Targeting the nucleotide metabolism of Trypanosoma brucei and other trypanosomatids

Abstract African sleeping sickness, Chagas disease, and leishmaniasis are life-threatening diseases that together affect millions of people around the world and are caused by different members of the protozoan family Trypanosomatidae. The most studied member of the family is Trypanosoma brucei, which is spread by tsetse flies and causes African sleeping sickness. Nucleotide metabolism in T. brucei and other trypanosomatids is significantly different from that of mammals and was recognized as a target for chemotherapy already in the 1970–1980s. A more thorough investigation of the nucleotide metabolism in recent years has paved the way for identifying nucleoside analogues that can cure T. brucei brain infections in animal models. Specific features of T. brucei nucleotide metabolism include the lack of de novo purine biosynthesis, the presence of very efficient purine transporters, the lack of salvage pathways for CTP synthesis, unique enzyme localizations, and a recently discovered novel pathway for dTTP synthesis. This review describes the nucleotide metabolism of T. brucei, highlights differences and similarities to other trypanosomatids, and discusses how to exploit the parasite-specific features for drug development.


Introduction
African sleeping sickness, or Human African Trypanosomiasis, is a fatal disease that is spread by tsetse flies in sub-Saharan Africa (Büscher et al. 2017 ). The human disease is caused by two subspecies of the protozoan parasite Trypanosoma brucei. Trypanosoma brucei gambiense causes c hr onic infections pr e v alent in western and central Africa and accounts for 97% of the cases, whereas the less common Trypanosoma brucei rhodesiense is responsible for an acute form of the disease in eastern Africa. When an infected fly bites, it injects parasites into the blood, where they m ultipl y and cause the undulating fe v er that c har acterizes the first stage of the disease . T he symptoms can easily be mistaken for other diseases, and it is often not until the second stage when the parasites infect the central nervous system that a pr oper dia gnosis is made . T his stage is characterized by a wide range of symptoms, including motor system and sensory disturbances, altered diurnal rhythm, personality changes, and in the end a fatal comatose condition. The pr ogr ession of sleeping sic kness is v ery dependent on the subspecies, and the c hr onic form caused by T. b. gambiense takes on av er a ge 3 years to become fatal, whereas T. b. rhodesiense kills the patient within weeks or months. Historically, the prevalence of the disease has been dependent on the political situation in Africa with major increases in times of war and political unrest. The situation was particularly urgent at the turn of the century with 25 000-30 000 r eported yearl y cases, whic h corr esponds to 300 000-500 000 total cases based on estimations of the reporting frequency and the average duration of the disease . T he situation has since then become m uc h better with a > 95% decrease in the yearly number of reported cases (Franco et al. 2020 ). Ho w e v er, 51 million people live in areas where there is a risk of becoming infected (Franco et al. 2020 ), and with a growing number of conflicts there is a risk that the number of cases will increase again. Key to disease control is proper surveillance and good treatment options to avoid vicious cycles where tsetse flies pick up the parasite from infected persons and spread it further.
Trypanosoma brucei belongs to a phylogenetic clade r eferr ed to as salivarian trypanosomes. In addition to the two human pathogenic T. brucei subspecies, this clade also includes T. b. brucei , T. b. evansi , T. vivax, and T. congolense , which cause anemia in a wide range of animals including cattle , horses , and camels (Kasozi et al. 2022 ). Most of the salivarian trypanosomes are spread by tsetse flies, which confines them to Africa. Howe v er, ther e ar e a few exceptions, including T. b. evansi and a South American subpopulation of T. viv ax , whic h ar e spr ead mec hanicall y by other flies without having life cycle stages in the insects. Another example is T. b. equiperdum, which causes a sexually transmitted venereal disease in horses. All salivarian trypanosomes are extracellular, and no major differences in their nucleotide metabolism have been re ported exce pt for some of the properties of their nucleoside/nucleobase transporters (described in the transporter section). Ho w e v er, it should be r emember ed that r esearc h on T. congolense and T. vivax is m uc h mor e limited than on T. brucei and that there could be differences that have not yet been discovered. Although animal trypanosomiasis poses a high economic burden and millions of cattle and other domestic animals die e v ery year because of it, all veterinary drugs in use against the disease were de v eloped mor e than 30 years ago and are limited by side effects and drug resistance (Giordani et al. 2016, Kasozi et al. 2022. This is in stark contrast to the human disease where major impr ov ements have been made in treatments as well as in the number of cases. African sleeping sickness is not the only human disease caused by trypanosomes. In South and Central America, triatomine bugs spr ead T. cruzi, a non-saliv arian trypanosome that causes Chagas disease in humans (Pérez-Molina and Molina 2018 ). The acute stage of the disease is generally mild but can occasionally give sev er e symptoms (1%-5% of the cases), and it often r ea ppears as a c hr onic condition 10-30 years later. Appr oximatel y 6 million people are estimated to be infected with T. cruzi , and 30%-40% of the c hr onic cases suffer fr om cardiac, digestiv e, and/or neur ological alterations that can sometimes be lethal. Acute cases can be successfull y tr eated with benznidazole or nifurtimox, wher eas tr eatment of the c hr onic cases fails in most cases. Another important disease caused by trypanosome-related organisms is leishmaniasis, which is spread by sandflies and is caused by se v er al species of Leishmania (Burza et al. 2018 ). The yearly incidence is 0.7-1 million cases, with 50 000-90 000 being the visceral form of the disease that is nearly always fatal if not treated. In addition, the most common variant, the self-healing cutaneous form, can cause problems with remaining scars after the disease is cured, and the related mucocutaneous form can lead to se v er el y disfiguring facial lesions. Current treatment for leishmaniasis depends on the symptoms as well as which species is causing it. Adverse effects and drug resistance are common problems.
Trypanosoma and Leishmania belong to the Trypanosomatidae famil y of par asites, and because T. brucei is compar ativ el y easy to manipulate genetically, it has served as a model organism for the famil y. This r e vie w will ther efor e focus primaril y on the nucleotide metabolism of T. brucei , but will also discuss the differences and similarities with T. cruzi and Leishmania , which affect more people. A major difference between the three trypanosomatids is that T. brucei is extr acellular, wher eas the main r e plicati ve form of T. cruzi and Leishmania in the mammalian host is the intracellular amastigote form. T. cruzi and Leishmania are taken up by the mammalian host cell into a par asitophor ous v acuole, but T. cruzi needs to escape into the cytosol to start proliferation.
Nucleotide metabolism has made headlines on many occasions in the r esearc h history of T. brucei and other trypanosomatids . Such disco veries include the nucleobase analogue allopurinol, which is still one of the most commonly used drugs against canine and feline leishmaniasis as well as the discovery of that downregulation of the T. brucei P2 adenosine transporter is a common way to ac hie v e r esistance a gainst two of the main classes of drugs used against African trypanosomiasis. In recent years, a m uc h gr eater knowledge about nucleotide metabolism in these parasites has been acquired. In most cells, including our own, nucleotide metabolism is a mainly cytosolic process, whereas studies of the trypanosomatids have revealed an extensive interplay between different organelles in the process, including a ne w mitoc hondrion-dependent pathway for dTTP synthesis not observed in other organisms . T he knowledge about trypanosomatid nucleotide metabolism has also led to advances in drug de v elopment suc h as or all y av ailable adenosine analogues that can cure T. brucei -infected animal models with brain infections. To cure T. brucei central nervous system infections by or all y av ailable drugs has been a holy grail of T. brucei research for decades. Despite major adv ances, the fe w r e vie ws that exist fr om the last fifteen years focus only on specific parts of trypanosomatid nucleotide metabolism. A major incentive behind this proposal is to fill in this ob vious lac k of r e vie ws by giving a broad and comprehensiv e pictur e of nucleotide metabolism in the trypanosomatids, to identify k e y drug targets, and to inspire new research in the area.

Trypanosoma brucei life cycle stages and curr ent tr eatment of African sleeping sickness
Vaccine de v elopment a gainst African sleeping sic kness has been hampered by the ability of T. brucei to switch its coat of variable surface gl ycopr otein, and we ar e completel y dependent on medicines rather than vaccines to combat the disease (Büscher et al. 2017 ). Trypanosoma brucei is an extr acellular par asite with se veral life cycle stages in the mammalian and tsetse fly hosts, and the target of chemotherapy is generally the long slender bloodstream form (BSF), which is the proliferating variant of the parasite in the mammalian blood, lymph, and central nervous system. It is enough to stop the pr olifer ation of the parasites to cure the disease because in the absence of cell division, the variable surface gl ycopr otein switc hing pr ocess cannot occur and the parasites are easily taken care of by the immune defense. A certain fraction of the long slender BSFs develop into nondividing short stumpy BSFs adapted for transmission to tsetse flies. Inside the tsetse fly's midgut, the ingested parasites transform into the procylic life cycle stage that is able to proliferate again. Most of the studies on T. brucei have been performed on the long slender BSFs, whic h ar e most r ele v ant for the disease, and pr ocyclics.
The treatment of African sleeping sic kness v aries depending on both the disease stage and the subspecies causing it (Kasozi et al. 2022 ) as shown in Table 1 . The first stage of the disease can be treated with pentamidine for T. b. gambiense and with suramin for T. b. rhodesiense , wher eas a c hallenge has been that centr al nervous system infections can only be treated with drugs that can efficientl y cr oss the blood-br ain barrier. The tr eatment of secondstage T. b. gambiense infections has gradually improved, with the standard treatment being a nifurtimox-eflornithine combination ther a py (NECT). The eflornithine component, which is also called difluoromethylornithine (DFMO), was initially used as a single treatment. The subsequent inclusion of nifurtimox made it possible to reduce the treatment period significantly, although the treatment is still quite demanding with slow infusions of 400 mg of DFMO e v ery 12 hours for 7 days (the nifurtimox component is giv en or all y). In r ecent years, the or all y av ailable drug fexinidazole has ther efor e become an attr activ e alternativ e a gainst T. b. gambiense infections in remote areas with limited hospital access (Kasozi et al. 2022 ). The medicine can be used against both stages of the disease, but NECT is still recommended for advanced central nervous system infections. Second stage T. b. rhodesiense infections are even more difficult to treat than T. b. gambiense , and the treatment is based on melarsoprol, an arsenical that in 5%-18% of the cases leads to a life-threatening encephalopathy (Büscher et al. 2017 ). A phase II/III clinical trial performed by the Drugs for Neglected Disease initiative (HA T-r -ACC study) is curr entl y e v aluating the effect of fexinidazole on T. b. rhodesiense infections (Lee et al. 2020, Alv ar ez-Rodriguez et al. 2022. Drug resistance has been a major problem for the treatment of both human and animal trypanosomiases, especially resistance against diamidines (e.g. pentamidine) and melaminophenyl arsenicals (e.g. melarsoprol). Interestingly, the first drug resistance gene discov er ed in T. brucei was Tb AT1, whic h encodes the purine transporter P2 and thereby makes a connection to nucleotide metabolism (Carter and Fairlamb 1993, Carter et al. 1995, Mäser et al. 1999 ). The natural substrates of this transporter are adenine Tbg ( + ) + oral * Melarsoprol is no longer a common treatment option against T. b. gambiense . * * Complicated treatment due to large infusion volumes of DFMO. and adenosine, but it was also found to be involved in the uptake of diamidines and melaminophenyl arsenicals . T his transporter is particularly important for the uptake of diminazene aceturate (de Koning et al. 2004 ), a diamidine used as a veterinary drug, whereas resistance to pentamidine and melarsoprol was later found to be primarily mediated by mutations in an aquaporin named AQP2 (Baker et al. 2012, Graf et al. 2013, Munday et al. 2015. Drug resistance against the components in NECT (DFMO + nifurtimox) or against fexinidazole has not yet become a problem, but laboratory experience has shown that selection for resistance to nifurtimox or fexinidazole gives cross-resistance to both of them (Sok olov a et al. 2010 ) and that drug resistance to DFMO can occur by losing the amino acid transporter AAT6 (Vincent et al. 2010 ). It is ther efor e important to de v elop mor e tr eatment alternativ es, and the man y differences in the nucleotide metabolism of T. brucei and other trypanosomatids as compared to humans can be exploited for the de v elopment of drugs with minimal side effects on the host cells.

General nucleotide metabolism
Food oxidation and general energy metabolism are central for the phosphorylation status of cellular nucleotides, with ATP driving the phosphorylation of the other nucleotides. Figure 1 A shows a simplified nucleotide metabolism scheme with deviations in trypanosomatids compared to mammalian cells marked in blue. ATP and other nucleoside triphosphates (NTPs) are dephosphorylated and rephosphorylated in a recycling process where the same nucleotides can be used over and over again. In contrast, the supply of new nucleotides via the de novo and salv a ge synthesis pathways is primarily needed for the expansion of the total nucleotide pool during cell division or to r eplace degr aded nucleotides . T he most striking difference between trypanosomatids and mammalian cells indicated in Fig. 1 A is the lack of de novo purine biosynthesis in the parasites (Ogbunude and Ikediobi 1983 ). This makes them dependent on salv a ging pr eformed purine nucleosides and bases from the host to make new nucleoside monophosphates (NMPs). Nucleosides can either be salv a ged dir ectl y or first cleav ed to liberate nucleobases that are used instead. The cleav a ge can be performed by nucleoside hydr olases, whic h ar e dominant in trypanosomatids but not present in mammalian cells, or by phosphorylases . T he trypanosomatids have a full set of enzymes to interconvert purine NMPs between each other, which allows them to survive with only one purine source. Hypoxanthine is gener all y the purine source in T. brucei growth media.
It is only de novo purine biosynthesis that is lacking in the trypanosomatids, whereas pyrimidines as well as deo xyribon ucleotides (including dTTP) can be produced by de novo biosynthesis . T he lack of de novo purine biosynthesis is very common among pr otozoan par asites and can be exploited by c hemother a py as de-scribed in the following sections. Figure 1 B shows an expanded view of the nucleotide metabolism of T. brucei long slender BSFs. Trypanosoma brucei long slender BSFs and procyclics differ in which transporters they express, but other aspects shown in the figure are similar in the two life cycle stages. Dephosphorylation and phosphorylation reactions of minor importance are not included in this summarizing figure but will still be described in subsequent sections. Most of the reactions in the left part of Fig. 1 B are also conserved in other trypanosomatids except for that the selectivities of the transporters are different and that Leishmania has the ability to deaminate adenine into hypoxanthine and thereby channel the majority of its purine salv a ge via hypoxanthine. Howe v er, it should be r emember ed that the nucleotide metabolism of other trypanosomatids have not been as extensiv el y studied as T. brucei , particularly the deo xyribon ucleotide metabolism shown to the right. Figure 1 B also includes the reactions catalyzed by ITP pyr ophosphatase, whic h is a housek ee ping enzyme that converts ITP (or dITP) and XTP to the corresponding monophospates (Vidal et al. 2022 ). ITP and XTP ar e deamination pr oducts of ATP and GTP, r espectiv el y, but can also be produced by the phosphorylation of the IMP and XMP intermediates in de novo purine biosynthesis . T he purpose of the reaction is to pr e v ent ITP/dITP and XTP from being incor por ated into nucleic acids. Knoc k out studies in T. brucei sho w ed that depleting the enzyme is not lethal by itself but sensitizes the trypanosomes to inhibitors of IMP dehydrogenase, most likely due to the buildup of IMP and consequently ITP (Vidal et al. 2022 ).

Gener a tion of NTPs from NDPs and NMPs
Glycosomes play a central role in ATP generation in T. brucei and other trypanosomatids (Per eir a et al. 2011 ). These peroxisomer elated or ganelles contain the enzymes needed for gl ycol ysis, nucleotide sugar metabolism, the pentose phosphate shunt, and many of the enzymes required for nucleotide metabolism (Guther et al. 2014 ). T he glycosomes ha v e a particularl y centr al r ole in T. brucei BSFs, which primarily live on carbohydrates and repress many of the mitochondrial reactions. In the insect midgut, there are limited amounts of carbohydrates a vailable , and the procyclic form of the parasite expands the size of its single mitochondrion to allow the use of amino acids such as proline and threonine as energy sources (Per eir a et al. 2011 ). In addition, ATP can also be gener ated fr om phosphoar ginine in T. brucei and other trypanosomatids (Per eir a et al. 2011 ) (Fig. 1 A). This r eaction is catal yzed by arginine kinase and has a similar role as creatine kinase in mammalian cells in r a pidl y gener ating A TP when the A TP generated from food oxidation is not sufficient. Although ATP is the main phosphorylation source in the cell, other NTPs participate as en-  Fig. 1 B indicate that all four NDPs can be converted to the corresponding dNDPs, and the P1 transporters are in plural (P1s) because they r epr esent an entire family of similar transporters. Minor uptake and phosphorylation/dephosphorylation activities are excluded, whereas Urd and Cyd are in brackets because their significance is unclear (only minor Urd uptake takes place in BSFs). Abbr e viations: PRTases, phosphoribosyltr ansfer ases; NMP/NDP/NTP, nucleoside mono-, di-, or triphosphate; P-Arg, phosphoarginine; P-creatine, phosphocreatine; X, xanthine; H, hypoxanthine (one letter abbr e viations ar e used for all nucleobases). ergy donors in dedicated reactions, including the usage of CTP in phospholipid metabolism, UTP in carbohydrate metabolism, and GTP in diverse processes such as protein translation and G-protein regulation. When used as energy sources, the NTPs themselves will be converted to NDPs and in a few processes to NMPs (e.g. in RNA turnover and phospholipid metabolism). In addition, NTPs, NDPs, and NMPs can also be dephosphorylated by side reactions of nonspecific phosphatases, but it is is unclear if this makes a significant contribution to the ov er all cellular nucleotide metabolism. The re-phosphorylation of NDPs and NMPs is catalyzed by NDP kinase and se v er al types of NMP kinases, r espectiv el y ( Fig. 1 A).
Energy metabolism is very efficient in most organisms, and a rule of thumb is that the ATP:ADP:AMP ratio is 100:10:1 (Hardie 2018 ). The high cellular activity of NDP kinase and the NMP kinases ensures that the ratio of NTPs over NDPs and NMPs favors the triphosphate form for other nucleotides as well. In T. brucei , nucleotide pool measur ements hav e mainl y focused on NTPs , dNTPs , and ADP. T he ATP pools are generally more than ten times higher than ADP but are very dependent on the supply of nutrients, and the ATP:ADP ratio can decrease within minutes upon growth medium removal in cell washing steps . It is ther efor e important to work fast and av oid w ashes when extracting nucleotides from T. brucei for analysis.
Despite their centr al r ole in nucleotide metabolism, knowledge about the NMP kinases and NDP kinase in trypansomatids is quite limited (Per eir a et al. 2011 ). In mammalian cells, the NMP kinases include adenylate kinases for the phosphorylation of AMP/dAMP, guanylate kinases for the phosphorylation of GMP/dGMP, UMP-CMP kinases for the phosphorylation of UMP and CMP/dCMP and thymidylate kinase for dTMP (Panayiotou et al. 2014 ). Ho w ever, the eukaryotic UMP-CMP kinases are not phylogenetically coherent. The majority of them are cytosolic and belong to a specific clade within the family of adenylate kinases but there is also a mitochondrial isoenzyme that is more related to thymidylate kinases (Fukami-K oba yashi et al. 1996, Xu et al. 2008. Gener all y, NDP kinase and most of the NMP kinases can phosphorylate ribonucleotides as well as deo xyribon ucleotides, and dTMP is the only deo xyribon ucleotide that needs a specific enzyme for its phosphorylation (thymidylate kinase). One factor that complicates the studies of the kinases in T. brucei is the many isoenzymes, including the following: r six classical adenylate kinases (ADK:A-F) r one UMP/CMP kinase with additional AMP specificity (ADKG) r one nuclear adenylate kinase (ADKn) r four NDP kinases (NDPK1-4) r two guanylate kinases r two thymidylate kinases The large number of adenylate kinases in T. brucei and other trypanosomatids is unique among unicellular or ganisms, whic h gener all y is in the range from 1 to 3 (Ginger et al. 2005 ), whereas humans and other m ulticellular or ganisms hav e e v en mor e members than the trypanosomatids (Panayiotou et al. 2014 ). Studies of the T. brucei adenylate kinases show that the isoenzymes differ in localization with one of them in the gl ycosome (ADKD), thr ee of them associated with the flagellum, and the remaining ones of uncertain location but thought to be in the mitchondrion and cytoplasm (Ginger et al. 2005 ). One of the studied enzymes ( T. brucei ADKG) was phylogenetically categorized as a cytosolic CMP-UMP kinase but contains an amino acid substitution that allows it to recognize AMP (Ginger et al. 2005 ). The substrate specificity in descending order is CMP > AMP > UMP (Ginger et al. 2005 ), and the high affinity for CMP ( ∼20 times lo w er K m than for AMP) could be an adaptation to the fact that T. brucei has unusually low cytidine nucleotide pools (Hofer et al. 2001 ). Subsequently, it was found that the trypanosomatids also have an adenylate kinase of the nuclear class (ADKn), and experiments in T. cruzi sho w ed that the enzyme shuttles between the nucleus and the cytoplasm (de los Mila gr os Cámar a et al. 2013 ). Most of the aden ylate kinases ar e conserved in the trypanosomatids, and all T. brucei members except one of the flagellar isoenzymes have homologues in T. cruzi as well (Bouvier et al. 2006 ).
The NDP kinases are also compartmentalized, and studies from T. cruzi show specific isoenzymes for the fla gellum/micr otubules (NDPK2), glycosome (NDPK3), and cytosol/nucleus (NDPK1) (Mir anda et al. 2022 ). Sur prisingl y, NDPK1 can also be excreted and it has been proposed that its extracellular activity might have an imm unor egulatory r ole important for virulence (Miranda et al. 2022 ). From a drug development perspective, it is unclear if it is enough to inhibit specific NDP kinases or NMP kinases in order to kill T. brucei or other trypansomatids. RNAi studies on the adenylate kinases in T. brucei show that reduced expression of some of them can at least reduce cellular growth (Ginger et al. 2005 ). The properties of the NMP kinases and NDP kinases are important to consider for drug de v elopment because they ar e centr al for the activ ation of nucleoside analogues, which in most of the studied cases need to be converted to the triphosphate form in order to have any effect. Ho w e v er, it should be pointed out that many of the recently discov er ed analogues wer e identified by phenotypic scr eening and that mechanism of action studies ar e ther efor e often lac king. For natural nucleosides, it is generally the first phosphorylation step fr om the fr ee nucleoside to the monophospate that is rate-limiting (Arnér and Eriksson 1995 ), but this is not always the case for nucleoside analogues. It would ther efor e be of value to know more about the specificities of the NMP kinases and NDP kinases for future nucleoside analogue development.

Nucleoside and nucleobase transport
It is essential for T. brucei and other trypanosomatids to be able to efficiently take up purines from the environment. Most of the uptake is performed by a family of equilibrati ve n ucleoside transporters (ENTs), which is the dominating nucleoside transporter family also in mammalian cells, but the ones studied in try-panosomatids are not equilbrative and are instead driven by the H + gr adient (Campa gnar o and de K oning 2020 ). T he H + gradient dependency has been verified for the T. brucei P1, H1, and H2 transporters and may be general for the trypanosomatid ENT transporters (de Koning and Jarvis 1997a, de Koning and Jarvis 1997b, de Koning et al. 1998 ). This active transport system can be exploited in drug discovery by using n ucleoside/n ucleobase analogues that are preferentially taken up by the parasites compared to mammalian cells . T he trypanosomatid tr ansporters ar e generally specific for either nucleosides or nucleobases, although there are some exceptions that transport both ( Table 2 ). Most of the transporters listed for Leishmania mexicana have also been verified in Leishmania major and Leishmania donovani , indicating that they seem to be conserved in Leishmania  ). In T. brucei , there is a fairly strict division between purine and pyridimidine specificities, whereas some of the T. cruzi and Leishmania nucleoside transporters have more relaxed specificities (especially Lm UU1). Some other general conclusions are that cytidine transport is limited in all three trypanosomatids shown in Table 2 and that adenosine transport is m uc h mor e efficient in T. brucei and Leishmania than in T. cruzi . In comparison, mammals have broadrange ENTs with only limited ability to discriminate between different nucleosides, and they often recognize some nucleobases as well (Wright and Lee 2021 ). In addition, mammalian cells have a few nucleobase transporters (Inoue 2017 ) and a separate class of sodium-de pendent concentrati ve n ucleoside transporters, but these gener all y hav e a mor e narr ow tissue distribution (Wright and Lee 2021 ).
The two adenosine transporter types in T. brucei , P1 and P2, have r eceiv ed the most attention (Table 2 ). Transporters from the P1 famil y hav e a gener al purine nucleoside specificity, and their main substr ates ar e adenosine , inosine , and guanosine , as well as the corresponding deo xyribon ucleosides (de Koning and Jarvis 1999 ). Ther e ar e se v er al P1 tr ansporter genes ( Tb NT2-7 and 9-10) distributed on different chromosomes, whereas the P2 transporter is encoded by the single Tb AT1 gene (note the different numbering of the protein and gene). Due to the multitude of P1 transporters, whic h ar e fairl y similar to eac h other in their substr ate specificities, it is considered more or less impossible to acquire resistance to drugs b y do wnregulating this transport activity. The opposite is true for the P2 tr ansporter, whic h is easily lost in drug-exposed parasites . T he P2 transporter is only expressed in BSFs and recognizes a pattern of amino groups in the adenine base, and this pattern is lacking in other purines but is shared in diamidines and melaminophenyl arsenicals . T he ribose moiety has only a modest influence on the interaction, and the P2 transporter is active with adenine as well as the corresponding nucleosides adenosine and deoxyadenosine . T he recognition patterns of P1 and P2 will be further described in the drug de v elopment sections. Inter estingly , T . congolense and T . vivax that cause animal trypanosomiasis lack P2 altogether and have only one major P1 transporting activity in each species (Ungogo et al. 2023 ). The single T. congolense P1 transporter is called Tco AT1 or Tco NT10 (not to be confused with Tb AT1 from T. brucei , which is a P2 transporter), and the only T. viv ax P1 tr ansporter that could be confirmed to be active is called Tvx NT3. Transport studies sho w ed that both Tco AT1 and Tvx NT3 are purine-specific and recognize typical P1 substrates (adenosine , inosine , and guanosine), but Tvx NT3 has a broader substrate specificity and can also bind purine nucleobases (not tested as substr ates). Similarl y to the T. brucei P1 transporters (but unlike P2), Tco AT1 and Tvx NT3 does not seem to transport drugs tested from the diamidine and melanophenyl arsenical families (Munday et al. 2013, Ungogo et al. 2023 ). In addition to the purine Table 2. Transporters in Trypanosomatidae. Substrates with affinities 30-300 times lo w er than than of the best substrate are in brackets (double br ac kets indicate e v en lo w er affinity). The table is based on the data fr om a r e vie w on tr ansporters in pathogenic pr otozoans (Campa gnar o and de Koning 2020 ) and includes mainly natural substrates. An exception is that the tubercidin recognition by Tcr NT2 is included in order to highlight that the transporter can bind purines. One-letter abbreviations are used for the nucleobases with H = hypoxanthine and X = xanthine.
nucleoside transporters, all three species have several genes encoding purine nucleobase transporters, although they have only been c har acterized in T. brucei . In T. brucei the expression pattern of H1-H4 varies with the life cycle stage with H2 and H3 being expressed in BSFs ( Fig. 1 B, Table 2 ). H2 is a high-affinity hypoxanthine/guanine transporter (with lo w er affinities for other substr ates), wher eas H3 has a ∼50-fold lo w er affinity than H2 and is likely to be of lo w er importance (de Koning and Jarvis 1997a , Wallace et al. 2002 ). In addition, ther e ar e tr ansporters specific for adenine (ADET1) and hypoxanthine (HXT1) that were identified in BSFs when other transporters were knocked out, but they ar e most likel y of minor importance in wild-type trypanosomes (Campa gnar o et al. 2018 ).
Pyrimidines can be made by de novo synthesis in trypanosomatids, and the tr ansporters, whic h ar e gener all y not of the ENT type, ar e ther efor e of less importance. T. brucei BSFs onl y hav e a single pyrimidine transporter (U3), which is highly specific for uracil (Ali et al. 2013a ). Surprisingly, the transporter does not efficiently use uridine (it has 10 000 times lo w er affinity than uracil), e v en though this is the major p yrimidine sour ce in the blood and cer ebr ospinal fluid (Eells and Spector 1983a ). Thus uridine transport is shown in double br ac kets in Table 2 . BSFs are also capable of low-affinity thymidine transport, but this activity is str ongl y inhibited by adenosine and inosine, which indicates that it is a side activity of the P1 transporter (Ali et al. 2013a ). In procyclics, there is a larger set of pyrimidine transporters with U1 being specific for uracil (and to a lesser extent uridine), U2 being specific for uridine, and C1 being specific for cytosine (Gudin et al. 2006 ). Ho w e v er, the physiological function of the C1 transporter is unclear; its activity is low and the trypanosomes do not have the necessary enzymes to salv a ge cytosine. It can ther efor e be questioned if cytosine is the main substrate or if the transporter has any other unknown function.
In addition to the nucleoside/nucleobase transporters in the plasma membr ane, ther e is also a need to tak e up n ucleotides into differ ent or ganelles. Most or ganelles need the nucleotides as ener gy sources, wher eas gl ycosomes and the mitoc hondrion participate dir ectl y in nucleotide synthesis and ther efor e hav e the need to also take up nucleosides and/or nucleobases (Guther et al. 2014, Moro-Bulnes et al. 2019. Gener all y, the uptake process r equir es tr ansporters, with the exception of the passa ge thr ough the nuclear and outer mitochondrial membranes, which have por es lar ge enough to let nucleotides pass fr eel y. As described below, glycosomal nucleobase transport might also be an exception. Very little is known about organellar transport in the trypanosomatids, but from other eukaryotes it is known that a general differ ence compar ed to the plasma membr ane is that man y of the or ganelles hav e the ability to tr ansport nucleotides, wher eas the plasma membrane only takes up nucleosides and nucleobases. T. brucei has some nucleotide transporter homologues (Colasante et al. 2009 ), but it is only the mitochondrial ADP/ATP carrier ( T. brucei MCP6) that has so far been confirmed to have a transporting activity (Peña-Diaz et al. 2012 ). In T. brucei pr ocyclics, this tr ansporter exchanges the ATP produced in the mitochondrial matrix for ADP needed as substrate in o xidati v e phosphorylation, wher eas the tr ansporter is oper ating in the r e v erse dir ection in BSFs that gen-erate most of their ATP in the glycosomes instead. It is not known how the glycosomes take up nucleotides, but there is evidence that they similarly to peroxisomes contain pores for the exchange of small hydrophilic molecules with the cytosol (Antonenkov andHiltunen 2012 , Gualdron-López et al. 2012 ). These pores are likely to be responsible for the uptake of nucleobases into the glycosomes but not bulkier metabolites such as ATP. For comparison, mammalian per oxisomes hav e an ATP-AMP exc hanger as well as transporters for coenzymes and other substances that cannot be taken up through the pores (Antonenkov and Hiltunen 2012 ).
Although extracellular nucleotides are not taken up directly by the plasma membr ane, suc h uptake can still occur indir ectl y by using ectonucleotidases (ecto-NTs), which are present on the surface of both the trypanosomatids (Cosentino-Gomes and Meyer-Fernandes 2011 ) and the host cells (Giuliani et al. 2020 ). The ecto-NTs dephosphorylate the nucleotides into nucleosides that can be taken up. There are several types of ecto-NTs including NTP diphosphohydrolases, 5´-ecto-NTs and 3´-ecto NTs. NTP diphosphohydrolases and 5´-ecto-NTs dephosphorylate NTPs/NDPs and NMPs, r espectiv el y, wher eas 3´-ecto-NTs can use 3´-nucleotides or nucleic acids as substrates. It is unclear how important the par asitic ecto-NTs ar e for n utrient uptak e (Cosentino-Gomes and Me yer-Fernandes 2011 ). Nucleosides/n ucleobases ar e mor e abundant than nucleotides in the extracellular environments in the mammalian host, whereas they have not been measured in the par asitophor ous v acuole wher e Leishmania amastigotes r eside. In contr ast, the cytosol wher e T. cruzi amastigotes pr olifer ate is heavily dominated by nucleotides that ar e pr esent in millimolar concentr ations (Tr aut 1994 ). Ho w e v er, it is also unclear in this case how important the T. cruzi ´s own ecto-NTs are for this process because there are also cytosolic 5´-NTs in the host cells that can dephosphorylate NMPs and thereby provide nucleosides to the intr acellular par asites.
Nucleotides/nucleosides are also important signalling molecules in plasma and other extracellular fluids, and it has been suggested that the main purpose of the 5´-ecto-NTs/NTP diphosphohydrolases of T. brucei and the extracellular life cycle stages of T. cruzi and Leishmania is to disrupt cell signalling in the host immune defense (Cosentino-Gomes and Meyer-F ernandes 2011 ). T he 3´-ecto-NTs are primarily important for Leishmania among the human pathogenic trypanosomatids (Freitas-Mesquita and Meyer-Fernandes 2017 ). When the parasite has been taken up by its insect vector, the 3´-ecto-NTs might possibl y hav e a r ole in pr oviding purines fr om nucleic acids ingested by the sand fly, and in the mammalian host they have been suggested to help the parasite to escape the immune defense by degrading the nucleic acid component of neutrophil extracellular tr a ps (Guimarães-Costa et al. 2014 ). These effects are dependent on the nuclease function of the enzyme, whereas it is unclear whether the 3´-nucleotidase activity serves any purpose.

Purine salvage
Trypanosoma brucei and other trypanosomatid pathogens are able to use all purine nucleosides and bases known to be salv a geable in nature and to interconvert the formed NMPs (Fig. 2 ). Interestingly, the cytosol can salvage all the nucleosides and bases and has a full set of enzymes to interconvert AMP and IMP, whereas some of the enzymes needed for the interconversion of IMP and GMP ar e onl y pr esent in the gl ycosomes in T. brucei (Guther et al. 2014 ) (Fig. 2 ). It is not known if there is any benefit from this division of labor between the two compartments, but a consequence is that a trypanosome that lives on adenosine or hypoxanthine, Dashed lines indicate low enzyme activities. Abbr e viations: IMPDH, IMP dehydrogenase; GMPR, GMP reductase; sAMP, succinyladenylate; AK, adenosine kinase; APRT1, adenine phosphoribosyltr ansfer ase 1 (cytosolic isoform); HGXPRT, hypoxanthine-guanine-xanthine phosphoribosyltr ansfer ase; IAG-NH, inosine-adenosine-guanosine-nucleoside hydrolase; IG-NH, inosine-guanosine-nucleoside hydrolase; ADSL, adenosylsuccinate l yase; ADSS, aden ylosuccinate synthase; GMPS, GMP synthase; AMPDH, AMP dehydrogenase; HGPRT-I, hypoxanthine-guanine phosphoribosyltr ansfer ase 1 (cytosolic isoform); MTAP, methylthioadenosine phosphorylase; NDT, nucleoside deoxyribosyltr ansfer ase. Isoenzymes that are inactive or not expressed in BSFs are not included (glycosomal enzymes APRT2 and HGPRT-II, r espectiv el y).
whic h ar e two major purines in the blood, can make AMP (and IMP) in the cytosol, but in order to make cytosolic GMP it needs to transport IMP to the glycosome , con vert it to XMP, and take it back to the cytosol to form GMP. A corresponding division of labor between the two compartments is also r equir ed to make AMP from GMP, where the first step occurs in the glycosome and the last two steps in the cytosol. The gl ycosomal membr ane ther efor e needs to allow the passage of NMPs, but it is currently not known how it is mediated. Both compartments have their own adenylate kinases, guanylate kinases, and NDP kinases for the further phosphorylation of AMP and GMP to their triphosphate forms. A distinction between the glycosome and the cytosol is that the glycosome only has enzymes to salv a ge nucleobases, wher eas the cytosol can also salv a ge nucleosides. Perha ps the small pores in the glycosomal membr ane allow mor e efficient passa ge of nucleobases than the larger nucleosides.
Nucleosides can be salv a ged in two different wa ys . In T. brucei , most nucleosides are first cleaved and the corresponding nucleobases ar e salv a ged into NMPs (P arkin 1996 ), with the notable exception of adenosine that can also be dir ectl y phosphorylated to form AMP (Lüscher et al. 2007, Vodnala et al. 2008. Adenosine is the purine source that has r eceiv ed the most attention in T. brucei , both because adenosine analogues have shown good promise against the disease and because the P2 adenosine transporter is involv ed in m ultidrug r esistance. Adenosine can be salv a ged in two different wa ys . T he first wa y is a two-step pr ocedur e to cleave the substrate via the enzyme inosine-adenosine-guanosine nucleoside hydrolase (IAG-NH) and then to salvage the formed adenine (Parkin 1996 ). The second way to salv a ge adenosine is to phosphorylate it dir ectl y with adenosine kinase . T his enzyme has a more than 100-fold higher adenosine affinity (and catalytic efficiency) than IAG-NH (Parkin 1996, Vodnala et al. 2008, which makes it likely to be the major salvage route under conditions when the substr ate concentr ation is low. The combination of having highly efficient transport and phosphorylation activities is an adv anta ge for de v eloping adenosine analogues that are specifically salvaged by the parasite. For comparison, mammalian cells also have a high-affinity adenosine kinase (Sahin et al. 2004 ) but less efficient adenosine transport (Campa gnar o and de Koning 2020 , Wright and Lee 2021 ). There are two isoforms of adenosine kinase in the parasite, but they are nearly identical in amino acid sequence and are considered to be functionall y equiv alent (Vodnala et al. 2008 ). T. brucei adenosine kinase can phosphorylate adenosine , deoxyadenosine , and inosine , but the high affinity for adenosine (K m = 0.041 μM) makes it the pr eferr ed substr ate (Vodnala et al. 2008 ). In contrast, the enzyme activity with inosine is m uc h lo w er and is marked in br ac kets in Fig. 2 (see the drug de v elopment section for enzyme kinetic data). Similarly to the mammalian enzyme, the T. brucei adenosine kinase is subject to substrate inhibition by adenosine in the micromolar range (Vodnala et al. 2008 ).
The extreme efficiency of adenosine salv a ge may also explain the symptoms of late-stage sleeping sickness. Experiments in mice with T. brucei brain infections sho w ed a disruption of the host adenosine signalling system known to be important for sleep regulation, and this might explain why the disease leads to dysregulated diurnal rhythm (Rijo-Ferr eir a et al. 2020 ). The adenosine concentration is normally only 0.12 μM in the cerebrospinal fluid of rodents (Eells and Spector 1983b ) and is below the detection limit of 0.1 μM in humans (Eells and Spector 1983a ). With the high affinity of adenosine kinase, it would not be surprising if the parasites can decrease the concentration further and disrupt the associated signalling pathwa ys , although mor e e vidence is r equir ed to support this hypothesis.
A comparison of purine salv a ge between T. brucei and Leishmania shows some differences. For example, L. donovani adenosine kinase is not subject to substrate inhibition, and it has a compar abl y low affinity for adenosine (K m = 33 μM) (Datta et al. 1987 ). The major route of salv a ge seems instead to occur via cleav a ge to adenine, which is deaminated by adenine amidotransferase, a Leishmania -specific enzyme that does not exist in other studied trypanosomatids or in mammalian cells (Boitz and Ullman 2013 ). Although the parasite also has APRT and can use adenine dir ectl y, the primary pathway in Leishmania is to deaminate the substrate and salv a ge the hypoxanthine using XPRT (homologous to T. brucei HGXPRT) or HGPRT (Boitz and Ullman 2013 ). Figure 2 also includes reactions catalyzed by nucleoside deoxyribosyltr ansfer ase (NDT), an enzyme present in most trypanosomatids , which can intercon vert deo xyribon ucleosides by re placing the n ucleobase attached to them with other nucleobases in the surroundings. Studies of T. brucei NDT, which is purine-specific , ha ve shown that although it is primarily used for the interconversion of purine deo xyribon ucleosides, it can also use ribonucleosides but with a 40-fold lo w er activity (Del Arco et al. 2019 ). The reactions are therefore indicated with question marks in the figur e. Perha ps the low activity can be compensated for by the higher concentrations of the ribonucleosides compared to deo xyribon ucleosides. Guanosine and inosine can fr eel y inter convert betw een each other but the higher physiological concentration of hypoxanthine compared to guanine is likely to favor inosine formation. Inosine (or guanosine) will in turn drive adenosine formation. The unidirectional arrow between the inosine/guanosine equilibrium and adenosine indicates that the oxopurines are superior ribosyl donors compared to adenosine (25fold difference).
Many of the enzymes in T. brucei purine salv a ge hav e been knocked out or knocked down, and the results support the expected roles of the different enzymes in purine salvage.
r HGPR T/HGXPR T. Knocking do wn both enzymes is r equir ed to give a strong effect on T. brucei proliferation with hypoxanthine as the purine source, whereas it is sufficient to knock do wn HGXPR T alone when xanthine is used (Doleželová et al. 2018 ).
r APR T. Knocking do wn APR T leads as expected to str ongl y reduced T. brucei proliferation when adenine is used as the purine source (Doleželová et al. 2021 ).
r Adenosine kinase. Knocking down the enzyme has no effect on T. brucei pr olifer ation with hypoxanthine or supr a physiological concentrations of adenosine as the purine source (Lüscher et al. 2007 ). This is in line with the observation that the enzyme is only important for high-affinity adenosine salv a ge (Vodnala et al. 2008 ).
r IA G-NH/IG-NH/MTAP. Knocking down IA G-NH, IG-NH, or MTAP (methylthioadenosine phosphorylase) giv es onl y a slight reduction in T. brucei proliferation using hypoxanthinecontaining medium, but over time an increased number of abnormal cells accumulate when MTAP or IG-NH is knocked down (Berg et al. 2010 ). The role of MTAP in regular nucleotide metabolism is pr obabl y of minor importance, but by being able to cleave methylthioadenosine it has an important role in the methionine cycle and polyamine synthesis, which might explain the accumulation of abnormal cells when the enzyme is knocked down.

Purine sources and inhibitors of purine salvage
Purine salv a ge has been seen as an ob vious tar get for drug discov ery a gainst the trypanosomatids e v er since the lack of de novo purine biosynthesis was discov er ed in the 1970-1980s (Marr et al. 1978, Gutteridge and Gaborak 1979, Fish et al. 1982 ). Subsequently, it has been shown that the plasma le v el of salv a geable purines is a limiting factor for the pr olifer ation of salivarian trypanosomes as shown in T. musculi -infected mice (Albright and Albright 1988 ). Ho w e v er, the major purine sources in human plasma available for salv a ge ar e a matter of contr ov ersy because the anal yses ar e v ery sensitive to how the samples are handled. Purines are continuousl y excr eted by cells and metabolized to uric acid, which is by far the most abundant purine in human plasma but is not possible to salv a ge. In order to quantify the plasma le v els of salv a geable purines, car e m ust be taken to immediatel y inactiv ate metabolizing enzymes and quic kl y r emov e the cells from the blood when collecting the sample to avoid changes in the purine concentrations during handling. Most studies indicate that hypoxanthine and xanthine are the major salv a geable purine sources, but when human blood samples were directly collected into a vial containing inhibitors of adenosine deaminase and xanthine oxidase follo w ed b y r a pid filtr ation to r emov e the cells, adenosine was high-est (2 μM) follo w ed b y hypoxanthine/xanthine (totally 0.7 μM) and inosine (0.2 μM) (Slowiaczek and Tattersall 1982 ). Hypoxanthine and xanthine were inseparable in the analysis and were therefor e giv en a collectiv e v alue. Guanosine and guanine ar e not detectable in most analyses and are probably of less importance. The natural purine concentrations also vary greatly over time depending on a wide range of factors such as physical exercise and diet, and it has been reported that the plasma hypoxanthine concentr ation can incr ease ∼20 fold after exer cise (Kay a et al. 2006 ). The concentrations of salvageable purines have also been measur ed in cer ebr ospinal fluid, and in this case hypoxanthine and xanthine are the dominating purines, whereas adenosine is below the detection limit in humans (Eells and Spector 1983a ). A general conclusion from all of these studies is that depending on the situation it will either be adenosine or hypoxanthine/xanthine that is the dominating purine source for T. brucei . For the intracellular trypanosomatids, the situation is e v en mor e unclear. The host cell cytoplasm is heavily dominated by nucleotides over nucleosides/nucleobases, and it is difficult to measure the latter r eliabl y due to their fast metabolism and sensitivity to disturbances during harvesting. Nothing is known about purine concentrations in the par asitophor ous v acuole wher e Leishmania amastigotes r eside.
From a drug development perspective, there are two main ways to exploit the salv a ge pathways: r Inhibiting (or hyper activ ating) important enzymes for salv a ge and interconversion.
r Using substrate analogues of salv a ge enzymes that convert the analogues into their nucleotide forms and subsequently inhibit downstream processes.
Despite major efforts to inhibit the salv a ge enzymes, onl y limited pr ogr ess has been made, and curr entl y the str ategy of using substrate analogues looks more promising. Generally, experiments on salv a ge enzyme inhibitors ar e not al ways str aightforw ar d to inter pr et. A complication is that the inhibition is very dependent on the purine source used in the growth medium, and it is ther efor e difficult to tr anslate the in vitro inhibition constants to the situation in the mammalian host. Furthermore, the purine concentrations vary in different animal models as well as in different body fluids and tissues. Other aspects to be aware of with cultured trypanosomes are that they need an adaptation period when changing purine source in the culture and that there can be purines in the growth medium coming from the serum component. This multitude of complications can possibly explain why drug de v elopment pr ojects on substr ate analogues cov er ed in the next section have been m uc h mor e successful than inhibiting salv a ge. Ne v ertheless, ther e ar e a fe w inter esting findings. Gener all y, the inhibitors can be divided into those that block early steps in salvage and those that block the interconversion between AMP , IMP , and GMP . Earl y step bloc kers include APRT inhibitors, dual HGPR T-HGXPR T inhibitors (Doleželová et al. 2018 ), and dual inhibitors of IAG-NH and IU-NH (Berg et al. 2010 ). Generall y, the inhibitors conditionall y inhibit trypanosome pr olifer ation in vitro and the parasites can still live on other purine sources. The HGPR T-HGXPR T inhibitors hav e the br oadest effect due to their ability to block the salvage of all oxypurines and thus force the trypanosomes to only live on adenosine/adenine as a purine sour ce. Ho w e v er, in vivo data would be necessary to shed light into whether it is sufficient to block oxopurine salv a ge to cure T. brucei infections or if the natur al le v els of adenosine/adenine are sufficient to circumvent the inhibition. An alternative approach is instead to target the enzymes responsible for interconversion between AMP , IMP , and GMP . Especially GMP synthase might be an attr activ e tar get because its inhibition makes the trypanosomes dependent on the salv a ge of guanine/guanosine, whic h gener all y are undetectable in the blood and other body fluids. Accordingly, GMP synthase knoc k out T. brucei can onl y be cultur ed in media containing guanine and cannot establish infection in mice (Li et al. 2015 ). Another way to block the pathway from IMP to GMP is to inhibit IMP dehydrogenase . Riba virin is a drug that inhibits T. brucei pr olifer ation in vitro by bloc king both IMP dehydr ogenase and the related enzyme GMP reductase (Bessho et al. 2016 ). Ho w ever, there is a possibility to circumvent the inhibition by the salv a ge of xanthine, and GMP synthase is ther efor e likel y to be a better target.
An alternative option is to inhibit the two enzymes needed for AMP synthesis from IMP (ADSL and ADSS). Knoc k out studies in Leishmania sho w ed that this made the parasites dependent on exogenous adenosine or adenine , and experiments in T. brucei sho w ed that reduced expression of either of the two enzymes slo w ed do wn pr olifer ation both in vitro (with hypoxanthine as the purine source) and in T. brucei -infected mice (Mony et al. 2014 ). Inter estingl y, the experiments on Leishmania sho w ed that the in vivo phenotype of the ADSS knoc k out was mild in comparison to the ADSL knoc k out, wher e the entr a pment of the purines as succinyl-AMP was hypothesized to be the reason for the severely compromised ability to infect mice . T he synthesis of succinyl-AMP is gener all y allostericall y inhibited by adenine nucleotides (AMP/ADP/ATP), and if AMP cannot be formed the regulation fails and IMP is used up for succinyl-AMP synthesis . T his leads to a general purine deficiency due to the fact that IMP is a substrate for both AMP and GMP synthesis.
A special twist in drug de v elopment a gainst the salv a ge enzymes was the discovery of a trypanocidal compound acting as a hyper activ ator of adenosine kinase by pr e v enting substr ate inhibition (Kuettel et al. 2009 ). Hyper activ ators ar e commonl y used to activ ate r eceptors (r eferr ed to as a gonists), but this was, according to the authors, the first time e v er that a metabolic enzyme was targeted by a hyperactivator.

Substr a te analogues of purine salvage enzymes
Substrate analogues of the salvage enzymes act as prodrugs that r equir e cellular conversion by kinases or phosphoribosyltransferases (PRTases) into their active nucleotide forms. Most of the r esearc h focus has been on the first step when the monophosphate form is pr oduced, whic h in the case of T. brucei purine synthesis can occur via HGPR T, HGXPR T, APR T, or adenosine kinase.
Pyrazolopyrimidines act as substrate analogues of purine PR Tases (HGPR T-like enzymes and APRT). The pyrazolopyrimidines are, despite their name, purine analogues, and allopurinol is the one that has r eceiv ed the gr eatest attention as a drug against the trypanosomatids . T his drug, whic h is widel y used for the treatment of gout, proved to be effective against leishmaniasis and Chagas disease in humans already in the 1980-1990s and is still in use today, although less fr equentl y for the human trypanosomatid-related diseases (Marsden et al. 1984, Gallerano et al. 1990, Mazzeti et al. 2019, Ma guir e et al. 2021, Madusanka et al. 2022. Instead, it has become a major treatment option against feline and canine leishmaniasis caused by L. infantum (Ribeiro et al. 2018, Garcia-Torres et al. 2022. Allopurinol resembles hypoxanthine, and its ribophosphorylation b y HGPR T-like enzymes leads to the formation of the corresponding IMP analogue, which needs to be aminated and then further phosphorylated to the corresponding ATP analogue to block the RNA synthesis of the parasites. In T. brucei , allopurinol is a good substrate for HGPRT, but the amination of the formed monophosphate is not efficient (Fish et al. 1985 ). A m uc h str onger antitrypanosomal effect can be ac hie v ed by aminopurinol, an aminated form of allopurinol that bypasses the amination step and is instead a substrate analogue of APRT (Natto et al. 2005 ), but the circumvention of the amination step also leads to greater toxicity in mammalian cells. This is because the selectivity of allopurinol is based on the fact that it is both a poor substrate of mammalian HGPRT (Keough et al. 1999 ) and is not aminated by mammalian cells (Spector et al. 1984 ). Aminopurinol is ther efor e not a good treatment option against T. brucei or other trypanosomatids. In contrast, allopurinol has very low toxicity and is ther efor e a m uc h better option against the trypanosomatids that are sensitive to it.
In the case of T. brucei , substrate analogues of adenosine kinase have been much more successful than the nucleobases described abov e and wer e initiall y dominated by inosine analogues, including formycin B as well as 9-deazainosine (9-DINO) that cured T. brucei -infected mice, although with only limited effect against second-stage disease (Bacchi et al. 1987 ). Similarly to allopurinol, these analogues first form the corresponding IMP analogue and subsequently need to be aminated and further phosphorylated to the corresponding ATP analogue in order to hav e an y effect on the trypanosomes. Formycin B monophosphate can also be aminated in mammalian cells and is ther efor e associated with side effects, wher eas 9-DINO is mor e selectiv e. Two adv anta ges with inosine analogues compared to many adenosine analogues are that they do not need to be combined with adenosine deaminase inhibitors to be stable in the blood and that they are primarily taken up by the T. brucei P1 tr ansporter famil y, whic h is not associated with drug resistance. Inosine was initially proposed to be phosphorylated by ATP-independent phosphotr ansfer ases (the pr e vious term for 5´-nucleotidases), but later it was found that it can be phosphorylated by adenosine kinase (Vodnala et al. 2008 ). The k cat with inosine is only ∼25% compared to adenosine, and the affinity is ∼10 000 times lo w er (K m = 570 vs. 0.041 μM). This may look like a negligible activity, but the catalytic efficiency with this substrate is comparable to adenine arabinoside, an adenosine analogue that was pr ov en to cur e T. brucei -infected mice in an adenosine kinasedependent manner, thus indicating that low-affinity substrates can also be used against the trypanosomes (Vodnala et al. 2008 ). It may seem like an enigma how the analogues have a chance to compete with adenosine despite 3-4 orders of magnitude lower affinity, but a possible explanation could be that adenosine entering the trypanosomes is metabolized so r a pidl y that it does not build up sufficient intracellular concentrations to act as a competitor. Although inosine analogues have shown success in acute mouse models, they have not been verified in mice with brain infections, and most of the de v elopment is instead focused on adenosine analogues to ac hie v e a higher affinity for the enzyme.
Ne v ertheless, ther e has been a r ene wed inter est in inosine analogues in recent years due to the fact that they are taken up by the P1 transporter and thus are not associated with drug resistance. In vitro results indicate that some alkylated 7-deazainosine analogues inhibit T. brucei with EC 50 values in the nM range (Hulpia et al. 2020a ) and an e v aluation of animal trypanosomiasis sho w ed that one of the analogues could cure T. congolense -infected mice (Mabille et al. 2022 ).
Adenosine analogues did not initially have any advantages over inosine analogues against T. brucei in animal models due to the fact that they are deaminated by mammalian adenosine deaminase present in the blood and are converted to the corresponding inosine analogues (Fig. 3 ). This situation changed when the analogues were combined with deo xycoform ycin, an adenosine deaminase inhibitor. When this inhibitor was combined with a pr e viousl y known antitrypanosomal adenosine analogue, cordyce pin (3´-deo xyadenosine), it became possible to cure mice with T. brucei brain infections (Rottenberg et al. 2005 ). Although efficient, cord yce pin has the disadv anta ge of pr edominantl y being taken up by the P2 nucleoside tr ansporter, whic h is associated with drug resistance when downregulated or deleted (Matovu et al. 2003, Geiser et al. 2005. The P2 transporter is compar abl y easy to lose because it is encoded by a single gene. For comparison, ther e ar e two adenosine kinase genes and m ultiple copies of the P1 tr ansporter. Accordingl y, the selection for cord yce pin-resistant trypanosomes led to mutations in the Tb AT1 gene (P2 transporter), whereas P1 transporters, adenosine kinase, or the other genes studied were not affected (Vodnala et al. 2009 ). It is ther efor e advisable to de v elop adenosine analogues that are either taken up by P1 or by both transporters but to avoid analogues that are predominantly taken up by P2.
The study of this analogue demonstrated that the selectivity of P1 over P2 cannot be fully based on competition experiments, which is the most common way to test transporter selectivity. Although competing str ongl y with the P2 transporter, it turned out that P1 was the major transporter as verified by knockout cells lacking the P2 transporter, which were as sensitive to the drug as the par ent str ain and knoc k out cells with r eintr oduced transporter. This indicated that the drug was an inhibitor of P2 rather than a competing substrate . T he FANA-A study also shed light onto the other criteria that should be fulfilled for efficient adenosine analogues against T. brucei . The criteria are summarized in Fig. 3 and ar e v alid for analogues active in their nucleotide form, which is the case for most of the studied adenosine analogues. First of all, the analogues need to be stable against both deamination in the blood and to cleav a ge inside the trypanosomes. IAG-NH is gener all y consider ed the main cleav a ge enzyme in T. brucei , but it is rather specific for the natural purine ribonucleosides (Parkin 1996 ). It is therefore more important to be aware of MTAP, which has a relatively broad sugar moiety specificity and cleaves adenosine, deoxyadenosine, and to a lesser extent cord yce pin (Vodnala et al. 2016 ). In the case of FANA-A, the 2´-fluorinated arabinosyl moiety made it resistant to cleav a ge (Ranjbarian et al. 2017 ). In addition to the stability and uptake efficienc y b y the P1 transporter, the analogues should also be efficiently phosphorylated in order to have a final effect on the trypanosomes. In the case of FANA-A, the triphosphate form of the drug is incor por ated into DNA and blocks further replication, whereas ribonucleoside analogues block RNA synthesis and/or interfere with catabolic , anabolic , or regulatory processes by mimicking ADP or ATP. An example of the latter category is tubercidin, which in its triphosphate form inhibits 3-phosphogl ycer ate kinase during gl ycol ysis (Dr e w et al. 2003 ). The selectivity of the analogues is thus based on a combination of the stability, uptake, phosphorylation, and sensitivity of the final target in the trypanosomes compared to mammalian cells.
The establishment of the factors presented in Fig. 3 made further drug de v elopment mor e r ational, especiall y concerning the role of MTAP, which has been neglected in the past. Figure 4 summarizes the influence of different positions of the adenosine molecule for stability against deamination and cleav a ge, P1/P2 transport, and phosphorylation by adenosine kinase. Neither cord yce pin nor FANA-A ar e stable a gainst deamination and  needed to be combined with deo xycoform ycin (pentostatin) to pr e v ent deamination in the blood (Rottenberg et al. 2005, Ranjbarian et al. 2017. Although deo xycoform ycin is a commonly used drug against some leukemias, experiments in rodents sho w ed that it can cause birth defects and it is ther efor e not recommended during pregnancy (Dostal et al. 1991 ). T hus , it is desirable to avoid deo xycoform ycin and instead use the analogues as single agents. One kno wn w ay to pr e v ent deamination is by 2-halogenization of the nucleobase moiety, but 2-fluorination makes the analogues poor er substr ates of T. brucei adenosine kinase (Vodnala et al. 2008, Ranjbarian et al. 2017. Another dr awbac k is that this modification will make 2´-deoxyadenosine analogues such as FANA-A (but not cord yce pin) mor e efficientl y phosphorylated by mammalian deoxycytidine kinase and thereby increase the risk of side effects (Arnér and Eriksson 1995 ). Experiments with 2F-cord yce pin sho w ed that it could cure T. brucei -infected mice as a single agent, although it r equir ed a 15-fold higher dose than cord yce pin itself and has not yet been verified in a second-stage disease model (Vodnala et al. 2013 ).
The studies of cord yce pin and FANA-A paved the way for the de v elopment of cord yce pin-tubercidin hybrid analogues ( i.e. 3´-deoxy-7-deazaadenosine analogues) based on the knowledge fr om pr e vious studies that tubercidin is resistant to T. brucei MTAP cleav a ge (Vodnala et al. 2016 ) and to deamination (Adamson et al. 1977 ). It was found that 3´-deoxy-7-deazaadenosine could cure second-stage disease, and its high stability made it possible to use or all y as a single agent (Hulpia et al. 2019 ). Analogues with varying P1/P2 specificity were identified, but it was only 3´-deoxy-7deazaadenosine that did not show obvious side effects in the mice, and this analogue was mainly taken up by the P2 transporter. The challenge for future drug development is to improve the P1 specificity, and Fig. 4 gives some hints as to what can be done . T he nucleobase recognition patterns of the P2 transporter and adenosine kinase are to a large extent overlapping, whereas the recognition pattern of P1 (as well as the sugar recognition pattern of adenosine kinase) is often altered in the analogues, with cord yce pin lacking one P1-binding site (3´-OH) and 3´-deoxy-7-deazaadenosine lacking two P1-binding sites (3´-OH and N-7). To maintain a reasonable affinity for the P1 transporter, it is therefore important to add compensatory modifications to regain the lost binding strength. In the case of 7-deazaadenosine analogues, this can be done by adding substituents to position 7 of the purine ring, and verification in P2 knoc k out trypanosomes sho w ed comparable drug sensitivities as the parent strain to these analogues, indicating that they are P1 specific (Hulpia et al. 2020b ). Some other variants of 7deazaadenosine analogues have also shown good activity against T. cruzi and Leishmania (Bouton et al. 2021a, Bouton et al. 2021b ). Figure 4 also shows the binding sites for T. brucei adenosine kinase. FAN A-A and cor d yce pin lac k a hydr oxyl gr oup at the 2´or 3ṕ osition, r espectiv el y, r esulting in a 2-3 orders of magnitude lower affinity than adenosine. Considering that the affinity for adenosine is exceptionally high to start with (K m = 41 nM), a decrease by a few orders of magnitude seems therefore to be compensated for by the activ e tr ansporters that can concentrate the drugs in the cell making both drugs active with EC 50 values in the nanomolar r ange a gainst the par asite. A take-home messa ge is that maintaining strong affinity for adenosine kinase is less critical than cleava ge/deamination r esistance and P1 transport efficiency when dev eloping ne w analogues a gainst T. brucei , and losing one inter action point with the substrate is still acceptable. A few analogues de v eloped a gainst T. brucei ar e also adenosine kinase-independent. These include MTAP substrate analogues that are modified at the 5´-position and thus cannot be phosphorylated at this site (Sufrin et al. 2008 ). MTAP has an important r ole in pol yamine metabolism via its cleav a ge of methylthioadenosine, and inhibitors of the enzyme could also cure T. brucei -infected mice, although the therapeutic effect was dependent on the specific strain of the parasite (Sufrin et al. 2008 ).

Biosynthesis of pyrimidine ribonucleotides
Pyrimidine ribonucleotide metabolism in T. brucei is quite the opposite compared to that of purines by being mainly based on de novo synthesis instead of salv a ge. UTP and CTP can both be synthesized de novo , whereas salvage is restricted to uracil and to a lesser extent uridine (Fig. 5 ). A more detailed review of pyrimidine metabolism in Trypanosomatidae was published in 2018 (Tiwari and Dubey 2018 ), and here we focus on the most interesting featur es fr om the vie wpoint of drug de v elopment a gainst T. brucei . The pyrimidine metabolism of trypanosomatids is in many ways differ ent fr om that in mammals: r The first three enzymes are individual in the trypanosomatids (Gao et al. 1999 , Tiwari andDubey 2018 ), whereas they are merged into a multifunctional protein in mammals called CAD from the initials of the constituent enzymes.
r The next enzyme, dihydr oor otate dehydr ogenase, is cytosolic in T. brucei and in other trypanosomatids instead of mitochondrial as in mammals (Hammond andGutteridge 1982 , Tiwari andDubey 2018 ). r Similarly to the mammalian enzyme, the UMP synthase in T. brucei is bifunctional consisting of two enzymatic activities (OMPDC-OPRT in Fig. 5 ), but it is located to the glycosomes instead of the cytosol in the trypanosomatids (Hammond andGutteridge 1982 , Guther et al. 2014 ).
r In T. brucei , CTP can only be made de novo , and the intracellular le v els of CTP ar e extr emel y low in long slender BSFs (Hofer et al. 2001 ). For comparison, mammalian cells have both de novo and salv a ge synthesis of CTP. Muc h less is known about the metabolism of CTP in other trypanosomatids, but cytidine/cytosine transport is limited and they gener all y hav e a CTP synthetase gene indicating that they ar e likel y to mainly depend on de novo CTP synthesis as well.
r A general theme is that uridine needs to be cleaved prior to its salv a ge in trypanosomatids . T his means that the salv a ge of both uracil and uridine is strictly dependent on uracil phosphoribosyltr ansfer ase (Hammond and Gutteridge 1982 ). This enzyme does not exist in mammals, which instead salvage uridine with uridine-cytidine kinase and use cleav a ge solel y as a step in pyrimidine degradation. In T. brucei the cleav a ge is performed by uridine phosphorylase (UP ase), whic h is structur all y differ ent fr om mammalian UP ase (Larson et al. 2010 ).
In addition, ther e ar e man y differ ences in the synthesis of dTTP, but that is cov er ed in the deo xyribon ucleotide metabolism section. The dTTP synthesis is dependent on a mitochondrial cytidine deaminase, which besides cytidine also uses deoxycytidine as substrate. Ho w ever, it is unclear whether cytidine is an important substrate of the enzyme because the par asite lac ks cytidine transporters and it is not known if significant amounts are generated by dephosphorylations of cytidine nucleotides. Cytidine is ther efor e in br ac kets in Fig. 5 , and the more important deoxycytidine substrate is described together with dTTP synthesis.
Most of the listed differences are universal in the trypanosomatids. Ho w e v er, uridine cleav a ge is performed by a broadspecificity nucleoside hydrolase that also recognizes inosine in Leishmania instead of UPase as in T. brucei and T. cruzi (Shi et al. 1999 ). Leishmania species also have a gene encoding a putative uridine kinase, but earlier studies indicate that there is no such activity in cell extracts indicating that the gene is perhaps not active (Hammond and Gutteridge 1982 ). As mentioned in the transporter section, the pyrimidine specificities vary among the trypanosomatids, and T. brucei BSFs use the U3 tr ansporter, whic h str ongl y fa vors uracil o ver uridine (Ali et al. 2013a ). The inefficient uridine transport is indicated with a dashed arrow in Fig. 5 . As indicated in the figure, uridine can also be pr oduced fr om cytidine by a mitochondrial cytidine deaminase . T his enzyme can also use deoxycytidine as substrate as described in the dTTP synthesis section. The reaction with cytidine is most likely of minor importance for the net production of pyrimidine nucleotides due to the lack of significant cytidine transport activities. It is therefore indicated in br ac kets in the figure. Ho w ever, it cannot be excluded that the enzyme has a role in taking care of cytidine produced internally from the dephosphorylation of cytidine-containing nucleotides. The many differences in pyrimidine de novo synthesis between T. brucei and mammalian cells is an adv anta ge for de v eloping specific inhibitors targeting the trypanosome-specific features in de novo UTP synthesis. One of the k e y questions has been whether the trypanosomes can overcome the inhibition of de novo synthesis by ur acil/uridine salv a ge despite the facts that uracil is barely detectable in the blood and that uridine is a poor substrate of the U2 transporter. Two studies of knockout parasites lacking UMP synthase in de novo synthesis sho w ed that the cells gr e w r eadily in vitro as long as there were sufficient levels of pyrimidines in the growth medium, but they had difficulties surviving and proliferating inside the mammalian host (Ali et al. 2013b, Ong et al. 2013. The parasitemia was low in one of the studies and nondetectable in the other. Ho w e v er, the trypanosomes could in the latter case regain virulence if they were accustomed to medium containing traces of uracil/uridine (using non-dialyzed serum in Figure 5. Pyrimidine metabolism in T. brucei with features different from mammalian cells indicated (labeled 1-6). The brackets around mitochondrial cytidine (and uridine formed downstream) indicate that cytidine is not likely to be a significant source of pyrimidines.in the par asite. Abbr e viations: CPS-II, carbamoyl phosphate synthase II; ATCase, aspartate transcarbamoylase (or aspartate carbamoyltr ansfer ase); DHOase, dihydr oor otase; DHOHD, dihydr oor otate dehydr ogenase; OMPDC, OMP decarboxylase; OPRT, orotate phosphoribosyltransferase; CDA, cytidine deaminase; UPase, uridine phosphorylase; UPRT, uracil phosphoribosyltransferase; CTPS, CTP synthetase. the growth medium) over 4 months before infecting the mice (Ong et al. 2013 ). Both studies concluded ther efor e that bloc ka ge of de novo pyrimidine synthesis is possible to circumvent, and it is ther efor e not a good target by itself. An alternative strategy could instead be to combine an inhibitor of de novo synthesis with a UPRT substrate analogue. An example of such a substrate analogue is 5-fluor our acil. Both UMP knoc k out studies sho w ed that the trypanosomes became > 10 times more sensitive to this drug, indicating that they compensated for the bloc ka ge of de novo synthesis with increased salvage (Ali et al. 2013b, Ong et al. 2013 ). The drug combination would thus serve the dual purpose of both blocking de novo synthesis and preventing compensatory upregulation of the salv a ge pathway.
An enigma in T. brucei pyrimidine metabolism is why BSFs have suc h low CTP le v els. Onl y about 0.5% of the total NTP le v el is CTP in T. brucei isolated from mice, whereas the corresponding percentage in cultured procyclics is 4% (Hofer et al. 2001 ) and is 6% in mammalian cells (Traut 1994 ). In contrast to mammalian cells, which can make CTP both by de novo synthesis and by cytidine salv a ge, the trypanosomes ar e not able to salv a ge c ytidine or c ytosine (Hofer et al. 2001 ). They are therefore completely dependent on CTP synthetase, which makes it a good target for chemothera py a gainst the par asite. Trypanosoma brucei lac ks uridine/cytidine kinase and can ther efor e not phosphorylate and thereby activ ate cyclopenten yl cytosine , 3-deazauridine , or other pyrimidine nucleoside analogues targeting CTP synthetase. Drug development has instead focused on glutamine analogues such as 6diazo-5-oxo-L-norleucine (DON), α-amino-3-c hlor o-4,5-dihydr o-5-isoxazoleacetic acid (acivicin), and 3-bromoacivicin (Hofer et al. 2001, Fijolek et al. 2007, Conti et al. 2011. Glutamine analogues hav e the dr awbac ks of causing side effects on other glutaminedependent r eactions suc h a mammalian de novo purine biosynthesis (DON) and GMP synthase (acivicin). De novo purine biosynthesis is the major target of DON in mammalian cells, and they can ther efor e be r escued by suppl ying hypoxanthine in the gr owth medium whereas the trypanosomes cannot circumvent the effect of the drug on CTP synthesis (Hofer et al. 2001 ). 3-Bromoacivicin was de v eloped as an impr ov ed v ersion of acivicin with the chlorine atom replaced by bromine resulting in higher T. brucei selectivity (Conti et al. 2011 ). Ho w e v er, it was not possible to cure T. brucei -infected mice with DON, acivicin, or 3-bromoacivicin (Hofer et al. 2001, Fijolek et al. 2007, Conti et al. 2011. In all three cases, the number of parasites decreased below the detection limit but r ea ppear ed 4-6 days after the tr eatment was terminated. Attempts have also been made with impr ov ed v ersions of acivicin by designing analogues based on the crystal structure of T. brucei CTP synthetase, but 3-bromoacivicin is still the most selective analogue (Tamborini et al. 2012 ). A possible alternative to the glutamine analogues could be to use pyrazolone-based hydr azones, whic h wer e r ecentl y discov er ed to str ongl y r educe CTP pools in the parasite indicating that CTP synthetase is a likely target (Marchetti et al. 2022 ).

De novo synthesis of deo xyribon ucleotides
Trypanosoma brucei has a full set of enzymes to produce dNTPs by de novo synthesis. As shown in Fig. 6 , ribonucleotide reductase (RNR) is a k e y enzyme for the de novo synthesis of all dNTPs. RNR uses NDPs as substrates and produces the corresponding dNDPs, which after a phosphorylation step by NDP kinase become the dNTPs needed for DNA synthesis. Similarly to nearly all RNRs, the T. brucei enzyme is subject to allosteric specificity regulation, where the binding of dA TP/A TP , dTTP , or dGTP determines which of the four substrates to reduce and thereby ensures balanced levels of dNTPs in the cell (Hofer et al. 1997, Hofer et al. 1998 ). The binding of dATP (or ATP) to this site stimulates the reduction of CDP and UDP, with CDP being the pr eferr ed substr ate of the two. This is not unique to T. brucei , and the mammalian RNR also has a pr efer ence for CDP ov er UDP. The main differ ence betw een the tw o RNRs is instead in the second allosteric site that normall y contr ols Figure 6. Synthesis of dNTPs in T. brucei takes place in the cytosol, mitochondrion, and nucleus. Note that dNTPs are needed both for nuclear and mitochondrial DNA synthesis, although only nuclear DNA synthesis is indicated in the figure. Reactions that are not studied are marked with question marks, wher eas r eactions supposedl y of minor importance ar e marked with dashed arr ows and associated enzymes in br ac kets (these ar e excluded from the metabolism summary in Fig. 1 B). Abbreviations: RNR, ribonucleotide reductase; AK, adenosine kinase; MTAP, methylthioadenosine phosphorylase; 5´-NT, 5´-nucleotidase; TK, thymidine kinase; dUTPase, deoxyuridine triphosphate pyrophosphatase; HD52, histidine-aspartate domain-containing protein 52; CDA, cytidine deaminase; DHFR, dihydrofolate reductase; TS, thymidylate synthase; NDT, nucleoside deoxyribosyltr ansfer ase.
the ov er all activity of the enzyme by binding dATP, whic h turns the enzyme off, and ATP that that turns it on. In T. brucei this site can bind the two nucleotides, but dATP is unable to switch the activity off, and it has been proposed that this function has been lost in evolution (Hofer et al. 1998 ). A possible benefit of losing dATP inhibition could be to enhance the synthesis of dCTP by making dATP a pur el y stim ulatory effector for the r eduction of CDP and to lesser extent UDP.
The lack of dATP inhibition has consequences for drug discov ery. Man y deo xyribon ucleoside analogues de v eloped a gainst mammalian cells in cancer ther a py act both by mimicking dATP to inhibit RNR via the ov er all activity site and by blocking DNA synthesis. It is important to be aware that these analogs do not gener all y affect RNR in T. brucei , whic h lac ks this allosteric site, and they often act as pure DNA synthesis inhibitors instead. Some anticancer nucleoside analogues are also substrate analogues of RNR, but in this case it is important to be aware that cytidine/deoxycytidine analogues (e.g. gemcitabine) are not likely to work because T. brucei does not hav e an y c ytidine/deoxyc ytidine kinase that can phosphorylate the analogues into their active form. As a target of cancer therapy, many classes of RNR inhibitors have been developed, including free radical sca vengers , ir on c helators, and sulfhydryl-r eacting a gents. A fe w of them hav e been tested on T. brucei , including hydr oxyur ea (a radical scavenger) and thiosemicarbazones (iron chelators), but these have so far not been specific enough in comparison to mammalian reference cells (Hofer et al. 1998, Ellis et al. 2015. Ne v ertheless, some of the most w ell-kno wn antitrypanosomal compounds act as indirect inhibitors of RNR activity by interfering with the metabolism of trypanothione, which is needed as a reductant of RNR via tryparedo xin (Dorme yer et al. 2001 ). Melarsoprol is an example of a sulfhydryl-r eacting a gent that scav enges trypanothione (and can possibly also react with sulfhydryl groups in RNR itself), whereas DFMO blocks the synthesis of trypanothione by inhibiting ornithine decarboxylase (Fairlamb et al. 1987, Fairlamb et al. 1989, Dormeyer et al. 2001, Larson et al. 2021. The synthesis of dTTP r equir es se v er al additional enzymes compared to the other dNTPs. A k e y enzyme in de novo dTTP synthesis is thymidylate synthase (TS), whic h conv erts dUMP to dTMP. The methyl group needed in the reaction is generally taken fr om methylene-tetr ahydr ofolate. In T. brucei and other trypanosomatids, TS is fused to dihydrofolate reductase (DHFR), which is needed for the r egener ation of tetr ahydr ofolate fr om the dihydr ofolate generated in the TS-catalyzed reaction (Garrett et al. 1984 ). Another specific feature of DHFR-TS is that in contrast to the corresponding mammalian enzymes it is mitochondrial and not cytosolic (Moro-Bulnes et al. 2019 ). Knockout studies in T. brucei sho w ed that DHFR-TS-depleted trypanosomes r equir e thymidine in the growth medium for survival and cannot establish infections in mice, indicating that the low thymidine le v els in the blood are not sufficient for the parasite to live on (Sienkiewicz et al. 2008 ). Knoc k out trypanosomes cultur ed in thymidine-fr ee medium died within 60 hours indicating that it is not onl y par asite pr olifer ation that is affected, but also parasite viability.
TS and DHFR ar e used as tar gets a gainst cancer, psoriasis (by inhibiting immune cell pr olifer ation), and infectious diseases. Studies of the r ecombinantl y expr essed T. brucei DHFR-TS protein sho w ed that typical inhibitors of mammalian DHFR (e .g. methotrexate) and TS (e .g. 5-fluor o-dUMP) wer e not selective enough and that microbial DHFR inhibitors such as pyrimethamine and trimethoprim were not very effective (Gibson et al. 2016 ). Ho w e v er, it is onl y the TS part of the protein that is essential. T. brucei and other trypanosomatids can compensate for inhibition of DHFR by upregulating pteridine reductase 1 (PTR1). The DHFR domain and PTR1 have been extensiv el y studied, whereas the full-length DHFR-TS protein is much less stable and more difficult to work with (Gibson et al. 2016 ). PTR1 is a br oad-substr ate enzyme that can r educe both biopterins and folates and thereby bypass the need for DHFR. Knoc k out studies sho w ed that it is essential in T. brucei b y being the only enzyme that can reduce biopterins (Sienkiewicz et al. 2010 ), but not in Leishmania amastigotes that also have a quinonoid dihydr obiopterin r eductase (QDPR) for biopterin metabolism (Ong et al. 2011 ). PTR1 and the DHFR domain of DHFR-TS have both been structur all y determined in complex with inhibitors, and it has been possible to de v elop drugs tar geting PTR1 alone as well as dual inhibitors affecting both enzymes with high specificity against the parasite over mammalian cells (Mpamhanga et al. 2009, Vanichtanankul et al. 2011, Khalaf et al. 2014, Pohner et al. 2022. The dual inhibitors have the advantage that they also can be used against Leishmania amastigotes, which have QDPR that can compensate for the inhibition of the biopterin metabolism but not the inhibition of folate metabolism (Ong et al. 2011 ). Only one class of dual inhibitors of DHFR and PTR1 has so far been tested in animal studies, but these inhibitors were considered toxic (Khalaf et al. 2014 ) and more in vivo data are needed to know if it is a pr omising a ppr oac h.
It is not only the DHFR-TS protein that is unique in T. brucei dTTP synthesis. Perha ps e v en mor e sur prising was the finding of a new pathway to produce dUMP, which is needed as a substrate of TS (Fig. 6 ). In most organisms, the majority of dUMP comes from the deamination of deoxycytidine nucleotides (dCMP or dCTP depending on the species). This is because RNR gener all y has a compar abl y high affinity for CDP and produces a surplus of dCDP compar ed to dUDP, whic h subsequentl y forms the dCMP or dCTP substrates for the deamination reaction. Ho w ever, T. brucei lacks dCMP deaminase and dCTP deaminase, which was initially belie v ed to be an adaption to save cytidine nucleotides that are present at very low levels in BSFs (Hofer et al. 2001 ). Surprisingly, it was instead found that T. brucei has a novel pathway for dUMP pr oduction fr om dCTP that is mainly mitochondrial (Moro-Bulnes et al. 2019 , Ya gue-Ca pilla et al. 2021 ). T he pathwa y starts with the r emov al of all three phosphates from dCTP by a mitochondrial orthologue of the mammalian enzyme sterile α-motif histidineaspartate domain-containing protein 1 (SAMHD1) (Yague-Capilla et al. 2021 ). The mammalian SAMHD1 dephosphorylates dNTPs both as a way to limit their concentrations when they are not needed (outside S-phase) and as a defense system against DNA viruses. Knocking out the T. brucei orthologue, HD52, led to increased dCTP and decreased dTTP pools, and it was concluded that dCTP must be an important HD52 substrate (Yague-Capilla et al. 2021 ). The gene was found to be conditionally essential, and the trypanosomes could be rescued by supplying thymidine in the growth medium, indicating that the enzyme is important for dTTP synthesis (Ya gue-Ca pilla et al. 2021 ). The second mitochondrial enzyme in the pathway is cytidine deaminase, which converts the deoxycytidine formed in the pr e vious step into deoxyuridine (Moro-Bulnes et al. 2019 ). The deoxyuridine is subsequently exported to the cytosol where thymidine kinase uses it for the synthesis of dUMP, which needs to be transported back into the mitochondrion to be used as a substrate of DHFR-TS (Moro-Bulnes et al. 2019 ).
It is not yet known how the dTTP synthesis pathway is regulated. In mammalian cells, dCMP deaminase is inhibited by dTTP and activated by dCTP, and it can thereby regulate the proportion of dCTP that is converted into dTTP. In T. brucei , DH52 is a possible candidate for the regulation of the dCTP-dTTP balance . T he corresponding mammalian orthologue, SAMHD1, is an allosteric enzyme, whereas cytidine deaminase, TS, and thymidine kinase gener all y lac k true allosteric r egulation (thymidine kinase can be inhibited by dTTP, but this is only a competitive inhibition via the catalytic site). An interesting conclusion from the study of T. brucei dTTP synthesis is the unique role of thymidine kinase, which is involved in both de novo dTTP synthesis and thymidine/deoxyuridine salv a ge (Leija et al. 2016, Mor o-Bulnes et al. 2019. The knoc k out/knoc kdown studies indicated that thymidine kinase is essential in contrast to cytidine deaminase and other enzymes in the pathway that were shown to be conditionally essential and could be rescued by supraphysiological thymi-dine concentrations (Leija et al. 2016 ). Ho w ever, the conditionally essential enzymes might also be of interest for drug discovery because, as mentioned above for DHFR-TS, thymidine auxotrophs ar e gener all y noninfectious because thymidine concentrations in blood and other body fluids are not sufficient to support parasite survival, although it remains to be verified that they cannot regain infectivity via gr adual ada ptation to thymidine starvation. Other trypanosomatids share some of the features with T. brucei dTTP metabolism-including DHFR-TS and thymidine kinaseand they also lack dCMP deaminase, but it is not known if the whole pathway is conserved.
An alternative way to make dUMP is by dephosphorylating deo xyuridine n ucleotides (produced by RNR) through the activity of dUTPase . T he trypanosomatid enzymes belong to the dimeric class of dUTP ases, whic h ar e mainl y bacterial and structur all y differ ent fr om the trimeric class of enzymes found in other eukaryotes as well as in some bacteria (Harkiolaki et al. 2004 ). Dimeric dUTPases use both dUTP and dUDP as substrates, and the T. brucei enzyme was found to be nuclear (Castillo-Acosta et al. 2008 ). Knoc k out/knoc kdo wn studies sho w ed that enzyme depletion led to growth arrest and increased incidences of mutations, str and br eaks, and v ariable surface gl ycopr otein switc hing (Castillo-Acosta et al. 2008, Castillo-Acosta et al. 2012 ), a phenotype that was found to be linked to abasic sites caused by dUTP incor por ation and subsequent base excision repair. The growth arrest could be rescued by adding high concentrations of thymidine in the growth medium. This was initially taken as evidence that the enzyme was needed for the supply of dUMP in dTTP synthesis (Castillo-Acosta et al. 2012 ), but this conclusion was made befor e the discov ery of the pathway to make dTTP from dCTP, and it could instead be an effect of increased levels of dTTP competing with dUTP for binding to DNA pol ymer ase.

Salvage synthesis of deo xyribon ucleotides
Salv a ge of deo xyribon ucleosides is limited in T. brucei ; ther e ar e only two enzymes able to phosphorylate deo xyribon ucleosides, and only the thymidine kinase is fully dedicated for such purpose (Fig. 6 ). This is in contrast to mammalian cells, which have four dedicated enzymes and the ability to phosphorylate all natural deo xyribon ucleosides in the c ytosolic as w ell as mitochondrial compartments. T. brucei does not have any mitochondrial deoxyribonucleoside kinases. A main difference between the organisms is that T. brucei and other trypanosomatids have a single mitochondrion that replicates its DN A in coor dination with the cell cycle, whereas mammals need to supply dNTPs for mitochondrial biogenesis and DNA repair also in resting cells with very low dNTP pools, and thus the ability to salv a ge deo xyribon ucleosides can be an additional source of dNTPs.
One of the enzymes able to phosphorylate deo xyribon ucleosides in T. brucei is the adenosine kinase described in the purine section, which phosphorylates both adenosine and deoxyadenosine. Ho w e v er, deoxyadenosine needs to be at supr a physiological concentrations to circumvent cleavage by MTAP and is not likely to be r ele v ant for r egular dATP synthesis (Vodnala et al. 2016 ). On the contrary , T . brucei grown in the presence of 1 mM deoxyadenosine expands its dATP pool to become the major nucleotide in the cell, and ATP that is used for deoxyadenosine phosphorylation drops to a third of its original level (Hofer et al. 1998, Vodnala et al. 2016 ). This situation is very harmful to the trypanosomes, which die within a few hours. At these high deoxyadenosine concentrations, MTAP cannot protect the trypanosomes due to that it becomes product-inhibited by the generated adenine (Vodnala et al. 2016 ). Although deoxyadenosine itself is not useful as a drug because of the high concentrations needed to avoid MTAP cleav a ge, analogues of deoxyadenosine that are resistant to the cleavage can be efficient antitrypanosomal agents (Ranjbarian et al. 2017 ). It was then concluded that the antipr olifer ativ e effect was due to DNA synthesis inhibition, whereas the cytotoxic effects seen at higher drug concentrations could be a result of the observed ATP depletion.
The second deo xyribon ucleoside salv a ge enzyme in T. brucei is thymidine kinase, which as described above is also involved in de novo synthesis . T his enzyme differs from most other thymidine kinases, including those from Leishmania and T. cruzi , by being a tandem enzyme with two homologous domains in T. brucei and four homologous domains in T. congolense (Ranjbarian et al. 2012 ). The tandem structure was also found in two unrelated pr otozoan par asites Tric homonas v aginalis and Giardia intestinalis , as well as in the parasitic round worm Ascaris suum indicating an ov err epr esentation in par asites (Ranjbarian et al. 2012, Kr ak ovka et al. 2022 ). Studies of the T. brucei thymidine kinase sho w ed that the C-terminal domain contains the enzymatic activity, while the N-terminal domain helps the active domain to ac hie v e str onger binding to its substr ate. Inter estingl y, the enzyme has a broader substrate specificity than other studied thymidine kinases. It uses thymidine and deoxyuridine as its main substrates, but it can also use deoxyinosine and deoxyguanosine. Ho w e v er, the addition of deoxyguanosine in the growth medium did not lead to expansion of the T. brucei dGTP pool, suggesting that its salv a ge is not very efficient (Hofer et al. 1998 ). Nevertheless, it should be remembered that the effect of a substrate analogue against the parasite that exploits thymidine kinase will also be dependent on the P1 tr ansporter, whic h has a m uc h higher affinity for the purines than thymidine (an ∼100-fold difference). A possible drug de v elopment str ategy could ther efor e be to de v elop thymidine analogues with additional functional groups to increase the affinity for the P1 transporter. It would then be an adv anta ge that the thymidine kinase has a br oad substr ate specificity and ther efor e can toler ate mor e c hanges in the substrate analogue than the mammalian enzyme. Two other factors that make the de v elopment of thymidine kinase substrate analogues interesting is that the enzyme seems to be essential (Leija et al. 2016 ) and that the transport is mediated by the P1 transporter family. Both factors minimize the risk of de v eloping drug r esistance. Studies of the pr otozoan par asite G. intestinalis indicate that substrate analogues of thymidine kinase can be efficient antiprotozoal agents (Krakovka et al. 2022 ). Figure 6 also contains some reactions with question marks, including reactions performed by an HD-domain 5´-nucleotidase pr oposed to hav e a r ole in deo xyn ucleotide metabolism (Leija et al. 2016 ). The enzyme can to various extents use ribonucleoside and deo xyribon ucleoside monophosphates as substr ates, especiall y dCMP , dUMP , and dTMP . Potential but not tested substrates of HDdomain 5´-nucleotidase (dAMP , dIMP , and dGMP) are marked with question marks in the figur e. A r elated uncertainty that is indicated in the figure is that it is not clear if significant le v els of dN-MPs ar e normall y pr oduced in the cell. Onl y limited amounts can be expected from nonspecific phosphatase activities, from the occasional use of dNTPs by enzymes using NTP-to-NMP conversions as an energy source, and from DNA repair. The only dNMPs that ar e pr oduced as a part of a pathway ar e dUMP and dTMP, whic h ar e pr oduced in the de novo synthesis of dTTP. Figure 6 also includes a reaction catalyzed by T. brucei NDT, which primarily interconverts deo xyribon ucleosides but to some extent also ribonucleosides as described in the purine salv a ge section. The enzyme exchanges the nucleobase on 6-oxopurines with free purine nucle-obases (A, G, or H) fr om the surr oundings and can ther eby establish an equilibrium between deoxyguanosine and deoxyinosine and also make deoxyadenosine from any of the two other purine deo xyribon ucleosides (Del Arco et al. 2019 ). Howe v er, the physiological r ele v ance of NDT fr om T. brucei and other trypanosomtids is unclear. Deoxyadenosine is the only one of the three deoxyribonucleosides that is efficiently salvaged in T. brucei , but it is gener all y cleav ed by MTAP under physiological conditions and e v en if some additional deoxyadenosine can be gener ated fr om deoxyinosine/deoxyguanosine, it is not likely to be sufficent to overcome MTAP cleav a ge.

Concluding remarks
Knowledge about nucleotide metabolism in T. brucei and other trypanosomatids has expanded significantly in recent years, leading to a better understanding of how the different pathways are compartmentalized and which targets are most promising to focus on for drug de v elopment. These findings can be summarized as follows: r Purines can only be made by salvage synthesis in trypanosomatids. Some reactions are cytosolic, whereas others are glycosomal. A full set of interconversion enzymes between AMP, IMP, and GMP makes it possible for the parasites to survive on any purine source.
r Purine uptake is dependent on high-affinity active transporters in trypanosomatids. In T. brucei , the P2 transporter is r elativ el y easy to downregulate and is associated with drug resistance. It is therefore preferable if purine nucleoside analogues are mainly taken up by the P1 transporter family instead.
r The most successful drug de v elopment pr ojects hav e been with purine analogues. Adenosine analogues can cure T. brucei -infected mice, and some of them also cur e br ain infections. Nucleobase analogues (primarily allopurinol) have been more successful against the other trypanosomatids and allopurinol is a common treatment against feline and canine leishmaniasis.
r Most adenosine analogues ar e activ e in their triphosphate form, and in order to be effective against T. brucei they should be mainly taken up by the P1 transporter, be resistant to deamination and cleav a ge, and be efficientl y phosphorylated. Figure 4 provides a guide for further development of adenosine analogues against T. brucei . r UTP can be made both by de novo pyrimidine synthesis and via ur acil/uridine salv a ge in trypanosomatids, but de novo synthesis is gener all y dominating. De novo pyrimidine biosynthesis is mainly cytosolic with the exception of UMP synthase, whic h is gl ycosomal. The m ultitude of differ ences between pyrimidine synthesis in T. brucei and mammals is highlighted in Fig. 5 . r CTP can only be made by de novo synthesis in T. brucei , and the CTP pools ar e extr emel y small in BSFs but not in procyclics. Mor e r esearc h is needed to determine the pur pose of having such low CTP pools in T. brucei BSFs.
r Trypanosoma brucei dTTP synthesis is dependent on an entir el y ne w mainl y mitoc hondrial pathwa y. T his pathwa y, which has not yet been studied in the other trypanosomatids, includes a mitochondrial dCTP hydrolase (DH52), mitochondrial cytidine deaminase, cytosolic thymidine kinase, and mitochondrial DHFR-TS (Fig. 6 ). Other dNTPs can be made via RNR. r CTP synthetase, RNR, and most likely thymidine kinase are the onl y completel y essential enzymes in T. brucei nucleotide metabolism. They are therefore potentially good targets for c hemother a py. Especiall y the r ecent discov ery that thymidine kinase seems essential came as a surprise . T his is the first time a thymidine kinase has been shown to be involved in both salv a ge and de novo dTTP synthesis in an y or ganism.
r Ther e ar e man y conditionall y essential enzymes, whic h when knocked out make T. brucei dependent on growth medium supplements but GMP synthase is the only one of these enzymes verified to be essential for establishing infections in animal models so far but also in this case, it remains to be verified that the parasites cannot adapt to the situation by the upregulation of compensatory pathwa ys .
r The nucleotide metabolism of T. brucei is to a large extent conserved in other pathogenic trypanosomatids but crucial experiments are often lacking. Consequently, it is not known if the nov el mitoc hondrial dTTP synthesis pathway in T. brucei described above exists in T. cruzi and Leishmania , and e v en less is known about the nucleotide metabolism in T. congolense and T. viv ax . Mor e r esearc h is needed to fill in this ob vious ga p of knowledge.