Right away, if you haven't heard already, you should know this: Another once-a-day antiretroviral combo failed.
And now for the really bad news: This one contained efavirenz (EFV).
The past year has endured a cavalcade of less than zesty nucleoside reverse transcriptase inhibitor (NRTI) concatenations, including two or more of the following agents: abacavir (ABC), didanosine (ddI), lamivudine (3TC), and the nucleotide tenofovir disoproxil fumarate (TDF). The most egregious examples -- once-daily 3TC/ABC/TDF and ddI/3TC/TDF -- hit the wall early in four separate trials.1-4 Trizivir (zidovudine [AZT]/3TC/ABC), an easy-to-swallow twice-daily regimen, came up on the short end of the antiviral stick in a contest with Trizivir plus EFV or Combivir (AZT/3TC) plus EFV.5
As this doleful drama unfolded, many questioned the wisdom of pooling three drugs that inhibited the same viral enzyme in the same way. Simply swapping one of the NRTIs for a more potent yet still-kind drug -- once-daily EFV, say -- seemed a sure ticket to durable potency.
And that worked in studies teaming EFV with 3TC/TDF,6 stavudine (d4T)/TDF,6 3TC/ABC,7 and ddI/3TC.8 But it didn't work in a small study of EFV with ddI and TDF [abstract 156]. The failures began within weeks.
Why? There is a one-word answer, a word that explains why HIV virologists, clinicians, number crunchers, and their ilk have gathered for 13 years in a fervid workshop that became a prototype of similar palavers: resistance.
Resistance doesn't explain everything about antiretroviral failure, but it explains a lot. It explains volumes, for example, in Los Angeles, New Orleans, New York, and Seattle, where the US Centers for Disease Control and Prevention (CDC) found that 79 percent of those with at least one genotype on their charts had at least one mutation that lets HIV laugh off antiretrovirals. And it explains plenty in a pilot trial by Daniel Podzamczer (Universitari de Bellvitge, Barcelona) [abstract 156].
With colleagues at other Barcelona centers, Podzamczer set out to compare the merits of a potent first-line regimen with a superpotent regimen in treatment-naive people with high viral loads. But early results abruptly transfigured the trial's focus from standard-versus-intense therapy to unexpected EFV failure. The Barcelona team planned to recruit 50 people with viral loads topping 30,000 copies/mL and randomize them to once-daily EFV, TDF, and ddI (200 or 250 mg depending on weight to counter the boosting effect of TDF) or the same three drugs plus lopinavir/ritonavir (LPV/RTV). They set the primary endpoint as the proportion with a viral load under 50 copies/mL at month 12, but they never got that far.
When the Data Safety and Monitoring Board found no virologic responses by some people after three months, they hastily ordered an unplanned interim analysis involving 26 people who finished three months of therapy. The 17 people originally enrolled in the three-drug arm and the 19 in the four-drug group matched well in proportions with AIDS and baseline CD4 count (median 195 versus 162 cells/mm3) and viral load (median 142,740 and 145,674 copies/mL). But their early response could hardly be more dissimilar. Defining failure as less than a 2-log drop in viral load by month three or a 1-log rebound from the lowest load reached, Podzamczer counted six failures among 14 (43 percent) in the three-drug group and none in 12 from the four-drug group (P = 0.017).
What went wrong? Intracellular concentrations of TDF and ddI metabolites did not differ much between responders and nonresponders in the three-drug group. But triple-drug takers who suffered virologic failure had significantly more advanced disease than those who did not:
Five of the six people in whom triple therapy tanked had the nonnucleoside G190S/E mutation with or without K103N at failure. These mutations popped up as early as 14 days into treatment and as late as three months. All six people had the L74V/I substitution tied to ddI failure, and two had K65R, common to failure of both ddI and TDF.
Apparently people who began the EFV triad with tropospheric viral loads could not bat down replication fast enough to halt evolution of resistant virus. And only one or two point mutations can sink a regimen like EFV, ddI, and TDF.
What's more, attendees of the XIII HIV International Drug Resistance Workshop speculated, subtle mutational tit-for-tats may have worked against EFV and its two nonthymidine teammates. Robert Shafer (Stanford University) and Daniel Kuritzkes (Brigham and Women's Hospital, Boston) observed that virus with thymidine analog mutations (TAMs provoked by AZT or d4T) withers before EFV even more readily than non-mutant (wild-type) virus. So EFV may favor non-TAMs like L74V and K65R. On top of that, pointed out Charles Boucher (Utrecht University, The Netherlands), years ago Hoechst researchers showed that the G190E nonnucleoside mutation prods the emergence of L74V/I nucleoside changes.9 All in all, an unhappy amalgam of selective pressures for people with high viral loads.
For a contrasting view of EFV prowess, look at what's been happening in Gilead's study 903 comparison of 3TC/EFV with d4T or TDF [abstract 157]. Gilead's Michael Miller reported that 73 percent in the TDF arm and 69 percent in the d4T arm have a viral load below 50 copies/mL after 144 weeks in a missing-data-equal-failure analysis.
A key difference between these study participants and Podzamczer's may be the more temperate baseline loads in the 903 group -- averaging below 100,000 copies/mL versus above 200,000 copies/mL in the Barcelona trial. EFV/3TC/TDF apparently squelched replication before many mutations evolved. If the theory holds that a threatened EFV regimen favors non-TAM mutations, inclusion of d4T in the other arm would be a plus. K65R emerged in only 10 people (1.7 percent) in the Gilead study, L74V in only two (0.03 percent), and G190A/E/S/Q in eight (1.3 percent).
EFV/AZT/3TC did just dandy in AIDS Clinical Trials Group (ACTG) protocol 384, perhaps -- once again -- because of the thymidine analog and the more moderate baseline loads (median 4.9 logs).10 But even in people with tuberculosis (TB) and an HIV load averaging 5.75 logs, once-daily ddI (at 400 mg), 3TC, and EFV pushed viral loads below 50 copies/mL in 15 of 17 who completed their anti-TB course.8
One may proffer two precepts from the ddI/TDF/EFV study, one micro-, one macro-, both cosmic. The microcosmic offering hardly needs iteration at this point:
The macrocosmic keepsake may be more important:
No one suggested that Podzamczer should have known that people with lofty loads ran a high risk of recruiting G190S/E, thereby smoothing the route to L74V, which may (as we shall soon see) pave the way to K65R. Just as no one rated GlaxoSmithKline rash for reckoning 3TC/ABC/TDF a once-daily mating made in antiviral heaven. Often only hindsight permits the thesis-driven scrutiny that tests a trial's underpinnings. Still, one must concede that two years of attempts to fashion potent yet complaisant once-a-day regimens has yielded some clunkers.
Why so? The most compelling evidence (indeed, the only evidence so far) points to HIV's evolutionary caprice. Two studies at this year's 5th International Workshop on Clinical Pharmacology of HIV Therapy divined no hints of intracellular antagonism between ABC and TDF.11,12 And in vitro work unveiled at the Resistance Workshop by Cecile Tremblay (Massachusetts General Hospital, Boston) discerned no interactions between or among ABC, TDF, 3TC, and ddI [abstract 41]. Indeed, she found "generally favorable" interactions between these NRTIs. Early looks at drug levels in the culprit studies -- such as Podzamczer's gauging of intracellular ddI and TDF levels -- disclosed no clues of pharmacokinetic hiccoughs.
But resistance mutations cropped up quickly in most people who washed out of triple-nuke trials involving ABC, TDF, 3TC, and/or ddI. The reason can be traced straight back to Charles Darwin and what slick-witted critics called his "law of higgledy-piggledy."13 Today's resistance wonks prefer to call the random processes that drive evolution "stochastic events," but it all boils down to the same thing: Base-pair switching on a given genome happens all the time in an utterly random fashion. But selection of the fittest variant -- whether hominid or host-cell dependent -- is anything but random.
Which brings us to the Canary Islands, (Figure 1) land of sempiternal spring and site of this year's Resistance Workshop. Before the Berbers, then the Arabs, then the Spanish, then the English, then the Spanish again claimed these volcanic peaks for their own, the primal residents, Cro-Magnon descendents called Guanche and Canario, led their simple Neolithic lives.14 As one might expect for a population exposed to year-round semitropical sun, their skin was brown. As one surely would not expect, their hair was blond and their eyes were blue. Stochastic events, for sure.
Figure 1. Tenerife -- a volcanic peak sticking out of the Atlantic -- has hosted the likes of Christopher Columbus, George W. Bush, and most recently attendees at the this year's Resistance Workshop.
Retroviruses such as HIV hold a huge evolutionary edge over Cro-Magnons, for example, because they replicate at warp speed with an AWOL error editor. And because, as John Coffin (Tufts Medical School, Boston) famously figured,15 since all single- and double-mutation errors surface in a single day of furious viral procreation, failure to stop HIV quickly favors selection of those mutants. This is especially bad news for regimens that falter in the face of just a few mutations. Like ddI, TDF, and EFV. And this is the game Darwin's disciples run to ground at the Resistance Workshop, as they strive to render evolution's laws ever less higgledy-piggledy.
Two alluring but flawed triple-nucleoside regimens met with a hasty demise in the past year. Three groups detailed early failures with 3TC, ABC, and TDF over the course of three meetings:
A fourth group studied ddI, 3TC, and TDF:
The four studies all enrolled treatment-naive people who suffered early virologic failures with the 3TC-linked M184V/I mutation, sometimes with K65R, which confers resistance to ddI, ABC, and TDF. With the precise role of K65R in this plot hanging in the balance, a bevy of researchers came to the Resistance Workshop with evidence to weigh, including delegates from ESS30009 2 and Tonus.3
The ESS30009 and Tonus teams reached several of the same conclusions:
Sifting virus from 41 ESS30009 enrollees in whom the three nukes failed by week 12, Lisa Ross (GlaxoSmithKline) found that all but one (98 percent) had the M184V/I change [abstract 159]. Of the 40 people with mutant virus by week 12, 18 (44 percent) had M184V/I alone and 22 (54 percent) had a position 184 change along with K65R. Some study participants had one mutant or the other after a mere two weeks of therapy.
All 40 people with mutations had virus with reduced susceptibility to 3TC, while eight (20 percent) had reduced susceptibility to ABC and only two (5 percent) lost leverage against TDF. Ross suggested that viral phenotype did not reflect virologic response as well as genotype in the early going of this ill-fated trial. Work by GlaxoSmithKline's Randall Lanier found that wild-type/mutant viral mixtures with less than 80 percent mutant virus appeared to be susceptible to TDF on phenotyping [abstract 130]. To the GlaxoSmithKline group, this finding suggested that "the level of TDF and 3TC resistance may be underestimated when K65R and/or M184V are present as mixtures." That underestimation, they argue, may explain why phenotyping of this virus indicates continued susceptibility to TDF.
The bottom line, Ross concluded, is that 3TC/ABC/TDF flunks fast because of a low barrier to resistance, or possibly because of pretreatment resistance. But she reported no pretreatment resistance in these 41 people.
Likewise, Tonus virologist Diane Descamps (Bichat-Claude Bernard Hospital, Paris) saw little hint of pre-treatment resistance in 21 people who suffered virologic failure in that trial [abstract 155]. The only soupçon of something amiss before treatment was a T215E change in one person. By treatment week four she found M184V/I in 10 people (48 percent) and K65K/R (the wild-type/mutant mix) in two (10 percent). One of the two with a position 65 substitution had M184I at week four. By week 12, though, M184V/I had emerged in 18 people (86 percent), and 13 of them (62 percent) had K65R. Clonal analysis showed that at week four the codon 184 and 65 mutations did not share the same viral genome. But Descamps noted that the selective PCR technique she used cannot categorically exclude early linkage of the two mutations.
Although Andy Warhol predicted that media saturation would guarantee all of us 15 minutes of fame, K65R has already racked up at least 15 months in the limelight. Kuritzkes nominated the lysine (K) to arginine (R) switch "mutation of the year" for 2003,16 and the story still has legs in 2004. In that time clinical research clearly showed that even one AZT a day keeps K65R away,17 a finding based on sound though subtle principles,18,19 and one confirmed by a Resistance Workshop study.
This retrospective analysis by Vincent Soriano (Hospital Carlos III, Madrid) involved 260 people who had taken a triple-NRTI regimen [abstract 166]. Defining virologic failure as never reaching a viral load below 50 copies/mL or rebounding from under 50 copies, he counted most of the 77 failures in people who started three nucleosides with a detectable load (45 percent), followed by those who started triple NRTIs as their first combination (23 percent), and those who simplified their regimen to three nukes (17 percent).
People taking an AZT-based regimen had fewer failures (16 percent) than those not taking AZT (42 percent) (P<0.001). Genotyping of 34 people taking a failed regimen turned up AZT-related thymidine analog mutations (TAMs) in no one with K65R. None of nine genotyped people in whom an AZT combination faltered had K65R compared with seven of 25 taking a non-AZT regimen.
Two other studies showed that K65R and TAMs cannot abide each other's company. Soriano's group made this point in a genotypic analysis of 1,846 viral isolates, including 1,630 from people with treatment experience [abstract 40]. K65R adorned 53 isolates (2.9 percent), but the incidence of this mutation rose as TDF entered clinical practice, from 0.6 percent in 1999, to 2.9 percent in 2002, 7.3 percent in 2003, and 11.5 percent in the first trimester of 2004. Ten of the 53 isolates with K65R (19 percent) came from people not taking TDF (four taking ddI/d4T, two 3TC/ABC, two d4T/3TC, one d4T/ABC, and one zalcitabine [ddC]). In his Workshop review of clinical data, Daniel Kuritzkes agreed that d4T may select K65R, as ddI, ddC, ABC, and TDF certainly do. In this respect, d4T appears to differ from its thymidine cousin, AZT.
TAMs showed up in 60 percent of treatment-experienced people without K65R but in only 12 of 53 (23 percent) with K65R (P<0.0001). And the only TAMs caught holding hands with K65R were K70R and K219E. The M184V change surfaced in 24 people (45 percent) with K65R, which rubbed elbows with three non-TAM mutations: Y115F (an ABC-related change) and the Q151M multinucleoside resistance complex, including V75I.
Although M184V and K65R often sat side by side in Soriano's analysis, he could discern only a trend toward their linkage in a univariate analysis (P = 0.1). On the other hand, a study by Andrea Antinori (National Institute for Infectious Diseases Lazzaro Spallanzani, Rome) suggested that M184V may protect against emergence of K65R [abstract 152]. Antinori combed through 1,392 genotypes from 771 people treated from 1999 through 2003. Multivariate analysis indicated that an AIDS diagnosis raised the risk of K65R 4.54 times, use of ABC upped the odds 13.5 times, and EFV raised the risk 14.32 times.
The link with EFV may reflect a tie between the L100I change and K65R, which emerged in a separate multivariate model. The second model also found, once more, that TAMs militate against emergence of K65R, as does M184V. Why would M184V protect against K65R in this analysis while ranking as a frequent fellow traveler in the triple-NRTI failure trials? Perhaps because Antinori's analysis involved people taking a range of regimens, not just risky triple nukes.
A study by Gilead's Michael Miller showed that K65R mutants hovering below the radar of standard sequencing may emerge later to threaten a TDF combo [abstract 158]. This small but enlightening inquest also nominated another variable that may jeopardize TDF therapy -- pretreatment with ddI or ABC, both of which may select the L74V mutation as well as K65R. (TDF does not select L74V.) Finally, Miller's analysis disclosed another example of the genotypic give-and-take that makes resistance both fascinating and frustrating.
The study started with 14 people in whom K65R emerged during the protocol 902 and 907 trials of intensification with TDF. Virus in four of those people harbored the L74V mutation when they started TDF. One of them had too little virus on ice to permit further study, while three had sufficient samples to allow cloning of virus from before TDF began and on treatment weeks four, eight, and 12.
Ultrasensitive single-genome sequencing spotted no K65R mutants in 91 clones of one person's baseline sample. But the other two had tiny traces of K65R in their baseline isolates -- traces that would go unnoticed on standard genotyping. One person was taking ddI and the other ABC when TDF started. Miller showed that, as TDF therapy continued, these K65R tidbits multiplied to mountainous heights, while L74V waned below the detection limit of standard sequencing (Table 1).
The baseline virus, with L74V in the foreground and K65R lurking over the assay's horizon, looked susceptible to TDF on phenotyping. But it wasn't. TDF clearly favored fast evolution of these K65R scraps, which helped defeat the TDF regimen in these two people. By implication, then, earlier treatment with ddI or ABC may imperil the response to TDF if either of those drugs promotes emergence of K65R -- even if standard tests miss that mutant.
Besides the bad news on ddI/TDF/EFV and the evolving K65R saga, the Resistance Workshop offered five other insightful studies that illumined four corners of clinical practice. One group found that transient viremic episodes -- better known as blips -- may betoken significantly lower CD4 counts. Two teams cataloged a relentless recruitment of resistance mutations during structured treatment interruptions (STIs). A meta-analysis of four- versus three-drug trials discerned no apparent reason to pile on the extra drug. And a single case report mapped a possible route to that rarest of evolutionary events -- LPV-provoked mutations.
Although one published study found a higher risk of sustained rebounds in people with viral blips,20 most studies have not.21-25 One of those reports descried no link between blips and dangerous CD4 drops.23 But a Resistance Workshop study did tie the most fleeting of blips to lower CD4 counts in people taking two NRTIs with a nonnucleoside (NNRTI), raising the prospect that one could become too blithe about blips [abstract 63].
Tracking 43 people who kept their viral load below 50 copies/mL for 18 months or more, Anne-Geneviève Marcelin (Pitié-Salpêtrière Hospital, Paris) counted eight (19 percent) who blipped above 50 copies then got below that mark again. Plasma levels of EFV or nevirapine (NVP) proved adequate in both blippers and non-blippers, but a souped-up RNA assay found slightly though significantly higher traces of low-level viremia in the blippers (7.5 versus 3 copies/mL, P = 0.008). People with blips had significantly lower CD4 counts after 12 and 18 months of follow-up (P = 0.04 and 0.02), although none of the blippers met the study definition of virologic failure -- consecutive viral loads above 50 copies/mL.
The study suggests that even occasional loss of viral control can trim long-term CD4 recovery. Since no one in this study had a virologic setback, the CD4 disadvantage in the blip group may have little clinical meaning. But it may mean more in regular blippers or people who start treatment at low CD4 counts.
Two studies -- one in Spain, the other in Italy -- confirmed that stopping drugs for structured treatment interruptions (STIs) can give rise to old or new resistance mutations, especially in people taking NNRTIs or 3TC.
Mireia Arnedo (University of Barcelona) genotyped virus from 35 people coursing through 112 interruption cycles in four protocols [abstract 142]. Mutations materialized in nine people (26 percent) during the STIs. Arnedo documented that half of the mutations had been tucked away after earlier treatment failures while the other half were newly minted. These mutants popped up during 20 of the 112 STI cycles (18 percent).
Among 13 people taking an NNRTI regimen, three (23 percent) had mutations -- all of them brand new. Among 18 people taking 3TC, nine (50 percent) had mutations, four of them (22 percent) first-timers. PI mutations arose in no one, and mutations conferring resistance to NRTIs other than 3TC were rare. The long half-lives of NNRTIs and 3TC -- and their vulnerability to resistance -- make them risky choices for STI trials. It will be interesting to see how TDF and emtricitabine (FTC), with their slow-drip clearance, fare in STI regimens.
Are mutations that arise during STIs dangerous? Yes, according to preliminary results of the ongoing ISS-PART trial, reported by Lucia Palmisano (Istituto Superiore di Sanità, Rome) [abstract 143]. This study randomizes people with viral loads under 400 copies/mL and CD4 tallies above 350 cells/mm3 to continue therapy or to suspend treatment twice for one month, twice for two months, then once for three months.
Mutant virus has already reared up in 28 percent of those in the STI arm, often in people with mutations cloistered in archival corners. Wielding a 3.5-copy RNA assay, Palmisano took a closer look at 49 people in the STI arm, 24 of whom (49 percent) had mutations spotted in plasma virus during an STI. Nine of these 49 people (18 percent) suffered a virologic failure, defined as a rebound above 400 copies/mL. The failure rate proved significantly higher in people with mutations (33 percent) than in those without them (12 percent) (P = 0.004).
The supersensitive RNA assay measured a significantly higher pre-STI load in people with mutations during a treatment break (17.5 copies/mL) than in those without mutations (4.9 copies/mL) (P =0.03). Mutations lurking in proviral DNA correlated with a higher pre-STI RNA load (P = 0.023). Palmisano concluded that RNA simmering below the standard 50-copy limit, as well as mutations hidden in DNA, can predict emergence of mutant virus in plasma during STIs. And mutations during STIs threaten virologic control.
Four antiretrovirals certainly did better than three in the small study of ddI/TDF/EFV with or without LPV/RTV (see the beginning of this article). But scrutiny of nine big randomized trials of four versus three drugs found that adding that fourth agent built little or no extra antiviral muscle [abstract 164]. As the just-discussed studies showed, today's standard triple regimens don't always pin down HIV enough to prevent the odd blip, not to mention barely simmering, if sub-50, replication. Findings like those led Andrew Hill (University of Liverpool) and Graeme Moyle (Chelsea and Westminster Hospital, London) to ask:
Is there an upper limit to antiviral efficacy from three-drug combinations of the current available classes of drugs that cannot be further improved by intensified therapy?
Except when the triple regimen has an unforgivable flaw, they found, the answer is yes.
That answer rests on an exacting appraisal of nine trials that randomized 2,137 (mostly) treatment-naive people to a three- or four-drug medley and had 48- or 52-week viral load results.26-34 Hill and Moyle figured missing-data-equal-failure results at the 50-copy level or, when that wasn't possible, the 400-copy cutoff.
Eight of the nine trials found no antiviral edge for four drugs over three. The one exception, CHARM,31 showed that AZT/3TC/ABC lagged that same trio plus NVP in antiviral oomph, though AZT/3TC/ABC no longer ranks as a first-line regimen. (The analysis did not include preliminary results of the more recent ACTG 5095,5 in which AZT/3TC/ABC trailed two still-blinded regimens, AZT/3TC/ABC/EFV and AZT/3TC/EFV.) Notably, four of the nine studies had antique unboosted PIs as the triple-therapy arm, but that did not detract from the overall triple-therapy results.
Hill and Moyle contended that the equivalent intent-to-treat results with three or four antiretrovirals do not necessarily reflect more toxic dropouts in the four-drug arms. As-treated analyses did not favor tetratherapy either. Still, they observed, "additional benefits from new agents or drug classes [in quad combos] cannot be excluded," so they recommend repeating the analysis as more three-versus-four results pass peer review. (A British study of resistance prevalence did find a lower likelihood of mutant virus among people who started four rather than three drugs. See abstract 135 under "Tales from the UK" below.)
If one calls the NNRTI resistance barrier low, the barricade before LPV/RTV resistance can only be called precipitous. Rebounding virus in people taking LPV/RTV as their first PI has always meant poor adherence or resistance to a weaker link in the regimen, typically 3TC. At the Resistance Workshop, though, Steven Deeks (San Francisco General Hospital) chronicled an unusual PI mutation set in a man taking LPV/RTV. But the reasons resistance overwhelmed this doughty double PI require little sleuthing [abstract 70].
A PI-naive man with undetectable viremia swapped EFV for LPV/RTV in a regimen that also included AZT, 3TC, and TDF. Then he decided to stop the AZT, 3TC, and TDF. He continued using speed, however, and missed at least one LPV/RTV refill over the ensuing months. No surprise, then, that his viral load zoomed to 14,000 copies/mL while sporting the PI-puncturing V82A mutation. But that familiar mutation yielded to an arcane collection of V32I, M46M/I, and I47A.
Phenotyping showed high-level (38-fold) resistance to LPV but undiminished susceptibility to other PIs and hypersusceptibility to saquinavir (SQV) (0.16-fold change compared with wild-type virus). Renewed treatment with SQV/RTV -- and some NRTIs for good measure -- sent the viral load skidding back below 50 copies/mL.
To rate the rarity of these mutations, Deeks snooped through ViroLogic's massive viral vaults. He found a bare 53 isolates with the I47A change, 40 of which also featured V32I. Only 14 isolates had only I47A and V32I. Deeks concluded that although primary resistance to LPV/RTV rivals hens' teeth in rarity, it can happen. And the kinky coupling of I47A and V32I may be one key.
Before proceeding, it may pay to set straight a few facts about the Canary Islands:
Figure 2. Tectonic shifts set off volcanoes -- like Tenerife's Teide. Although Teide did not blow during the Resistance Workshop, attendees palpated some seismic tremblings in the scope of worldwide resistance.
Vulcanism is perhaps a touchy topic in the resistance world, where arguments routinely erupt over rates of mutant virus -- before and during treatment. How widespread is resistance, anyway? Do prevalence estimates rest on reliable models of people exposed to antiretrovirals? Do the populations sampled represent the populations treated? Is resistance in untreated people on the way up, down, or sideways? Can the developed world (which has made a hash of resistant and cross-resistant virus) use resistance to urge circumspection in treating the 90 percent who can't afford resistance assays? Is resistance a good reason to reject NVP for mother-to-child transmission?
These volatile issues -- which seethe more magma as one nears the epidemic's epicenters -- cannot be escaped. And resistance mavens at this year's Workshop could not escape the shadow of Teide -- at 12,198 feet (3,718 meters), Spain's highest peak and the raison d'être for Tenerife. But the meeting did afford an opportunity for some dispassionate dialectics on resistance North and South. These findings emerged:
Susan Little (University of California, San Diego) offered a cogent review of 13 resistance prevalence studies, seven involving untreated people (Table 2) and six treated people (Table 3). Regardless of treatment status, as NNRTIs displace PIs in the affections of physicians, resistance to nonnukes has grown apace. This held true in Canada, the United States, Italy, Spain, the United Kingdom, Japan, and in two massive database surveys probing over 150,000 isolates [abstracts 79 and 80].
Little asked the apposite question: Given the ever-escalating rate of resistance to NNRTIs -- which typically knocks out all three licensed drugs -- and the persistence of these mutations once they arise, should clinicians reappraise use of EFV and NVP as first-line favorites? And she tossed more sparks on the tinder with her own study showing higher viral set points in people infected with NNRTI-resistant virus (see "Transmitted Resistance and Viral Set Point" below).
John Mellors (University of Pittsburgh) called even more widespread resistance to nonnukes "a disaster waiting to happen." The World Health Organization's promotion of single-dose NVP to avert vertical transmission, he said, followed by the group's preference for first-line NNRTI regimens, could touch off a tectonic surge in NNRTI resistance. Giving single-dose NVP to women with high viral loads is among today's most dangerous antiretroviral tactics, added Andrea De Luca (Catholic University, Rome).
Douglas Mayers, top virologist for NVP's marketer, Boehringer Ingelheim, met those arguments with the blunt reality that some struggling countries still lack the means to gear up for single-dose NVP prophylaxis -- never mind triple therapy. He urged flexibility in antiretroviral planning for countries short on means and long in lists of untreated mothers with HIV. When multidrug therapy becomes more feasible in poor countries, suggested Allan Hance (Bichat-Claude Bernard Hospital, Paris), a reasonable approach would be to start treatment with drugs less susceptible to resistance, like PIs, then switch to more friendly nonnukes after the viral load drops.
The ugly threat of nonnucleoside resistance is not just a problem for poor countries, added John Coffin (National Cancer Institute, Frederick, Maryland). The class-killing K103N mutation, for example, can drop to an undetectable 10 percent of the viral population and dig in its heels for a long time. The resulting impression that someone has virus susceptible to NNRTIs could scuttle many a treatment plan. But Robert Grant (Gladstone Institute, San Francisco) counseled moderation in weighing the overall merits of nonnucleosides in developed countries. For most people NNRTIs remain potent, tolerable, and durable -- three traits to look for when planning a first regimen.
People in the United States have been taking antiretrovirals for nearly two decades, and it shows. Three recent surveys found alarming resistance rates among untreated people, treated people, and those who shared beds or needles with them.
Scouting for mutant virus in 10 US cities, the CDC figured that 8.3 percent of 1,082 untreated, HIV-infected people have mutations conferring resistance to reverse transcriptase or protease inhibitors.35 Surveys at 39 clinics from 1997 through 2001 showed that prevalence appeared to track with access to antiretrovirals among the mutant virus donors. Rates proved higher in gay men (11.6 percent) than in women (6.1 percent) or heterosexual men (4.7 percent). With a resistance prevalence of 13 percent, untreated whites easily outdistanced Hispanics (7.9 percent) and African Americans (5.4 percent). Among untreated people with a partner taking antiretrovirals, 15.2 percent had resistance mutations.
"Depending on the characteristics of the patients tested," the CDC concluded, getting a genotype before starting therapy "would identify a substantial number of infected persons with mutations associated with reduced antiretroviral-drug susceptibility."
A Bunyanesque US survey aimed to reckon resistance among (1) 132,500 adults treated since 1996 and with a viral load above 500 copies/mL late in 1998 and (2) 208,900 adults treated at urban clinics since 1996 and surviving into late 1998.36 Nearly two thirds of the target population (63 percent) had a viral load above 500 copies/mL. So this cohort clearly represented people with poor responses to early triple regimens, probably many whose clinicians made the mistake of adding a PI to two floundering NRTIs. Like it or not, that's what happened to a lot of people from 1996 through 1998, and today's US resistance archives reflect the fallout.
Among the 63 percent with detectable viral loads, a whopping (estimated) 76 percent had virus resistant to one or more antiretrovirals. Overall, more than half of those treated in that period had resistant virus. Resistance proved significantly more likely in people with:
At the Resistance Workshop the CDC's Amanda Smith confirmed this high US prevalence in a four-city study, and in one way her results gave even greater reason for dismay [abstract 101]. She eyed trends in 5,939 people treated in Los Angeles, New Orleans, New York, and Seattle from November 2001 through October 2003, so this smaller snapshot frames a later segment of US treatment history than the just-described study. The group included 774 people (13 percent) with at least one genotype on record (a signal that their treatment may have gone awry). The most rueful news comes here: 149 of those 774 (19 percent) got a regimen that did not include a PI, an NNRTI, or AZT/3TC/ABC. In other words, clinicians treating one in five people in these four cities prescribed something other than HAART in 2001-2003.
Looking at genotypes from all 774 people, Smith figured the following mutation rates:
Those numbers probably underestimate actual resistance rates in these four cities because people getting potent regimens were 2.3 times more likely to have a resistance test than people getting substandard therapy. To put it another way, clinicians serving up outmoded regimens in the early 2000s tended not to bother genotyping.
As Susan Little's review showed (Tables 2 and 3), US residents do not lack company on the resistance rolls. Europeans, Australians, and Japanese are doing their bit to feed the profit motive at Virco and ViroLogic. Britain's scept'red isle is no exception.
Studying 4,487 people starting three or more antiretrovirals at six clinics in London and Brighton, Andrew Phillips (Royal Free and University College Medical School, London) charted an ever-ballooning resistance bubble since the survey's start in January 1996 [abstract 135]. Most people began treatment with an NNRTI (56 percent) or a PI (41 percent). Median follow-up covered 183 weeks.
Defining virologic failure as consecutive loads above 1,000 copies/mL after 24 weeks of treatment or one load over that mark and a drug switch, Phillips figured an overall failure rate of 26 percent. Among people treated up to six years, 42 percent fulfilled failure criteria. At the six-year mark, 27 percent had at least one resistance mutation, 18 percent had mutations involving two classes, and 3.5 percent had triple-class resistance. After two, four, and six years of follow-up, resistance prevalence rose in each class:
Resistance proved more likely among people who started treatment with a viral load above 100,000 copies/mL, AIDS, or three versus four drugs. Another finding suggests that NNRTIs' fabled low barrier to resistance may have less clinical import than commonly imputed: People taking a PI without an NNRTI proved 18 percent more likely to come away with resistant virus than people taking an NNRTI without a PI (P<0.005). Perhaps PI tolerability problems inspire enough poor adherence to make PIs a bigger early resistance risk than NNRTIs (see "Poor Adherence and NNRTI Resistance" below).
Phillips observed that these resistance estimates may lowball actual rates because not all cohort members got genotyped at virologic failure and because standard assays miss minority mutants. The steadily spiraling resistance prevalence in this survey led Phillips to the somber conclusion that resistance "is a general phenomenon, and not one restricted to relatively small subgroups of people who might face particular problems with achieving and maintaining adherence."
Resistance rates are notoriously difficult to estimate accurately, as Deenan Pillay (University College of London) demonstrated by reckoning resistance among Britons by two methods [abstract 77].
Both methods rely on the mutation list compiled by the IAS-USA. The analysis involved over 10,000 people who began antiretroviral therapy and were alive after 1998. Method A -- one commonly used in resistance research today -- showed a moderate, stable level of triple-class resistance. The novel method B discerned a lower but rising rate (Table 4). The absolute number of tests registering triple-class resistance by method A bounded from 43 in 1998 to 191 in 1999 but then changed little through 2002. Method B counted 118 cases of triple-class resistance in 1998, 377 in 1999, 590 in 2000, 718 in 2001, and 816 in 2002.
Pillay cautioned that Method B probably underestimates the rate of triple-class resistance because testing occurred at only 16 percent of drug switches, when spotting resistant virus would be more likely. Although his analysis supports Phillips's finding that the overall burden of resistance is on the rise in Britain, the proportion of people saddled with multiclass resistance remains low -- at least lower than in the United States, if Amanda Smith's four-city numbers reflect the treated HIV population at large.
On the other hand, newly transmitted resistance seems more bestrewn in Albion than in the United States or in other countries with recent data, according to scrutiny of 2,191 tests by David Dunn (MRC Clinical Trials Unit, London) [abstract 96]. This study also showed that resistance rates based on viral susceptibility testing lag rates based on mutation lists such as the IAS-USA compendium Pillay used.
Estimating drug susceptibility with the Stanford HIVDB algorithm, which calls virus susceptible, intermediate, or resistant, Dunn rated 8.9 percent of tests run on untreated people from 1997 through early 2003 resistant. If he combined the resistant and intermediate strata, the prevalence of reduced susceptibility in untreated people reached 11.9 percent -- still well short of the 18 percent ciphered with the IAS-USA checklist. Dunn traced this overestimate with the IAS index to resistance calls at protease position 33 and reverse transcriptase position 118. (In a separate study, Deenan Pillay showed that transmission of virus bearing the V118I change does not compromise response to a first regimen [abstract 106]).
Dunn's year-by-year accounting of resistant virus in treatment-naive people suggested that lowered susceptibility to NRTIs leveled off after 2000, whereas rates of resistance to PIs and NNRTIs continue to describe linear upward trajectories. Transmitted triple-class resistance looked rare by Dunn's reckoning, amounting to only 1.3 percent of samples. The most common double-class resistance involved PIs and NNRTIs (6.3 percent).
Two CDC teams proposed a novel means of tracking down transmitted resistance, both relying on bargain-basement point-mutation assays. Diane Bennett fashioned this parsimonious approach to solve a problem in CDC surveillance for resistance in newly diagnosed people [abstract 85]: The CDC wants to expand the project to 22 sites, but sites in most states don't do genotyping. So Bennett devised an algorithm based on efficient point-mutation tests, aiming to pick up 90 percent of resistant strains by focusing on a mere eight positions in reverse transcriptase (RT) and two in protease. Only samples that come up positive in the point mutation screen go on to genotyping.
In certain positions the algorithm instructs the point mutation assay to look for a wild-type amino acid (RT 69, 184, and 215 and protease 46), and absence of wild-type calls for full genotyping. In other positions the algorithm wants a specific mutation whose presence triggers a call for genotyping (RT 41, 67, 70, 103, and 108 and protease 90). Testing the algorithm on 1,082 isolates from recently infected people -- all with known genotypes -- Bennett's decision tree picked out 86 of the 90 resistant strains (96 percent). Resistance prevalence by this method measured 7.9 percent, not significantly different from the 8.3 percent figured with full genotyping. Bennett then validated these results in three other data sets. But because this screening strategy would miss some resistant virus, she cautioned, the algorithm can't replace geno-typing in clinical practice.
The CDC's Jeffrey Johnson gave this concept a limited test with a real-time polymerase chain reaction (PCR) point mutation assay trained to track down the hallmark K103N nonnucleoside mutation and the M184V mutation engendered by 3TC [abstract 76]. In tests of viral specimens known to contain no mutations or to carry K103N or M184V, the real-time PCR test proved highly sensitive and specific. Johnson calculated that it sniffs out K103N mutants making up only 0.4 percent of a viral population and M184V mutants making up only 2 percent.
Using the PCR probes to scout for the target mutants in untreated people who had other reverse transcriptase mutations detected by conventional assays, Johnson rousted K103N from four of 165 isolates (2.4 percent) that appeared to be free of that mutation on standard genotyping. The M184V PCR assay betrayed three positive samples (1.7 percent) in 173 isolates rated M184V-free by conventional testing. Johnson proposed that sensitive, high-throughput assays like these can prove useful in screening for resistant HIV.
Every year's Resistance Workshop kicks off with a parade of data on new antiretrovirals, or, to be honest, unlicensed antiretrovirals. Five years have elapsed, for example, since Virco scientists raised workshoppers' eyebrows -- and more than a little hope -- with news that the PI tipranavir mopped up 90 percent of virus resistant to other protease drugs (in the lab).37 And researchers working on capravirine have made the meeting rounds long enough to instruct all on this NNRTI's proper pronunciation (cap-ruh-VYE-reen, not cap-RAV-uh-reen). Merck's Daria Hazuda dutifully appears every year with the latest results of her epic toils to bring an integrase inhibitor to market, but neither she nor Merck's competitors have many candidates to show for their Sisyphean efforts.
Developers of nascent protease and reverse transcriptase inhibitors invariably label their antiviral incunabula "second-generation" drugs, to the growing ire of old virology hands, who all but banned this term from future meetings. And anyone daring to claim they failed to select virus resistant to their fledgling candidate after umpteen rounds of serial passage had better be good at ducking barbs, bomblets, and overripe produce.
Results of phase 3 tipranavir (TPV) trials may appear later this year, but Resistance Workshop attendees had to settle for a phase 2 study of TPV/RTV versus other boosted PIs in people with lots of antiretroviral experience [abstract 147]. Study participants had virus resistant to the first three antiretroviral classes, had taken two or more PI-based regimens (including their current, failing combination), and had three or four mutations at protease positions 33, 82, 84, and 90.
Prestudy sequencing found that 83 percent of enrollees had three mutations at positions 33, 82, 84, and 90, while the rest claimed all four mutations. These veteran antiretroviralists shouldered a median of six PI mutations, six NRTI mutations, and two NNRTI mutations. Baseline phenotyping showed a 4.7-fold change in 50 percent inhibitory concentration (IC50) for tipranavir and 41- to 361-fold changes for other PIs.
Two weeks of treatment lowered viral loads substantially more in the TPV group than in the groups taking other boosted PIs:
At week four, after everyone added TPV/RTV to their drug diet, viral loads began bouncing back up. Mayers attributed this across-the-board failure to difficulties in constructing a good background regimen for people with such deep treatment experience (one in five people had already tried the fusion inhibitor enfuvirtide [ENF]).
But data disclosed at the Pharmacology Workshop earlier this year dealt a somber hint that TPV may mess with the pharmacokinetics of other PIs.38 There Boehringer Ingelheim's Kevin Curry reported that adding TPV sliced LPV's trough by about 45 percent, APV's by about 50 percent, and SQV's by about 80 percent. Total exposure and peaks of the other PIs also fell after TPV entered the mix, perhaps because the CYP3A4-inducing effect of TPV undermines the inhibiting effect of RTV.
Mayers said early parsings of phase 3 results suggest that TPV/RTV works best in deep salvage when it can team with T-20 in people who have not already taken the fusion inhibitor.
A younger rival of TPV, Pfizer/Agouron's AG-001859, also outmuscles some mutants resistant to current PIs, but its clinical moxie will remain unknown for some time since the drug has just entered phase 1 study. Pfizer's Jennifer Hammond did bring some lab dish data to Tenerife, from a study that pitted AG-001859 and current PIs against an army of viral isolates [abstract 13]. Half had at least one key PI mutation, and 82 percent proved more than 10-fold resistant to PIs already on pharmacy shelves.
Isolates with high-level resistance to APV (IC50 100 nM), IDV (IC50 51 nM), LPV (IC50 41 nM), nelfinavir (NFV) (IC50 49 nM), and SQV (IC50 62 nM) largely retained susceptibility to AG-001859 (respective median fold changes in IC50 of 4.8, 3.1, 1.8, 2.6, and 3.2). The Pfizer/Agouron drug selected virus with three mutations and 10-fold resistance relative to wild-type virus after 200 days of serial passage, but Hammond did not divulge those mutations at the Resistance Workshop.
Two new or newish nonnucleosides of note had advocates at the Resistance Workshop, Roche's TMC125 and Pfizer's capravirine.
In serial passage studies that used EFV as a control, TMC125 selected virus with a 10-fold loss in susceptibility after passage 5 whereas EFV picked out 10-fold resistant virus after passage 3, reported Roche's Gabrielle Heilek-Snyder [abstract 16]. The L100I and K103N mutations provoked high-level resistance to EFV but only moderate resistance to TMC125. Two unusual substitutions, V179F/Y and L214F, coupled with the familiar Y181C, yielded high-level resistance to TMC125 and cross-resistance to other NNRTIs. V138K and M230L also figured in lowered susceptibility to the Roche drug.
Capravirine suffered a setback when high doses caused vasculitis in animal studies. But that problem has not plagued humans, and clinical trials are back on track. Hammond offered some long-term results from a phase 2 study that stopped when the vasculitis scare arose [abstract 15]. People enrolled in that study had all started capravirine/NFV (1,400/1,250 mg twice daily) plus two NRTIs after another NNRTI foundered, and most had classic NNRTI mutations. They could continue unblinded treatment with capravirine/NFV if they had a viral load below 400 copies/mL, and 36 did.
Sixteen people (44 percent) kept their viral load under the 400-copy mark for 41 to 51 months, while five stopped because of side effects, six because of virologic failure, and nine for other reasons. The median time to failure in the six people who lost control of HIV measured 19 months (range four to 27 months). Neither capravirine nor NFV provoked new mutations in two of those people, while three picked up NNRTI-linked changes at positions 101, 108, 188, and/or 190. The sixth person had no genotype at virologic failure; phenotyping indicated sustained susceptibility to capravirine but resistance to NFV.
Hammond proposed that capravirine can rein in NNRTI-resistant virus in some people for prolonged periods.
Before many folks focused on entry inhibitors as the next class of antiretrovirals, integrase inhibitors seemed the likely fourth horseman in HIV's apocalypse. But as Kuritzkes observed at the Resistance Workshop, that horseman rode not to battle but into a "zone of chaos" from which it has yet to emerge. It is not for lack of effort.
Research has now fashioned at least four sets of integrase inhibitors -- the diketo acids (already with their own abbreviation, DKAs), napthyridines, styrylquinolines, and carbazoles. There is some good news: Mutations that render virus resistant to some of these candidates do not make it resistant to others. The bad news is that none of the candidates seems set for the rigors of clinical trials.
Merck's Daria Hazuda works with two types of agents that inhibit an integrase step called strand transfer -- the DKAs and the napthyridines [abstract 1]. Her early research pinpointed integrase sites 66, 151, 153, 154, and 155 as resistance hot buttons for DKAs. When she also turned her attention to napthyridines, she uncovered a whole new set of mutation sinkholes -- at residues 74, 121, and 125. Recombinant viruses confected to carry this latter mutation set proved resistant to napthyridines but not DKAs. And virus bearing the DKA mutation ensemble proved largely susceptible to napthyridines. There was one exception. The N155S swap rendered Hazuda's viral constructs 20-fold resistant to DKAs and 12-fold resistant to a Merck napthyridine.
The DKA- and napthyridine-specific mutations lie on different wrinkles in integrase's active site, but they all sit within that site. N155 may confer cross-resistance, Hazuda proposed, because it could be a pivotal strut in the active site's architecture. But mutations that confer resistance to another candidate class, the styrylquinolines, all perch outside the active site [abstract 3]. BioalliancePharma's Arnaud Chéret (GlaxoSmithKline) reported that a C280Y change makes virus five-fold resistant to the alluringly labeled FZ41, a styrylquinoline lead compound. The V165I/V249I substitutions inspire nine-fold resistance to FZ41. Around the time of the Resistance Workshop, BioallancePharma published details of work indicating that virus resistant to DKAs remains susceptible to styrylquinolines.39
Another trying tale of integrase inhibitor development came from Wataru Sugiura (AIDS Research Center, National Institute of Infectious Diseases [NIID], Japan), who winnowed through 12,000 small-molecule compounds before he harvested one, a carbazole derivative, that showed promise in perturbing integrase [abstract 2]. With this compound as a template, he engineered 15 constructs with sundry side chain ribs branching from the carbazole backbone. Among those 15 he found eight that stymied strand transfer.
But these compounds had a wafer-thin therapeutic window. Their IC50s ranged from 0.78 to 5.3 µMü while their 50 percent cytotoxicity concentration spanned 1.9 to 5.04 µM. Hence, the carbazole derivatives do not merit development. Sugiura suggested they may point the way to other agents, but the frustrating outcome of his efforts underlines the difficulty of this endeavor.
Why the heck is it so hard to find a good integrase inhibitor? All drug development is tortuous, Hazuda offered, and that goes double for would-be integrase inhibitors. The enzyme target shares few similarities with those inhibited so far. For example, Arnaud Chéret noted, integrase has no host cell counterpart, whereas reverse transcriptase and protease do.
On top of that, antiretroviral development is a different game in 2004. Ten or 15 years ago, the goal was to find something fast and to worry about toxicity later. (When researchers discussed resistance to AZT at the premier Resistance Workshop, some argued that what resistance there was could be overcome simply by giving more drug.) Times have changed, Hazuda noted. New drugs must now boast tolerability and ease in taking. But surely even a hard-to-take integrase inhibitor would find customers. People who need ENF learned how to use needles.
The Workshop offered updates on other novel compounds, including Bristol-Myers Squibb's attachment inhibitors [abstract 5] and Pfizer's CCR5 antagonist [abstract 6] (see note 40). But perhaps the most intriguing new antiretroviral is not a drug at all, but a monoclonal antibody that muzzles HIV the same way ENF does, by binding to a stretch of HIV's envelope glycoprotein 41 (gp41) tagged heptad repeat 1 (HR1) [abstract 9].
Searching for human single-chain antibodies that bind to HR1, Merck's Michael Miller found one, designated D5, that curbed HIV entry in a high-throughput entry assay and shut down viral replication in single- and multiple-cycle assays. This antiviral activity did not suffer after he converted D5 into a human IgG1. But these early steps may be the first on a steepening stairway, because Miller learned that the potency of D5 IgG1 varied much in neutralizing an array of HIV isolates. Some isolates shrugged off the monoclonal antibody entirely.
Still, Miller positioned his work as proof-of-concept that human IgG can bind to gp41 HR1 and thereby block HIV entry. He proposed that synthetic antigens can be designed to bear an HR1-derived neutralizing epitope.
Whenever speakers toss off terms like "probabilistic model" and "fuzzy operators," or abbreviations like "ddNp4ddN" and "ddTp4ddT," you know you've wandered into the deep end of the high-science pool. That can happen with riptide precipitance at the Resistance Workshop, where most speakers only reluctantly insult their audience by explaining, for example, that fuzzy operators are "data-mining approaches suitable to find and represent complex associations in a transparent way" [abstract 107] -- as we well know -- not apologists for next year's federal budget or hung-over surgeons.
As fearsome as this terminology sounds, though, it repays study, because sometimes the most gritty data yield a pearl of lambent clinical merit. For example, it takes some inkling of ribosomal frameshifts and transframe proteins to appreciate just how HIV can become resistant to a PI without conjuring up protease mutations. That's what Noortje van Maarseveen (University Medical Center, Utrecht) found when she started working with Ro-033-4649, an experimental PI with an apparently high genetic barrier to resistance. Despite that admirable trait, van Maarseveen conjectured, "one can hypothesize that HIV, like other microorganisms, could exploit alternative resistance mechanisms" [abstract 36]. And so it does.
In vitro work selected virus six to eight times less susceptible to Ro-033-4649 than to wild-type virus. But the selected species had no mutations in the gene that codes for protease. A closer look at viral scaffolding showed changes at gag positions 436 and 437. When van Maarseveen reran the experiment, she found gag mutations at the same spots. A site-directed mutant with the 436 and 437 substitutions proved two- to three-fold resistant to Ro-033-4649 and two- to five-fold resistant to licensed PIs. But this mutant remained susceptible to AZT and NVP.
These genetic flip-flops rest in the ribosomal frameshift site, where pol gene products (such as protease and reverse transcriptase) get made. The same region figures in the evolution of three gag sites that cleave protease into the smaller chunks that help make new virions virulent. These findings hint that HIV can get gag to help it resist inhibition by protease drugs.
In fact, earlier work found mutations at gag cleavage sites in PI-resistant virus. But these were compensatory mutations, mere handmaidens to the major changes HIV needs to sidestep PIs. Charles Boucher, senior author of the Ro-033-4649 study, proposed that the gag mutations van Maarseveen found may explain some current PI failures in people with good adherence but without protease mutations. A search for these changes in such people could confirm or challenge this theory.
Five years ago Peter Meyer (University of Miami) and Michael Parniak (now at the University of Pittsburgh) showed up at the Resistance Workshop with independent studies defining the mechanism of resistance to AZT. Meyer and Parniak were back this year with even better news -- a little biochemical tinkering may be enough to unhinge these viral high jinks.
Meyer, Parniak, and others showed that resistance to AZT results when the analog plugged into the growing viral DNA chain gets stripped back out, an extirpation that lets that viral chain keep adding links. To stop that stripping, Parniak focused on a bisphosphonate numbered BPH-218, which hindered excision of AZT monophosphate at an IC50 of 2 µM [abstract 26]. But BPH-218 didn't thwart normal DNA synthesis -- a signal that it selectively blocks the excision trigger.
By itself, BPH-218 weakly stifled replication of thymidine analog mutant (TAM) virus resistant to AZT. But when Parniak unleashed both BPH-218 and AZT against TAMs, the bisphosphonate significantly cranked up AZT's anti-TAM activity. AZT slows TAM replication at a 50 percent effective concentration (EC50) of 3.5 µM, compared with a keyed-up 0.3 µM versus nonmutant (wild-type) virus. But bolstered by BPH-218, AZT's anti-TAM EC50 improved to 0.01 µM. BPH-218 apparently rendered AZT even more potent against TAMs than it is against wild-type virus, a finding implying that uprooting AZT from budding DNA also helps some wild-type virus duck AZT's sickle. That hypothesis would explain why AZT lacks the intrinsic antiviral activity one would expect.
Can research fashion a clinically feasible bisphosphonate? Parniak predicted that prosaic "medicinal chemistry approaches" can get the job done. Whether such a drug helps people harness HIV is another question.
Since AZT excision is a chemical reaction, Peter Meyer posited, a little chemistry should reverse it. But he took a tack different from Parniak's, testing the ability of dinucleoside tetraphosphates containing dideoxynucleosides -- ddNp4ddNs -- to inhibit DNA synthesis by wild-type or AZT-resistant HIV reverse transcriptase [abstract 31].
Exposing virus to either dideoxynucleotide triphosphate (ddTTP) or ddTp4ddT, Meyer found that DNA polymerization led to concentration-dependent drops in incorporated dinucleoside triphosphate (dNTPs) because incorporation of the inhibitor into the DNA broke the growing DNA chain. Wild-type reverse transcriptase proved strikingly less sensitive to inhibition by ddTp4ddT than by ddTTP (IC50 about 10 µM versus about 140 nM). But reverse transcriptase loaded with the AZT mutations 67N, 70R, 215Y, and 219Q was only 5-fold less sensitive to inhibition by ddTp4ddT than by ddTTP (IC50 about 0.9 µM versus about 0.2 µM). And reverse transcriptase groaning with the mutations 41L, 69S-AG, 210W, 211K, 214F, and 215Y proved equally sensitive to inhibition by ddTp4ddT and ddTTP (IC50 s about 0.35 µM and 0.32 µM).
What do these numbers mean? Meyer deduced that dinucleoside polyphosphates act as substrates (targets) for DNA polymerization by HIV reverse transcriptase and insinuate themselves into reverse transcriptase tricked out with AZT resistance mutations much more readily than into plain-Jane wild-type reverse transcriptase. So dinucleoside polyphosphates must have what it takes to prevent or reverse evolution of mutations resistant to AZT. Next, Meyer will fiddle with phosphate, base, and sugar groups of dinucleoside polyphosphates in hopes of boosting their inhibition of HIV reverse transcriptase, improving their uptake by cells, and heightening their stability.
If Parniak and Meyer are right -- even if one of them is right -- the fruition of this work would yield a sweet harvest from yeoman research first seeded years ago. But it may not weed out the tangles of AZT resistance entirely, if rhizomes unearthed by other plowmen prove productive. Vinay Pathak (National Cancer Institute, Frederick, Maryland) and colleagues at the University of Pittsburgh suggested a novel way in which nucleosides may choke off HIV replication, and it implies a whole new resistance root [abstract 20]. Meyer and Parniak traced out three steps to AZT resistance:
Pathak proposed a fourth:
If RNase H erodes the RNA template before DNA synthesis resumes, reverse transcription -- and viral replication -- would stop. By this reasoning, mutations that slow RNA erosion would promote resistance to drugs like AZT by prolonging the season in which nucleosides can be harvested -- or excised -- from DNA.
Pathak tested this thesis by dousing three viral constructs with AZT, d4T, ddI, 3TC, or EFV: virus with wild-type reverse transcriptase, TAM-trammeled virus, and virus with two RNase H mutations -- H539N and D549N. Compared with wild-type virus, D549N boosted resistance to AZT 10-fold and to d4T 2.6-fold, rates similar to those engendered by TAM-laden virus. H539N did more, catapulting resistance to AZT 180-fold and to d4T 10-fold, levels much higher than those seen with TAMs. The RNase H mutations modestly boosted resistance to ddI and had no effect on 3TC or EFV.
Logic suggests, then, that trimming RNase H activity can pump up resistance to thymidine analogs with no help from TAMs. Another bit of evidence supporting this theory came in virus Pathak cloned from a person with nucleoside-resistant virus: Along with TAMS, it bore the D549N change.
Eradication of HIV resists complete erosion from the HIV research agenda, and for good reason. Although work so far shows that evicting the retrovirus will take more than several years of pious antiretroviral adherence, there are ways to pry latent HIV from its hideaways. Even if such tactics can't cleanse the body of HIV, whisking it from archival corners could make treatment more durable and perhaps abet on-and-off therapy.
But a big hurdle blocks easy extraction of HIV from snoozing CD4 cells: Agents that roust HIV from resting cells typically rouse the cells themselves. And those freshly roused cells offer a freshet of new targets for HIV, a process that quickens viral replication. That conundrum led David Margolis (University of Texas Southwestern Medical School, Dallas) to fish for compounds that prompt cells to spill HIV without waking the cells themselves [abstract 68]. He landed two, including one already found on many a formulary.
The prime candidate is valproic acid, marketed as the antiepileptic Depakene. Among other activities, valproic acid inhibits histone deacetylase, which in turn inhibits HIV gene expression and virus production and may contribute to HIV's knack for napping in sleepy CD4s.
In work presented at the Resistance Workshop and published around the same time,41 Margolis showed that valproic acid can knock the "de-" off histone deacetylase, acetylating HIV into action. But the drug did not rouse slumberous CD4s or expose them to new rounds of HIV infection. Looking in resting CD4s gathered from five antiretroviral-treated people with undetectable viremia, he found that valproic acid induced the cells to oust HIV at doses people can take. In a control experiment he could not recover HIV from resting cells stimulated by interleukin 2. A separate study showed that interleukin 7 encouraged resting CD4s from four aviremic people to spill their HIV.
Valproic acid is no innocuous drug. It can cause liver failure and life-threatening pancreatitis and may promote birth defects if taken by pregnant women. But as an eye-opener for latent virus, it would be taken only in brief courses.
Summing up progress in resistance technology, Richard Harrigan (British Columbia Centre for Excellence in HIV/AIDS, Vancouver) characterized the state of the art in its infancy, circa 1997, this way:
Now more than four years into the new millennium, where does resistance technology stand? Well, Harrigan averred, it's fairly slow and expensive, it misses minority variants, interpretation is complex, and analyses lack cross-validation. But this waggish if wise audit of the state of the art belies genuine technologic progress, two examples of which address conspicuous concerns in Plague Year 2004.
The first, a test that pinpoints resistant virus in dried blood spots, came from Robert Lloyd (Research Think Tank, Alpharetta, Georgia) [abstract 121]. The assay may also simplify viral load monitoring in clinical trials with low-tech sites and tight budgets. Lloyd offered a direct comparison of his blood spot kit, SampleTanker, with standard frozen specimens. Randomly picking from archived iced samples, Lloyd and coworkers prepared duplicate 1-mL frozen and dried specimens for shipping. The SampleTanker spots were air dried and shipped off in individual sealed tubes exposed to ambient temperatures.
Technicians who got the shipments reconstituted the dried 1-mL samples to yield an average 0.998 mL of plasma. They readily measured viral load and determined genotypes of all frozen and dried specimens. Viral loads proved consistently lower with SampleTanker spots, by an average 0.36 log copies/mL, but that would not prove problematic if the difference from standard assays remains consistent. Genotyping resistance-related mutations did not vary from one assay to the other regardless of viral load, storage time, or shipment method.
In the developed world about one third of those with HIV assailing their CD4s also have hepatitis C virus (HCV) hectoring their hepatocytes. Coinfected people respond to interferon and ribavirin worse than those harboring only HCV, and resistance may play a role in that poor response. Michael Kozal (Yale University, New Haven, Connecticut) beat the big assay makers and anti-HCV marketers to the punch with a test that delves for mutants resistant to both viruses [abstract 122].
With the help of Affymetrix, Kozal devised a microarray that scours HCV genotypes 1a and 1b for NS5A and NS5B mutations and combs 1,040 bases of the HIV-1 subtype B pol gene that rules reverse transcriptase and protease. Alternate arrays aimed at specific resistance mutations let Kozal gauge the makeup of mutant and wild-type mixtures.
Probing viral isolates from coinfected people as well as cloned virus, Kozal sequenced 6,102 HIV-1 pol bases and 11,448 NS5A and NS5B bases. The assay's accuracy hit 99 percent in making the right HIV pol base calls and 98 percent for the HCV genes. The test spotted minority variants that made up only 1 percent of the viral population.
Kozal proposed that an assay like this could simplify the study of resistance to interferon and ribavirin even as it decodes the effect of those drugs on HIV genes. He plans to continue fine-tuning the microarray for HCV genotype 1 variants because of their high prevalence in countries where people have access to therapy, and because HCV genotypes 1a and 1b evade antiviral control more readily than other genotypes.
Mutations in the retroviral genome may arise according to Darwin's law of higgledy-piggledy, but they evolve -- as larger species do -- because natural selection favors one genetic palimpsest over another. Science has ratified this rule so thoroughly for HIV-1 that one would expect the phenotypic -- or clinical -- consequences of the law to be clear by now. But they are not.
True, nearly 10 years have passed since ACTG investigators tied AZT resistance to a higher risk of progression or death.42 But does the same tie bind resistance to mortality in the age of triple therapy? Two Resistance Workshop reports disagreed. And what of that most ethically loaded of resistance quandaries -- whether to promote single-dose NVP prophylaxis to prevent mother-to-child transmission? The Resistance Workshop yielded more evidence that a single swallow of this long-lasting drug may not be in the mother's best interest. But the debate over promoting unidose NVP rages still. With NNRTI resistance -- and transmission of resistant mutants -- on the rise in the developed world, what clinical residua can one expect there? A US study held an unhappy clue:
A decade of research shows that viral loads settle to a set point soon after infection in people who don't start antiretrovirals. A high set point, Susan Little reminded colleagues, favors:
So Little asked whether transmission of resistant virus swayed the set point, and how much [abstract 49]. The study involved 340 people with untreated primary HIV infection, 24 (7 percent) with high-level resistance to PIs, 35 (10 percent) with high-level resistance to NRTIs, and 53 (16 percent) with high-level resistance to NNRTIs. Nearly all (94 percent) were men, and 68 percent were non-Hispanic whites.
When Little compared people infected with drug-sensitive (less than 2.5-fold resistant) virus and those with resistance to any drug class, she found no significant difference in viral set point (4.85 logs without resistance and 4.68 logs with resistance, P = 0.187). But the 35 people with NRTI resistance and the 24 with PI resistance had significantly lower set points when compared with the drug-sensitive group (-0.7 log, P<0.001; -0.5 log, P<0.05). And the 53 people infected with NNRTI-resistant virus had a 0.6-log higher set point than the drug-sensitive group (P<0.001).
The lower set points in people with NRTI- or PI-resistant virus support the concept that virus resistant to these drugs has a lower replication capacity than wild-type virus. But the higher set point with NNRTI-resistant virus had nothing to do with replication capacity, Little found, because transmitted NNRTI mutants spewed progeny virions with the vim of drug-susceptible virus.
Although these findings do not address the value of antiretrovirals in people infected with resistant virus, Little cautioned, she proposed that higher set points in people with NNRTI resistance call for close study to determine:
A study in Côte d'Ivoire addressed this contentious third point:
Tracking resistance in 63 women and 26 of their children in Abidjan, Marie-Laure Chaix (CHU Necker, Paris) forged a link between high NVP levels and resistant virus and documented persistence of mutant virus in some of the children [abstract 160]. The women began 300 mg of AZT twice daily at 36 weeks of gestation or later, then took 600 mg of AZT plus 200 mg of NVP just before labor. The neonates got a single dose of NVP on day two or three plus AZT syrup for one week. The transmission rate for all 381 women enrolled in the trial came to 6.4 percent.
Four weeks after delivery, 21 of the 63 women in the resistance substudy (33 percent) had mutant virus -- usually K103N -- in their plasma. None had AZT-related mutations. Chaix looked for mutations in peripheral blood mononuclear cell (PBMC) DNA in 20 of these women and found it in 15. Three women who gave PBMC samples 12 months after delivery had no evidence of cell-bound NNRTI mutations at that point.
Chaix figured a mean NVP half-life of 69 hours in these women. The median NVP concentration measured 598 ng/mL in women without resistant virus and 851 ng/mL in those with resistance (P = 0.014), but NVP levels overlapped considerably in the two groups. A multivariate analysis picked out two factors that independently predicted NVP resistance in mothers:
Among four women who took two NVP doses, three ended up with resistant virus. Work presented earlier this year involving Thai women who took single-dose NVP found a significantly worse six-month virologic response to a regimen containing NVP among those with postpartum resistance.47
In Chaix's study six of 26 infected neonates (23 percent) had NVP-resistant virus at week four. One child whose PBMCs were checked at month three still had resistant virus in DNA, as did one checked at month 12.
A revealing secondary analysis found NVP resistance in 18 of 50 women (36 percent) infected with HIV-1 subtype CRF-02 and in none of eight infected with subtype A (P = 0.0484). This clue that certain subtypes favor emergence of NVP resistance after single-dose prophylaxis held true in a bigger substudy of the HIVNET-012 trial [abstract 50]. Susan Eshleman (Johns Hopkins Medical Institutions, Baltimore) logged mutation trends in 147 Ugandan women infected with HIV-1 subtype A and 98 infected with subtype D.
Seven days after delivery, the two groups didn't differ in mutation rates. But at the six- to eight-week point, 35 women with subtype D (36 percent) and 28 with subtype A (19 percent) had NNRTI mutations in plasma. Those rates translated into a 2.52 times higher risk of resistance in the subtype D-infected women (95 percent confidence interval 1.04 to 8.90).
Eshleman attributed the shifting resistance prevalence from day seven to week eight to two factors:
Given the polychrome global diversity of HIV-1, these developments merit continued scrutiny in NVP-treated women. (For a debate on the implications of NVP prophylaxis, see "Will NNRTI resistance go global?" above.)
An emerging tenet of antiretroviral therapy holds that clinicians should favor easy-to-take, relatively nontoxic NNRTIs for people whose adherence may prove problematic. But results from the REACH cohort of poor people in San Francisco suggest that may not be a great idea [abstract 163]. David Bangsberg (San Francisco General Hospital) discerned resistance trajectories that varied with drug class and adherence resolve.
The study involved 54 people taking a suppressive NNRTI regimen and 54 taking an effective PI combination for more than six months. Bangsberg checked their adherence with unannounced pill-count visits at their usual places of residence. Whenever a viral load climbed above 50 copies/mL, he genotyped a plasma sample. If the viral load stayed below that mark, he assumed drug resistance had not developed.
The NNRTI resistance rate (54 percent) stood much higher than the PI rate (21 percent, P<0.01) in the lowest adherence quartile (0 to 53 percent). As adherence improved, though, the risk of NNRTI resistance dropped while the risk of PI resistance rose. A logistic regression model controlling for treatment duration, prior NRTIs, and baseline CD4 count found a 25 percent lower risk of NNRTI resistance for each 10 percent jump in adherence (P = 0.03) and a 36 percent higher risk of PI resistance with each 10 percent improvement in adherence (P = 0.07).
Bangsberg concluded that NNRTIs may not always be the better choice when a physician fears wobbly adherence, an idea confirmed in a British study (see "Tales from the UK" above).
At the end of the day, the end of a life becomes the most important clinical outcome. Logic suggests that more resistance means fewer antiretroviral options, that fewer options mean rising viral loads and falling CD4s, and that those markers augur an untimely demise. But in the earliest days of resistance research, many experts declined to make that deductive leap. Only when the ACTG tied AZT resistance to early death42 did some concede the point.
That study involved paleozoic AZT monotherapy, whereas more recent work showed stable CD4 counts after virologic failure48 and a lower risk of death with multidrug therapy despite highly resistant virus.49 Did the mortality tables turn again as stronger combos replaced feeble nuke confections? Richard Harrigan addressed the question in British Columbia, as Mauro Zaccarelli (National Institute for Infectious Diseases Lazzaro Spallanzani, Rome) did in Rome. They came up with different answers.
Harrigan, who published his findings around the time of the Resistance Workshop,50 weighed resistance in two cohorts -- 554 people with HIV who had a nonaccidental death from July 1997 through December 2001 and 1,220 treated people with virologic failure but alive at the end of that span [abstract 84].
Among the 554 people who died, the 58 (10.5 percent) who never touched an antiretroviral and the 147 (26.5 percent) with a sub-500 viral load probably had no resistance when they died. Harrigan also grouped 99 briefly treated people (18 percent) in the nonresistant group, but one could argue that their median two months of therapy offered plenty of opportunity for resistance in people with spotty adherence. Death did prove significantly more likely in people who tried one or two NRTIs before starting triple therapy (P<0.0001). Looking for mutations in the last plasma sample of those who died with a detectable load, Harrigan found significantly less single-, double-, and triple-class resistance than in the 1,220 living people with virologic failure (Table 5).
Harrigan concluded that "an exhaustion of treatment options due to drug resistance was not the primary driver of mortality in this cohort." But resistance did take the driver's seat in a 623-person Italian study [abstract 139]. As Harrigan had, Zaccarelli used IAS-USA mutation tables to chart resistance in his cohort, but otherwise this study used utterly different methods. Zaccarelli looked only at treated people, logging deaths from any cause or a new AIDS diagnosis over a median 574 days of follow-up.
Almost half of the cohort, 306 people (49 percent), had no resistance mutations, while 215 (34.5 percent) had resistance to one class, 78 (12.5 percent) to two classes, and 24 (4 percent) to three classes. A median of three regimens (interquartile range one to four) had fizzled in these people.
A Cox proportional hazards model found four independent predictors of progression or death:
After 36 months of follow-up, rates of survival or staying free from AIDS fell as one class after the next succumbed to resistance:
It is interesting that single-, double-, and triple-class resistance rates in the Roman cohort come much closer to those of Harrigan's mortality cohort than his failure cohort (Table 5). But comparing these studies' results makes little sense since they approach the mortality question so differently. Both studies show that people with resistant virus die; Zaccarelli's math ties resistance to death.
Mark Mascolini writes about HIV infection, often as IAPAC Monthly's writer-at-large (email@example.com).