Nucleoside- and nucleotide-analogue reverse-transcriptase inhibitors (NRTIs) require intracellular phosphorylation for anti–human immunodeficiency virus (HIV) activity and toxicity. Long-term toxicities associated with NRTIs may be related to overactivation of this process. In vitro experiments have shown increased rates of NRTI and endogenous nucleoside phosphorylation to be associated with cellular activation. Patients with advanced HIV disease often have overexpression of cytokines, which corresponds to an elevated cellular activation state. These patients also have higher rates of NRTI phosphorylation and NRTI toxicity, suggesting an interaction between a proinflammatory biological state, NRTI phosphorylation, and toxicity. Studies suggest that women may have higher rates of NRTI phosphorylation than do men, as well as increased risk for NRTI-induced toxicity. Future research is needed to understand the NRTI activation process and improve the long-term toxicity profile of NRTIs. Such research should include comparisons of NRTI phosphorylation according to sex and cellular activation state (i.e., elevated vs. low).
Treatment of HIV infection with combination antiretroviral therapy (ART) is a long-term undertaking [1, 2]. Although ART has been proven to reduce morbidity and mortality, drug toxicity issues have dampened enthusiasm for these accomplishments [3–5]. Nucleoside- and nucleotide-analogue reverse-transcriptase inhibitors (NRTIs) play a central role in the treatment of HIV. Eight NRTIs are currently available: abacavir, didanosine, emtricitabine, lamivudine, stavudine, tenofovir disoproxil fumarate (nucelotide analogue), zalcitabine, and zidovudine. Several other NRTIs are in various stages of clinical development. Long-term toxicities attributed to NRTIs include hyperlactatemia and lactic acidosis, hepatomegaly with steatosis, peripheral neuropathy, myopathy and/or cardiomyopathy, ototoxicity, cytopenias, pancreatitis, and lipoatrophy [3, 4, 6–10]. Cross-sectional studies, summarized in table 1, identify recognized risk factors for the most common of these events. Limitations of the analyses—such as study design, lack of a uniform toxicity definition (particularly in the case of lipodystrophy), and evaluations during ART compared with those during monotherapy or dual NRTI therapy [10, 27, 35, 38]—should be considered when evaluating these risk factors.
Although NRTI toxicities are a major problem for patients infected with HIV, pharmacological research in this area is limited, in large part because measuring the active intracellular NRTI triphosphate concentrations in patients is difficult. The objective of this review is to address cellular NRTI activation as it relates to NRTI toxicity and to identify focus areas for future research. Information was included from epidemiological studies of NRTI toxicity, descriptive studies of NRTI phosphorylation in patients, and in vitro experiments.
For all NRTIs, it is essential to consider drug exposures in terms of intracellular NRTI triphosphate concentrations, because these are the moieties that exert antiretroviral and toxic activities. This creates a distinctive therapeutic index for NRTIs, as depicted in figure 1, in which cellular NRTI activation can be one factor that influences antiretroviral effects and drug toxicity.
The cellular activation of NRTIs produces at least 2 distinct sets of pharmacokinetic dispositions, one for the biologically inactive drug in plasma and the other for the active NRTI phosphate in cells . Although underlying relationships probably exist between plasma NRTI concentrations and intracellular NRTI phosphate concentrations, these relationships are currently unpredictable in patients [44, 45]. This is likely the result of rate-limiting or saturated-phosphorylation steps and the overall biological complexity of the system. A simplified diagram of the intracellular activation of the currently available NRTIs is shown in figure 2 [46–48].
Most of the clinical manifestations of NRTI toxicities resemble mitochondrial diseases, and histologic evidence demonstrates abnormal mitochondria and/or mtDNA depletion in affected tissues [22, 47, 49–51]. Studies show that NRTI triphosphates competitively inhibit mtDNA polymerase γ in vitro [52, 53]. This, in turn, may decrease the number of mitochondrial respiratory chain proteins, inhibit aerobic respiration, induce oxidative stress, increase mutation in mtDNA, and result in mitochondrial and/or tissue failure . This mechanism of toxicity is a selectivity/specificity problem, and, therefore, toxicity will strongly depend on drug dose and concentration, as shown in figure 1. As such, elevations in NRTI triphosphate concentrations will significantly impact the mitochondrial toxicity of NRTIs.
To date, the cellular activation state for a given cell type has been determined to be the most influential factor for increased generation of intracellular NRTI phosphate concentrations, although this was elucidated in vitro. Cells treated with phytohemagglutinin or granulocyte-macrophage colony-stimulating factor (GM-CSF) generated 2- to >150-fold higher triphosphate concentrations of zalcitabine, lamivudine, stavudine, zidovudine, and didanosine (2′,3′-dideoxyadenosine triphosphate) than resting cells in vitro [55–59]. Elevated cell activation results in high nucleic acid synthesis and an upregulation of kinases that phosphorylate NRTIs [57, 60]. It must be noted that zalcitabine, lamivudine, and didanosine are more active virologically in resting cells than in activated cells, which has been attributed to a more favorable ratio of NRTI triphosphate to endogenous nucleoside triphosphate (e.g., lamivudine triphosphate to 2′-deoxycytidine-5′ triphosphate ratio) in resting cells [57, 58]. Nevertheless, the actual concentrations of lamivudine, zalcitabine, and 2′,3′-dideoxyadenosine triphosphates were ∼2–4-fold higher in activated versus resting cells [55–58]. It is not known whether the same cytoplasmic (whole cell) ratio is as relevant to NRTI toxicities in the mitochondrial compartment as it is to antiviral effects in the cytoplasmic/nuclear compartment.
The higher incidences of pancreatitis and peripheral neuropathy observed in patients receiving hydroxyurea—a ribonucleotide reductase inhibitor—were hypothesized to be caused by decreased endogenous deoxynucleotide pools in cytoplasm and, thus, a ratio favoring the NRTI triphosphates, particularly didanosine (2′,3′-dideoxyadenosine triphosphate) . However, a study in patients treated with hydroxyurea could not detect changes in endogenous deoxynucleotide pools, including deoxyadenosine triphosphate pools . Hydroxyurea may also upregulate salvage pathways of nucleoside phosphorylation, and it arrests cells in the G1-S phase, which may upregulate generation of NRTI phosphates and could also explain increased toxicity . These multiple possibilities of hydroxyurea effects make it difficult to resolve the importance of the cytoplasmic ratio for NRTI toxicities in the mitochondrial compartment.
In the mitochondrial compartment, there are kinases that potentially phosphorylate NRTIs, but a nucleotide carrier protein may also shuttle nucleotides from cytoplasm into the mitochondrion [47, 63]. This nucleotide carrier has a stronger affinity for many NRTI phosphates than the corresponding endogenous nucleotides, which may influence the importance of the ratio with cytoplasmic endogenous nucleotides . As a possible illustration of the activity of this carrier, one in vitro study with zidovudine could not generate mitochondrial toxicity during exposure of isolated mitochondria to zidovudine. Instead, zidovudine was extremely toxic to mitochondria in activated whole cells, compared with resting whole cells, which may indicate significant NRTI phosphate shuttling into mitochondria to elicit the toxicity .
In Vivo Versus in Vitro NRTI Triphosphate Concentrations and Toxicity
Thus far, only 1 study has addressed a possible relationship between NRTI phosphate concentrations in patients and NRTI toxicity: higher total zidovudine phosphate concentrations (mono-, di-, and tri-) in PBMCs were associated with reduced hemoglobin levels during zidovudine monotherapy, whereas no such relationships were found with plasma zidovudine concentrations . Table 2 compares the typical NRTI triphosphate pharmacological data descriptively measured in patient's PBMCs, along with the NRTI triphosphate concentrations that inhibit mtDNA polymerase γ in vitro. Intracellular concentrations of 2′,3′-dideoxyadenosine, emtricitabine, lamivudine, and stavudine triphosphates reach levels near the in vitro binding affinity for polymerase γ. Table 2 also shows the in vitro therapeutic index of each NRTI; a low number indicates a narrow therapeutic index. Lamivudine values may need to be viewed in light of the exonuclease function of polymerase γ. Lamivudine monophosphate is excised 750-fold more rapidly than zalcitabine monophosphate, which presumably lessens the toxicity of lamivudine relative to that of zalcitabine . The other NRTIs are excised at rates between those of these 2 agents .
Clinically, zidovudine, didanosine, stavudine, and zalcitabine all underwent dose de-escalations during development to approximately one-half of the present-day dose, because the probability of toxicity was unacceptably high (mainly cytopenia and myopathy were associated with zidovudine, and pancreatitis and peripheral neuropathy were associated with didanosine, stavudine, and zalcitabine) [6, 74–78]. This narrow margin of drug exposures that elicited the increased risk of clinical toxicity is indicative of a low therapeutic index, as depicted in figure 1. Zidovudine seems to have more clinical toxicity than would be predicted on the basis of the values in table 2. Zidovudine may exert activity on cellular and mitochondrial processes other than the inhibition of polymerase γ, such as inhibition of adenylate kinase, impairment of the mitochondrial ADP-ATP translocator, and uncoupling of the electron transport chain [47, 79].
Of importance, the exact biochemical events that elicit clinical toxicities are not well understood, and specific pharmacological questions remain: Are concentrations of NRTI mono- or diphosphates important? Do changes in endogenous nucleotide pools in cytoplasm versus those in mitochondria contribute to toxicity? How do NRTI phosphates appear and disappear in the cytoplasmic versus mitochondrial cell compartments? Why do different tissues have different NRTI toxicity profiles?
Intracellular NRTI Phosphate Levels and Clinical Toxicities in Women
In small descriptive studies of NRTI phosphorylation, one characteristic that correlates with higher intracellular PBMC concentrations is female sex. One study found median lamivudine triphosphate and zidovudine triphosphate concentrations that were 1.6- and 2.3-fold higher, respectively, in a group of 4 HIV-infected women than those in a group of 29 men who initiated zidovudine, lamivudine, and indinavir (P < .01), although zidovudine or lamivudine plasma concentrations did not differ according to sex . A second study reported mean total zidovudine phosphate levels that were 2-fold higher in a group of 5 women than those in a group of 16 men (P = .004), and another report described carbovir triphosphate (the active triphosphate for abacavir) levels in a single woman that were 2–8-fold higher than the levels in a group of 4 men [43, 66]. These data provide pharmacological insight that is consistent with epidemiological studies that showed 4-fold more lipodystrophy during dual therapy with NRTIs, disproportionately high rates of lactic acidosis and pancreatitis, and stronger antiviral responses to NRTIs in women versus men, respectively [27, 11–13, 80]. Of interest, women with cancer experienced 1.5-fold–more severe toxicities with the use of the anticancer nucleoside analogue fluorouracil than did men, after adjustment for study, dose, body size, and age (P < .0001) . Fluorouracil requires cellular activation for biological activity, using steps similar to those of zidovudine and stavudine .
Intracellular NRTI Phosphate Levels in Patients with Advanced HIV Disease
It has also been found that the presence and severity of HIV disease correlates with higher concentrations of intracellular NRTI phosphates in PBMCs [83, 84]. Among persons who received zidovudine monotherapy, the lowest intracellular zidovudine phosphate (mono-, di-, and tri-) concentrations were in healthy volunteers, followed by 5- and 12-fold higher concentrations in HIV-infected patients with CD4 cell counts of >100 and <100 cells/mm3, respectively . One small study also found high zidovudine monophosphate concentrations in patients with low CD4 cell counts, although the zidovudine triphosphate levels were lower .
A relationship between HIV disease severity and NRTI phosphorylation was also manifested in rates of NRTI phosphorylation in patients with advanced disease that were higher during the early stages of treatment relative to rates during later stages. In ART-naive patients initiating therapy with zidovudine, lamivudine, and indinavir, zidovudine triphosphate levels were 2.5-fold higher early in therapy (at week 2) in patients with a CD4 cell count of <100 cells/mm3 than they were in patients with higher CD4 cell counts [44, 87]. However, after 1 year of therapy, the zidovudine triphosphate levels in patients with advanced disease were reduced to the concentration range observed in patients with mild disease [44, 87]. In another zidovudine monotherapy study, the highest zidovudine phosphate concentrations were observed on day 1 of therapy; these concentrations decreased to ∼30% of this value when measured again 6 months later . These data suggest that the presence of HIV infection and advanced HIV disease status are associated with higher NRTI phosphate concentrations, particularly just after initiation of therapy.
This provides pharmacological insight that is consistent with the advanced disease risk factor for NRTI toxicity. For example, a CD4 cell count of <100 cells/mm3 was the strongest predictor of lipoatrophy for NRTI-treated subjects in the HIV Outpatient Study (HOPS) cohort . NRTI-associated peripheral neuropathy is ∼2-fold more common in persons with CD4 cell counts of <100 cells/mm3 than it is in persons with higher CD4 cell counts . Neuropathy risk in the HOPS cohort was highest in patients with CD4 cell counts of <100 cells/mm3 during the first months after initiating NRTI therapy, but it subsequently decreased over time. This finding is similar to the initially elevated and later decreased NRTI phosphate concentrations described in patients with advanced disease (Lichtenstein et al., unpublished data). In the dose de-escalation studies described above, patients with CD4 cell counts of <100 cells/mm3 experienced significantly increased NRTI toxicities [77, 88].
Some investigators hypothesized that the genetic sequence of the polymerase γ enzyme may vary among individuals, resulting in increased or decreased affinity for NRTI triphosphates. Specific sequence variants of the mtDNA polymerase γ gene obtained from 14 patients with NRTI-associated lactic acidosis or peripheral neuropathy were compared with sequences from 45 patients without NRTI toxicity. The investigators found no correlations between variations in the DNA polymerase γ gene sequence and these NRTI-associated toxicities. Instead, low CD4 cell counts were found to be predictive of neuropathy .
Relationship Between Cellular Activation and NRTI Phosphorylation in Patients
Just as cellular activation increases NRTI phosphorylation in vitro, an elevated state of cellular activation in patients with advanced disease may be a biological mechanism for increased NRTI phosphorylation in vivo. HIV infection and advanced HIV disease are associated with highly elevated concentrations of proinflammation and cellular activation markers, such as proinflammatory serum cytokines and molecules, IFN, TNF, and soluble TNF receptor type 2 (sTNFrII), and lymphocyte activation markers, such as CD38+ cells and HLA-DR+ [89–93]. For instance, compared with healthy volunteers, TNF and sTNFrII levels were 4–25-fold higher in patients with Centers for Disease Control and Prevention (CDC) class A and B HIV infection and 10–40-fold higher in patients with CDC class C HIV infection . As described in the previous section, these findings are similar to the apparent rates of NRTI phosphorylation in the same patient groups . In addition, treatment of advanced disease is known to significantly reduce the concentration of cell activation markers; this finding also corresponds to the reduction in the intracellular level of NRTI phosphate observed in patients with advanced disease after the disease had been controlled [44, 94]. Together, this points to a possible relationship between proinflammatory cellular activation and intracellular NRTI pharmacology and toxicity, but this possibility has not been adequately studied. One clinical study assessed total levels of zidovudine phosphate in PBMCs of patients initially treated with zidovudine alone followed by the cytokine GM-CSF plus zidovudine. GM-CSF is known to increase cell activation in patients, as shown by increased expression of HLA-DR+ on monocytes . An overall trend toward higher zidovudine phosphate levels was observed during combination GM-CSF–zidovudine therapy in the entire cohort (P = .07), and higher doses of GM-CSF led to higher total zidovudine phosphate levels (P = .01) .
Table 3 presents relationships among proinflammatory cellular activation, NRTI triphosphate concentrations, and the epidemiological NRTI toxicity risk factors identified in table 1—including female sex, white race, age, hepatitis B and C coinfection, nadir CD4 cell count, and concomitant protease inhibitors. To provide some additional considerations, in animal models, NRTI toxicity could not be generated unless experimentally induced HIV infection was also present . In phase 1–2 studies with fialuridine—an investigational anti–hepatitis B virus nucleoside analogue that was urgently halted during the developmental phase because 7 of 15 patients died or required liver transplantation—analysis of liver biopsy specimens obtained before initiation of treatment showed higher inflammation scores for the 7 patients who died or underwent liver transplantation than for patients who averted serious fialuridine toxicity [117, 118].
Coinciding with the possibility that advanced HIV disease and the corresponding cellular activation may increase NRTI triphosphate concentrations and thereby increase toxicity, inflammation may also cause biological harm that increases susceptibility to NRTI toxicities in an additive way [119, 120]. For illustration, ART-naive HIV-infected patients already show evidence of mtDNA depletion in PBMCs, compared with healthy volunteers, and, thus, NRTI-induced mtDNA depletion could be adding to preexisting mtDNA depletion . In addition, the inflammation associated with HIV infection itself may harm bystander tissues, which occurs in HIV-associated peripheral neuropathy. Pathological analyses of nerve specimens obtained from these subjects indicate that activated macrophages infiltrate nerve tissue and may secrete destructive molecules . Consistent with this finding, HIV-associated peripheral neuropathy occurs most frequently in the later stages of HIV infection. Patients with a history of non–drug-related neuropathy are at increased risk for NRTI-associated neuropathy [24, 122].
These observations identify a need to investigate how inflammation and cytokines affect NRTI cellular pharmacology and toxicity. One avenue for study should include determining how certain cytokines activate certain tissues and how this may stimulate high rates of NRTI phosphorylation in that tissue or cell type, which could contribute to the tissue selectivity of NRTI toxicities. For example, patients with ART-associated lipoatrophy and/or lipodystrophy express more TNF mRNA in subcutaneous fat and exhibit higher serum levels of sTNFrII and TNF than do control subjects [100, 101, 104, 123]. TNF exerts activity on fat [91, 124], and elevated TNF levels could specifically upregulate NRTI phosphorylation and/or susceptibility to NRTI toxicity in fat cells and tissue.
Special Considerations for Cellular NRTI Activation with Concomitant Anti–Hepatitis C Virus Therapy
In the mid-1990s, studies evaluated the use of the proinflammatory cytokine IFN in combination with NRTIs to treat HIV infection. Increased NRTI toxicities were observed. One randomized study of IFN-α with or without zalcitabine-zidovudine demonstrated significantly more peripheral neuropathies (P = .02) and cytopenias (P = .03) in the combination IFN-α arm than in the zidovudine-zalcitabine only arm . In a phase 1–2 study of didanosine alone followed by IFN-α–didanosine, excess clinical pancreatitis or elevations of serum amylase/lipase was observed during IFN-α–didanosine combination therapy (didanosine doses of ⩽250 mg b.i.d. were associated with a pancreatitis incidence of 24%) .
Present concerns for drug-drug interactions between IFN-ribavirin and ARTs are focused on intracellular interactions between ribavirin and NRTIs, because in vitro studies showed that ribavirin enhanced conversion of didanosine to the active 2′,3′-dideoxyadenosine triphosphate [108, 125]. In vitro studies with ribavirin also showed that zidovudine and stavudine were less efficiently converted to their triphosphates [108, 126, 127]. Clinically, the risk of lactic acidosis and/or pancreatitis was 5% among patients coinfected with HIV/hepatitis C virus and cotreated with ART plus IFN-ribavirin, and the toxicity risk was significantly higher for didanosine recipients than for patients who received other NRTIs (OR, 18.3; P = .0002) . However, the individual effects of ribavirin versus those of IFN on the increased risk of didanosine toxicity are not separable.
An increased number of cases of hyperlactatemia have been reported among patients treated with stavudine combined with IFN-ribavirin, which is counterintuitive on the basis of the in vitro ribavirin data . This raises concern about the possible pharmacological interaction between IFN and stavudine. A small study of intracellular stavudine triphosphate concentrations in 15 patients before and after initiating IFN-ribavirin therapy found no statistical change in median peak concentrations 1 month after initiation of therapy, compared with baseline concentrations, although the range of peak concentrations increased (from a median of 26.4 and a range of 0.9–47.9 fmol per 106 PBMCs at baseline to a median of 14.7 and a range of 1.1–148.3 fmol per 106 PBMCs at month 1) . Future in vivo studies are needed to investigate intracellular interactions between NRTI phosphates and ribavirin, and such studies should also investigate the effects of the proinflammatory cytokine IFN.
Conclusions and Clinical Implications
First, this review underscores the pressing need for more research in the area of cellular NRTI pharmacology as it relates to NRTI toxicity, because these adverse events severely affect many patients. Research activity may be stimulated by new mass spectrometry methods that could simplify the process for measuring NRTI phosphate concentrations in patients . Although PBMCs are currently the only biological specimens in which NRTI phosphate concentrations can reasonably be measured, this should not discourage research because NRTI toxicities presumably occur at the mitochondrial level in tissues of nerves, the liver, fat, the pancreas, and so on. The data in this review were based on measurement of NRTI phosphate levels in PBMCs, and possible research avenues were identified on the basis of these data.
Second, the observations in this review highlight specific research questions that warrant high priority, because the clinical implications are important. For example, the strategy of postponing administration of ART to ART-naive patients as long as possible to circumvent toxicities—as is recommended by the current national guidelines—may in fact predispose these patients with a state of elevated biological activation to increased NRTI toxicity . Investigations are needed to address a possible association between inflammatory cytokines and cellular NRTI pharmacology in patients. Drug-drug interaction studies involving HIV and hepatitis C virus treatments should also investigate whether IFN upregulates NRTI phosphorylation in patients. Studies are needed to determine whether NRTI triphosphate levels are higher in women than in men, which could also provide pharmacological insights into the predisposition to fluorouracil toxicity in women. If NRTIs can be widely supplied to developing countries, these same issues may become more important, because women constitute the majority of the HIV-infected population of sub-Saharan Africa, and ARTs would presumably be rationed to patients with more-advanced disease on the basis of need . Ultimately, to provide the safest, most informed, and rational use of NRTIs for patients, additional studies of cellular NRTI pharmacology will be needed.
We thank Dr. Carlos Catalano and Dr. Dorie Hoody for helpful discussions.