My interest in AZT and HIV came about as a result of a news report I heard on the radio in 1996. According to that report, HIV had started to develop resistance to AZT, which at that time was the drug of choice for treating AIDS. This idea did not make much sense to me - AZT is a chemical that operates on a molecular level to stop DNA replication. It does this by binding to certain points on the strand of DNA to prevent the rest of the chain from being attached during the replication process. This process occurs on a very low level. It didn't seem possible that HIV, or any virus for that matter, would be able to circumvent that action, seeing as how it takes place on the molecular level.

I therefore decided to do a survey of the medical literature on the topic of HIV drug resistance to examine the evidence and determine the validity of the premise. I spent some months pouring over the scientific papers, and ultimately concluded that the evidence did not support the idea of HIV resistance to AZT or to any of the nucleoside analogs.

I wrote a paper with my findings and published it in 1997. Over the years I have seen nothing to make me reconsider my original conclusion, and therefore I am included the paper here. What you see below is a rewrite - I was never particularly satisfied with the editing, so I have rewritten it for clarity and grammar. Although AZT and nuceloside analogs are no longer as heavily used as they were in the first decade of AIDS, there are nonetheless lessons to be learned and further thought to the issue of nucleoside analogs. As far as I know, nobody has proven that a virus is capable of developing resistance to this particular group of chemicals. I suspect it is impossible for such a thing to happen.

The original publisher gives me permission to post my reprint.The link to the digital object identifier is as follows: doi:10.1016/S0306-9877(97)90208-5. Go to and enter the DOI to pull it up.

Research has not proven the existence of HIV resistance to nucleoside-analog drugs


The medical community believes that mutations in the pol gene of the human immunodeficiency virus (HIV) confer resistance to nucleoside-analog chemicals. However, laboratory techniques that test sensitivity to the drugs fail to provide direct evidence that this is the case. A rational and unbiased review of the clinical and laboratory evidence indicates to the contrary, that drug resistance does not take place under natural circumstances. A method to confirm or disprove this statement is provided in this paper.


The topic of HIV drug resistance is of much importance in AIDS research and therapeutic strategy. The ‘demise of monotherapy’ treatment strategies in favor of the combination approach is partly due to the belief that the virus develops resistance to antiviral compounds (1-5). With this in mind many researchers have tested the combination therapy approach (6-8). Using stronger drugs at higher doses, their goal is to stunt the emergence of drug resistance under real conditions (9).

A quantity of papers claim to show the virulence of drug resistant virus isolates in vitro, claim to correlate the in vivo appearance of viral mutations after the start of drug therapy (5), claim that the effectiveness of therapy diminishes and the patient clinical condition worsens as a result of the new viruses (9-13). According to these papers, drug resistance usually appears in patients after variable lengths of time on therapy, and the drug resistant viruses gradually begin to replace the drug sensitive viruses (9,12,14). Supposedly the power of the drug resistant strains varies with the drug, and cross-resistance may occur (5,6,15).

Resistance-conferring mutations

The medical community believes that the mechanism of HIV drug resistance arises in part from mutations in the pol gene coding for reverse transcriptase (RT) (3,5,15,16), specifically the following several amino-acid residues which researchers associate with “strong” resistance to AZT (17,18): codons 41, 67, 70, 215, 219 of HIV-1 pol (7). Likewise researchers have identified multiple sites of mutation corresponding to resistance to other inhibitors (6,8). Multiple mutations in a single virus seems to provide greater resistance than just a single mutation (3,5). Supposedly drug resistance is correlated with the following items: the appearance of the mutations; the start of drug therapy; and length of therapy. Patients who have been on drug therapy for longer periods of time contain larger percentages of mutated viruses (9,15,16,19,20). Mutant strains evolve in spite of aggressive therapy. Mutations conferring resistance to one drug fail to confer resistance to a different drug (5). Supposedly, subinhibitory amounts of AZT actually encourage the emergence of drug-resistant viruses (10,11). These observations seem to imply the necessity for higher doses of medication as part of combination therapy.

Postulated biochemical basis of resistance

Despite the mutation hypothesis, the molecular details for the biochemical basis of resistance remain unexplained (10.) The postulated mechanism involves the intracellular action of RT at the genetic level (5). The scientific literature proposes that mutations make HIV polymerase more selective about incorporating the analog and/or less inhibited by the competitive presence of the analog’s anabolites. According to those papers, one possible way this could occur is if the amino-acid residues help to discriminate between the legitimate 3’-OH group of endogenous nucleotides and the 3’-OH group belonging to the analog (3).

We don’t need flights of fancy to determine the theoretical ways by which a virus may become resistant to a nucleoside-analog. We only need to know how AZT suppresses HIV replication, and fortunately for us this is a well understood process.

How nucleoside analogs work

AZT suppresses DNA replication through the action of the triphoshate anabolite (AZT-TP). This anabolite is the only antiviral constituent of AZT(5). To get to its phosphate form, AZT is metabolized by cellular kinases (14,21). HIV on its own cannot metabolize it (22). Once in triphosphate form, AZT inhibits HIV replication through two actions: 1. as a linear competitive inhibitor of the process by which reverse transcriptase incorporatees the endogenous deoxythymidine-triphosphate (dTTP) into viral DNA; 2. as a DNA chain terminator (23-25). Nucleotide analogs other than AZT also inhibit HIV replication through similar action (26), the only difference being that the efficiency or antiviral strength varies according to the efficiency with which the 5’TP is formed (23). For example, among the major nucleotide analogs, ddCyd  (dideoxycytidine) has the greatest antiviral effect because it generates the most intracellular 5’TP. In contrast, ddThd (dideoxythymidine) generates the lowest concentration of 5’TP and has the least potent effect (23). The presence of 5’TP is the important consideration, not the method that produces it. For example, although it is believed that ddG, ddI, and ddC are activated by  kinases different than those responsible for AZT (21), these agents once in 5’TP form are equally as active and effective in inhibiting viral DNA polymerase (23).

The 5’TP is the key ingredient and how does it come into being? It depends on several factors: the ease with which the parent compound enters and leaves the cells, the affinity of the compound for the appropriate kinases, the activity and concentration of the kinases, and the susceptibility of the 5’TPs to degradation by phosphatases (23). The effectiveness of AZT and other dideoxynucleosides (ddNs) is the direct result of these factors.

Theoretical ways by which a virus may be resistant to these drugs

The above discussion tells us everything we need to know to identify all the possible ways by which drug resistance may occur. There are six:

  1. Reverse transcriptase is more selective. It imports fewer AZT anabolites into the DNA;
  2. Reverse transcriptase activity is less inhibited by the competitive presence of AZT anabolites;
  3. HIV deaminates AZT-TP or causes the cellular phosphatases to do  this;
  4. HIV prevents phosphorylation of AZT, either directly or by interfering with cellular kinases (since HIV does not carry its own kinase);
  5. HIV prevents AZT from entering and leaving the host cell;
  6. HIV upsets cellular nucleotide pools; there is evidence that the concentration of these pools affects the nucleotide analogs (21);

Does laboratory evidence show that drug resistance exists?

It should be noted that there is convincing evidence that in the real world the  resistance of the host cell to AZT may affect antiviral activity (10,11,14). But this is an unrelated process that should not be confused with the question of HIV drug resistance.

In my literature survey, I encountered the scientific examination of only one item on the list of six - item #2 - and have concluded that reverse transcriptase activity does not correlate with the strength of HIV resistance to AZT or other inhibitors. The affinities of ddN 5’TPs for reverse transcriptase do not vary in magnitude despite a 1000-fold variance in the antiviral potencies of the parent drugs (6,23). Richman (5) and Hao (23) found no enzymologic difference between wild and mutant RT (23). Larder and Kemp (3) found no correlation between RT activity and strength of resistance. The reverse transcriptases associated with both non-drug-resistant and mutant viruses are equally susceptible to inhibition by and show equal affinity for AZT-TP (3,19). Darby and Richman (19) demonstrate their prowess in logic by first dismissing the importance of the lab results by saying that cell-free enzyme assays are not accurate relfections of the mechanism by which AZT-TP inhibits the the transcription complex in the cell in vivo, and we should take it on faith that in the real world there is a different mechanism at work.

Despite the lack of test results for the other items in the itemized list of six, some evidence exists to descredit the likelihood of #1 and #3. Rooke found that the Ki and ID50 values of AZT-TP did not differ between his drug-resistant and drug-sensitive strains (27). Indeed, studies which make more effort at greater molecular analysis find fewer and fewer differences. For example, Lacey et al (28) admit paradoxical findings. Unfortunately, the next step in their logic is to conclude that the lack of evidence points to “incomplete understanding of AZT inhibition of HIV in the cell.”
What?! Since when does a rational person conclude that something must exist despite the lack of evidence that it exists?

Continuing to look at item #1, there are several unexplained flaws in the idea that HIV may learn how to discern between the proper endogenous physiological nucleotides and the nucleotide analogs. First is the incompatibility of the lab results with the clinical outcomes. For example, despite the supposed loss of viral sensitivity to ddNs there is no noticeable clinical or viriological consequence (7,19,29), even in patients whose viral population after a year of therapy mostly consists of “drug-resistant” strains or in patients whose viral population has undergone a 100-fold increase in the strength of resistance to drugs (3, 5, 6, 12,19). It makes no sense that there is no clinical consequence at all. One would expect a dire change in the condition of the patient. But the conclusion is that drug resistance is not causally associated with a change in the patient’s clinical response to AZT (2).

Despite the rapid mutation rate of HIV, there are no consistent findings of drug resistance in either the lab or the clinical environment. It takes six months for resistant virus to emerge and sometimes there is a dominant emergence while in other cases it is a slow trickle.  Not all samples exhibit drug resistance. If one were to go on these results, one would assume that mutations are not necessary nor sufficient for drug resistance. Suprisingly, it has been found that drug resistance is reversible with cessation of therapy, leading to the re-emergence of drug-sensitive strains of the virus that nonetheless retain the target mutations, the very mutations that supposedlyl cause drug-resistance (34)! But why be logical? (3) states that multiple mutations are necessary for high-level AZT resistance and other researchers take this as truth and continue to publish their papers.

Next let’s examine the discrepancy between the results obtained from different viral parentages. Larder and Kemp used site-directed mutagenesis to create two reverse transcriptase mutants with the target codon changes. The mutants exhibited reduced sensitivity to inhibition by AZT-TP, but ‘viruses constructed from infectious molecular clones containing these mutations displayed hypersensitivity to zidovudine, not resistance, when tested in culture’ (see ref 3, p. 1157). In other words, the viruses most resembling real-world conditions were more sensitive to the effects of AZT. If one were to draw broad conclusions from this experiment (which one should not because broad conclusions require repeated experiments with consistent results), one would say that the target codons play little to no role in drug resistance., meaning HIV mutation is not involved and one needs to look elsewhere for any causes of drug resistance.

Such inconsistent and conflicting reports are common throughout the literature. Researchers have found that exposure to drugs is not necessary for target mutations to arise. Sometimes  random mutation produces drug resistant strains in patients who do not have a history of drug therapy (12,33). If this result is reliable, then it would indicate that drug resistance is not dependent upon exposure to drugs, which would imply that there should be either an equal distribution of observed drug resistance across all patients or other insofar unidentified variables. In either case, the problem is that there is no adequate measurement or definition of “drug resistance” (see below the next section). Until one exists, observations are not reliable.

In the literature, the inconsistency of results is not only in respect to the mutation hypothesis but across the board. For example, some researchers observe cross-resistance (6,15,30,31) while others observe a lack of cross-resistance (8,16,19,32) even within the same report (5)!

Some researchers compare HIV with other viruses. Unfortunately they compare and oranges. For example, a ‘single amino acid change in the viral thymidine kinase [of herpes simplex virus] or DNA polymerase are sufficient to confer drug resistance’ to acyclovir (3, p. 1157, also see refs 19 and 33). Not suprising considering that HSV carries its only thymidine kinase while HIV does not (24). Unfortunately Larder and Kemp ignore this fact and express their puzzlement as to why HIV doesn’t behave like HSV.

Finally let’s look at the inconsistencies surrounding the different classes of nucloeotide analogs. The literature typically reports 100-200 fold increase in resistance to AZT and 4-10 fold increase in resistance to ddI and ddC. There have been intense efforts to create highly resistant genotypes (5,16) but no impact on the observed numbers:. In fact, the numbers don’t make sense at all. If drug resistance were a real phenomenon, one would expect the virus equally resistant to all nucleotide analogs.

All nucleotide analogs have equal affinity for HIV reverse transcriptase

Each of these chemicals is equally effective in destroying the virus’s ability to replicate. Therefore any mechanism that inhibits one class of nucleotide analog will just as effectively inhibit any other. Of the list of six possible methods by which HIV could resist a nulceotide analog,, #2 is the only list item that researchers have examined with any rigor. As I stated above, the results rule out #2. Nobody has tested  the remaining five items in a controlled environment that compares non-drug-resistant and mutated viruses while varying AZT dosage, type of host cell and viral parentage.

The  value in a literature survey is in taking a step back in order to see the forest and not the trees. In doing so, we see that neither the clinical nor experimental evidence supports the existence of a real world, bodily manifestation of HIV drug resistance.

Inconclusive laboratory results

Measurements for HIV drug susceptibility are not direct measurements. Rather they test for secondary effects assumed to be the consequence of resistance. There are a handful of tuch tests :  the plaque reduction assay; examination for syncitium formation; the ELISA for p24 antigen production; RT activity in cell-free supernatants; and the simple fact of cell death (see bibliography). Once the results of these tests are obtained, some researchers apply interpretive analysis: 50% inhibitory concentration (IC50); the Michaelis constant Km; and the inhibition constant Ki , to name a few. The final interpretation is usually expressed as inhibitory concentration. For example, to define drug resistance they may choose an arbitrary number like  IC50 > 0.05 micro-mol AZT.

Identification of nucleic acid sequences of drug-sensitive and drug-resistance viruses is done with PCR and other sequencing techniques.

As I went through the literature looking for the basis of measurement, I only was able to find studies with limited numbers of trials and small sample sizes - typically three to five experimental runs (14,30) with 10 patients (29), 24 and 26 patients (7), five patients (15), one (16), eighteen (35), twenty six (17) and twenty (8). Some researchers actually admit that these numbers are too small to draw conclusions. Lader et al state this in one of their papers(19) . Staszewski et al (33) state that they cannot extrapolate the in vivo effects of the observed NNRTI mutations due to the small samples size of their experiments. Most researchers, however, ignore this and proceed to use their results to construct measurements of drug resistance which they and other researchers later use as part of drug therapy recommendations.

There are two big problems with the ways in which HIV drug susceptibility are defined  First, the small sample sizes are not sufficient to draw conclusions. This is particularly relevant considering that HIV has a high mutation rate. In any population of HIV there are ‘huge reservoirs of genetic variants’ (33). The HIV reverse transcriptase itself  is particularly prone to errors. Any study using small numbers of patients or few iterations of the lab experiment fails to account for and control the genetic variance among the viral population.I A huge no-no if indeed , as everyone assumes, mutations are responsible for confering drug resistance.

Second, the definitions of drug resistance are based not on any standard. Most researchers assume that vrial mutation are responsible, but there is insufficient evidence to identify particular mutations in a causal role. Nielsen et al ‘cannot exclude random variation in susceptibility to zidovudine among HIV isolates’ (9, p. 329).

In general, the lab tests are not reliable because they don’t accurately reproduce what’s going on in the patient’s body. The cultivation of HIV in the lab heavily favors the dominance of genetic variance that doesn’t resemble the patient, due to infrequent clinical samples followed by cultivation and propagation in the lab. No researcher has ruled out the possibility that the adaptation of HIV to cell culture may perturb results (12).

Several pieces of evidence indicate drug resistance does not occur

Some results observed in the laboratory support the idea that there is no such thing as resistance to nucleoside analogs. Considering that these are the same laboratories producing less than stellar work as already noted in this survey, the results must be taken lightly. However, moving forward if researchers can duplicate the results consistently and with proper methods, it could be significant.

One observation is the intracellular amount of AZT concentration in T-cell lines. Both drug-resistant and drug-sensitive cells show the same amount of AZT uptake  and the same concentration of regular nucleoside triphosphates (11). This implies the effectiveness of AZT should be the same in both cells, and therefore there is no difference in “drug resistance.”  Refer to list items #5 and #6 in the  above section “Theoretical ways …”

Another observation is the amount of ddN activity towards reverse transcriptase. Recall list item #2. If HIV reverse transcriptase were to show less inhibition by ddN anabolites, this would be a legitimate indication of drug resistance. However, researchers in (23) did not find this to be the case. They observed that once phosphorylated to 5’TP form, the strength of affinity towards reverse transcriptase is the same across all the ddNs. In a template primer system, all six ddNs competitvely inhibit HIV RT. There is no correlation between a ddN modifying the dN pool and its antiviral activity in H9, MOLT4, CEM, or ATH8 cells.

In another study, the introduction of virus to ATH8 cells produces no change in the cell’s ability to phosphorylate ddCyd (26). This is evidence that any drug resistance is not due to the mechanism I described in list item #4.

If one is to make a case for HIV drug resistance, it is imperitive to control for cell type, the concentration of ddN that is phosphorylated to 5’TP, and viral parentage. If drug resistance truly exists, then when these parameters are controlled, it should be apparent that the concentrations of the physiological nucleotide pool and the ddN anabolite pool within cells infected by drug-resistant HIV are markedly different than cells infected by drug-sensitive HIV. For example, there may be a greater concentration of AZT tri-phosphates and a reduced concentration of physiological nucleotides in cells infected with drug-resistant virus. A readily noticeable difference is expected due to the extreme effect of ddN compounds on viral expression (3,5).

I suggest a method of a testing drug susceptibility: Use the fact that  the affinity of ddNs  ofr HIV reverse transcriptase is much greater than for cellular DNA polymerase alpha (22, 36). Recall that HIV does not influence the phosphorylation of AZT, meaning there are equal concentrations of AZT anabolites in cells infected by drug-resistant HIV as cells infected by drug-sensitive HIV. Therefore in cells with drug-resistant virus, there should be a greater concentration of active ddN anabolites and thus a greater amount of damage to cellular DNA polymerase.

Indeed there are only two legitimate methods of testing drug susceptibility: (a) by measuring intracellular levels of dN-TPs and ddN-TPS; and (b) measuring the disruption of cellular DNA.  Both methods will be sensitive to Yarchoan’s ‘unidentified metabolite of AZT’ (2, p. 866), if such metabolite exists. The technology exists to perform these tests - for example, PCA extraction allows the extraction of bases and nucleotide or nucleoside anabolites from cells and determination of their concentrations (11). However, nobody has used it to analyze viral drug resistance. Based on the literature survey, one can conclude only that the laboratory evidence to date does not support intracellular effects due to drug resistant strains of HIV.

I will go a step further, and say that HIV resistance to antiviral drugs in general, including protease inhibitors is unlikely. How could a virus become so efficient in its selective ability, yet its replicative ability is unchanged and the observed levels of cellular nucleotide pools are not disturbed (39).


A rational examination of the literature shows that there is no standard for defining drug resistance, that the lab experiments use sample sizes too small to be conclusive, and that there are too many unexplained differences between in vivo and in vitro observations. A scientific person must conclude that the phenomenon of drug resistance is a fanciful notion with no basis in reality.

Unfortunately the mutation hypothesis encourages the design and use of more powerful antiviral drugs at higher doses and the use of drug resistance as a prognostic tool (37, 38). Combination therapies are encouraged including old chain-terminators like AZT along with the newest drugs (6, 8, 32). These efforts continue despite the lack of therapeutic benefit. Levels of HIV still persist despite a three-drug combination (4). Drug-resistant strains still evolve despite a two-drug combination of AZT and ddI (7). Some researchers even claim drug resistance is unavoidable (6). All this, despite the lack of correlation between poor clinical outcome and drug resistance (4, 7, 34).

It is not possible for a virus to resist a nucleotide analog. There is no such thing as HIV resistance to AZT or any other nucleotide or nucleoside analog. This statement instantly clears up the issue of any ‘biochemical paradoxes’ baffling so many researchers (21). The use of ddN therapies is not justified.