparisons (Dunn’s) of each clone to 3D7 A10. In the EC50 could not be determined with range of tested, the that growth was used as a *p≤0.05; See also Table 2, Supplementary file and Table data 1. The DHODH double mutant a resistant phenotype to single The dose-response phenotypes of the DHODH line selected in this study C276Y the DHODH C276Y line previously selected with DSM265, and the DHODH line previously selected with (Mandt et al., were Shown is a representative dose-response curve for DSM265 and TCMDC-125334 for Error bars show standard deviation of technical replicates within a single For these experiments, we used the DHODH C276Y clone that we had previously isolated from in vitro selections with DSM265 (Mandt et al., 2019). We treated three 25 mL flasks of 108 DHODH C276Y parasites with nM TCMDC-125334. After parasites populations were treated with a second of nM TCMDC-125334. Resistance was observed in all three bulk parasite populations recovered from this second pulse (Figure and While these parasites were less sensitive to TCMDC-125334 relative to the DHODH C276Y parental line, they were still more sensitive than Pf3D7 A10 parasites with a wildtype dhodh (Figure isolated from these selected populations had a similar dose-response phenotype for TCMDC-125334 (Figure and Table 2, Supplementary file Table data 1). However, compared to the DHODH C276Y parent, they had increased resistance to DSM265, with nM (Figure Table 2, Supplementary file Table data 1). revealed that parasites the DHODH C276Y mutation, and additional mutations in than to wildtype (Table 2, Table data 2, Table data 3). isolated from flask 1 and flask 2 had increased dhodh copy in addition to the DHODH C276Y mutation, as confirmed by (Table 2, Figure supplement 1, Table data from flask 3 had an additional mutation (Table 2, Table data 3). (Note that Figure 3 and Table 2 show two representative

Source: eLife

DOI: 10.7554/elife.85023.sa2