ABSTRACT
The error-prone replication of human immunodeficiency virus type 1 (HIV-1) enables it to continuously evade host CD8+ T-cell responses. The observed transmission, and potential accumulation, of CD8+ T-cell escape mutations in the population may suggest a gradual adaptation of HIV-1 to immune pressures. Recent reports, however, have highlighted the propensity of some escape mutations to revert upon transmission to a new host in order to restore efficient replication capacity. To more specifically address the role of reversions in early HIV-1 evolution, we examined sequence polymorphisms arising across the HIV-1 genome in seven subjects followed longitudinally 1 year from primary infection. As expected, numerous nonsynonymous mutations were associated with described CD8+ T-cell epitopes, supporting a prominent role for cellular immune responses in driving early HIV-1 evolution. Strikingly, however, a substantial proportion of substitutions (42%) reverted toward the clade B consensus sequence, with nearly one-quarter of them located within defined CD8 epitopes not restricted by the contemporary host's HLA. More importantly, these reversions arose significantly faster than forward mutations, with the most rapidly reverting mutations preferentially arising within structurally conserved residues. These data suggest that many transmitted mutations likely incur a fitness cost that is recovered through retrieval of an optimal, or ancestral, form of the virus. The propensity of mutations to revert may limit the accumulation of immune pressure-driven mutations in the population, thus preserving critical CD8+ T-cell epitopes as vaccine targets, and argue against an unremitting adaptation of HIV-1 to host immune pressures.
Adaptive immune responses are designed to control, and in most cases eradicate, foreign pathogenic organisms. As an indispensable arm of the adaptive immune response, CD8+ cytotoxic T lymphocytes (CTL) have been shown to play a key role in controlling human immunodeficiency virus type 1 (HIV-1) infection. HIV-1-specific CTL responses correlate with the initial control of viremia (11, 40), and in simian immunodeficiency virus (SIV)-infected animals viral loads rise dramatically after antibody-mediated depletion of CTL (35, 62). Moreover, major histocompatibility complex class I molecules mediate the specificity of the CTL response, and the role of particular major histocompatibility complex class I alleles in the control of HIV-1 and SIV infections is also well established (18, 37).
Limiting the effectiveness of CTL responses is the error-prone replicative nature of HIV-1, which generates highly variable viral quasispecies in infected hosts. Viruses with mutations that diminish or abrogate CTL responses have been shown to be preferentially selected for, revealing that the viral quasispecies continuously evolve to evade CTL responses (2, 4, 12, 33, 38, 51, 55). More dramatically, in some cases particular CTL escape mutations have been shown to contribute to the loss of control of viremia and disease progression (4, 12, 33, 38, 50, 55). Recent data from both HIV-1 and SIV now reveal that more than 50% of the non-Env amino acid sequence variations arising over the course of infection are associated with CTL responses, revealing that these responses represent a major driving force of HIV-1 and SIV evolution (2, 51). HLA-associated mutations have also been observed at the population level (6, 41, 48), suggesting the potential for HIV-1 to gradually adapt to host immune pressures (13, 32, 64, 68). It is still unclear, however, if most of the transmitted mutations, especially CTL escape mutations, are stable upon transmission and thus may become fixed in the population (41).
Recently, CTL escape mutations have been observed to revert upon transmission of HIV-1, SIV, and even hepatitis C virus to an HLA-disparate host (3, 27, 42, 56, 63). In these cases, it was presumed that in the absence of the original selective pressure some mutations may no longer provide a tangible benefit to the virus and therefore revert within the new environment (24, 27, 42, 54). The specific impact of some CTL escape mutations on viral replication in both HIV-1 and SIV now provides a clearer understanding of the forces driving some mutations to revert upon transmission (27, 44, 46, 54). Notably, the rate at which transmitted mutations revert may also reveal the relative impact of a mutation. One study of SIV escape mutations in macaques demonstrated that a mutation in Gag quickly reverted to the wild type, likely because of tighter structural constraints in Gag (27). In contrast, mutations in Nef and Tat reverted more slowly or not at all. Similarly, in HIV-1, reversions in both an HLA-A3 (3) and an HLA-B57 (42) epitope have been described, with the relative rate of reversion again corresponding to the conservation of these residues in circulating strains. Therefore, while HIV-1 may demonstrate a propensity to evade host CTL responses through the development of escape mutations, many immune pressure-driven mutations may exact a significant enough cost on viral fitness, which we define here primarily as viral replicative capacity, as to prompt rapid reversion upon transmission to a subsequent host. Supporting this hypothesis is a recent report illustrating that HIV-1 Env commonly evolves via reversions toward an ancestral state shortly after transmission and prior to accumulating new host-specific adaptive mutations (34).
HIV-1 evolution is thus shaped both by selective immune pressures and by purifying pressures attempting to maintain protein structure and function. With the increased feasibility of screening for genome-wide CTL responses (1, 5, 9, 22, 26, 59, 67) and sequencing of HIV-1 (58, 60), numerous studies have undertaken comprehensive analyses to examine the interplay between host CTL responses and the evolution of HIV-1 (2, 10, 16, 21, 30, 31, 36, 39, 53, 65). In particular, two studies investigated HIV-1 sequence variations across the entire genome and revealed a strong influence of CTL-driven immune responses on HIV-1 evolution both within the first month and many years after infection (2, 10). Most studies, however, have focused on selective immune pressures and have not addressed the potential role of reverting mutations and purifying selective pressures in driving HIV-1 evolution. We previously observed in a longitudinal study of four acutely infected subjects that a substantial number of amino acid mutations arising over the course of infection represented reverting substitutions (2). Here we have undertaken a broader analysis to evaluate the impact of these competing selective forces in driving HIV-1 evolution early after infection.
MATERIALS AND METHODS
Subjects.Seven acutely HIV-1 subtype B-infected subjects were enrolled from the Australian HIV Seroconversion Cohort, having been identified during primary infection. HLA typing was performed at the Centre for Clinical Immunology and Biomedical Statistics at the Royal Perth Hospital and Murdock University (Perth, Australia) as shown in Table 1. Peripheral blood mononuclear cell (PBMC) samples were obtained longitudinally approximately every 6 months from all seven patients, with two or three time points available for each subject (Table 1). None of the seven subjects was treated with antiviral therapy during the course of the study. The study was approved by the local Institutional Review Board, and all subjects gave written informed consent.
HLA typing and sampling times of seven subjects
Viral sequencing.Genomic DNA was extracted from PBMC samples (5 million cells each) with the QIAamp DNA Blood Mini Kit (QIAGEN catalog no. 51104). Nested PCR protocols with limiting dilution adapted from references 60 and 61 were used to amplify nearly full-length HIV-1 genomes with EXL DNA polymerase (Stratagene catalog no. 600344). The sequences of the primary forward and reverse PCR primers used, respectively, are 5′-AAATCTCTAGCAGTGGCGCCCGAACAG-3′ and 5′-TGAGGGATCTCTAGTTACCAGAGTC 3′, while the nested forward and reverse primer sequences are 5′-GCGGAGGCTAGAAGGAGAGAGATGG-3′ and 5′-GCACTCAAGGCAAGCTTTA-TTGAGGCTTA-3′. PCR cycling conditions were as follows: 92°C for 2 min; 10 cycles of 10 s at 92°C, 30 s at 60°C, and 10 min at 68°C; 20 cycles of 10 s at 92°C, 30 s at 55°C, and 10 min at 68°C; and a final extension of 10 min at 68°C. Five independent PCR products of each sample were pooled and purified with the QIAquick PCR purification kit (QIAGEN catalog no. 28104) and directly population sequenced at the Massachusetts General Hospital DNA Sequencing Core facility with 70 clade B consensus sequencing primers as previously described (7).
Data analysis.Sequence data were manually edited with Sequencher 4.6 (Gene Codes Corporation). If the secondary peak reached 25% or more of the height of the primary peak at a given position, a mixed residue was called. Sequences were highly concordant with sequences derived from plasma and bulk sequencing approaches (data not shown). Changes from a single residue to a mixed residue and from a mixed residue to a single residue were both considered single-residue substitutions. Nucleotide sequences were conceptually translated and aligned with MacVector 7.2.3 (Accelrys). The clade B HIV consensus sequence (2002) from the Los Alamo National Laboratory (LANL) HIV Sequence Database was used as the reference sequence to compare with our sequencing data. Replacement of a clade B HIV-1 consensus amino acid residue with a nonconsensus residue was considered a forward mutation, while replacement of a nonconsensus residue with a clade B HIV-1 consensus residue was categorized a reversion. Epitope mapping was based on a recent comprehensive collection of reported epitopes by Frahm and Brander (25). Fisher's exact test and nonpaired t test were used in statistical analyses with PRISM 4.0 (GraphPad Software).
Entropy scores of each residue were calculated by normalized Shannon Entropy, distinguishing 21 amino acid symbols based on sequences from the LANL HIV Sequence Database with a web-based algorithm at http://www.ebi.ac.uk/thornton-srv/databases/cgi-bin/valdar/scorecons_server.pl.
RESULTS
The rate of HIV-1 evolution peaks early after infection.We focused our efforts to examine the forces influencing the evolution of HIV-1 within the first year of infection, during the time frame when CTL responses first arise and a viral set point is established (20). To assess the scope and dynamics of HIV-1 sequence variations during this critical phase of the infection, we obtained full-length longitudinal sequence data from seven untreated patients (Table 1). Across the seven HIV-1 genomes, a total of 643 mutations arose at the nucleotide level over the observation period, amounting to a mutation frequency of 0.97% per nucleotide per year. Of these 643 mutations, 366 represented nonsynonymous (NS) substitutions resulting in amino acid changes, equivalent to a mutation frequency rate of 1.66% per amino acid residue per year. The two prior studies focusing on HIV-1 evolution of non-Env regions in different stages of infection (2, 10) enabled a comparison of the kinetics of HIV-1 evolution to the 1-year window examined here. In these studies, Bernardin et al. examined viral evolution in nine subjects during the first month of infection (10) while Allen et al. examined viral evolution in four subjects upward of 5 years after infection (2). In comparing these three studies, as expected, the total number of NS mutations commonly accumulated over time. However, the rate of mutation accumulation in the non-Env region appeared to peak early after infection, as in the study of Bernardin et al. (1.2%/year), and then subside over the course of infection (Fig. 1), suggesting that early events may have a particularly marked impact on HIV-1 evolution. In addition, on the basis of protein lengths and the numbers of mutations harbored in each protein in our study, amino acid substitutions preferentially occurred within the accessory, regulatory, and Env proteins (P < 10−5), with substitution rates (per amino acid residue per year) for Nef of 2.4%, for Tat of 2.1%, for Rev of 1.6%, for Vif of 1.3%, and for Env of 2.9% (Fig. 2). These findings are consistent with our previous study illustrating greater evolution within highly variable proteins (2) and are reflective of the overall sequence diversity of these proteins in the LANL HIV Sequence Database (68). Despite being one of the most conserved proteins, Gag was also heavily targeted, with a substitution rate of 1.2%, perhaps reflective of the frequent targeting of this protein by CTL responses (15, 45, 49, 69).
The rate of mutation accumulation peaks early in HIV-1 infection. The mutation frequencies per amino acid residue per year in non-Env regions were calculated in two recent studies, with Bernardin et al. focusing on the first month after infection (Acute) (10) and Allen et al. focusing on 3 to 5 years after infection (Chronic) (2). The frequencies of NS mutations in non-Env regions in these two studies and our data (Early) are illustrated.
Mutations preferentially arise within nonstructural and Env proteins. NS amino acid mutations were mapped onto individual proteins. After adjustment for length, the relative mutation frequencies at the amino acid level of each protein are shown, with nonstructural and Env proteins represented by solid bars.
CTL responses are influential in driving HIV-1 evolution.CTL immune responses represent a substantial selective force in the evolution of HIV-1 (2). The observation of HLA imprinting across the HIV-1 genome at the population level has provided further support for this (2, 10, 48). In the present study, 6.3% of all mutations and 8% of all non-Env mutations identified within 1 year of infection could be mapped onto previously described CTL epitopes restricted by each subject's HLA alleles (25). Consistent with our data, 9.8% of the total non-Env amino acid mutations arising during acute HIV-1 infection were reported to be associated with described CTL epitopes in a previous study (10). On the basis of the number of amino acid residues within defined CTL epitopes and the total number of residues across the seven HIV-1 genomes, we evaluated if forward mutations preferentially occurred in CTL epitopes. Although a considerable proportion of the forward mutations targeted CTL epitopes in this study (11%), overall forward mutations were not significantly more likely to occur within described CTL epitopes (P = 0.30) than outside, as has previously been observed (2, 10). A number of mutations outside of CTL epitopes have been reported to interfere with CTL epitope presentation (3, 23, 47, 66). By including “neighboring” residues that were located immediately before or after (within one residue) a known CTL epitope restricted by the host's HLA type, 15% of the forward mutations fell into this group and reached statistical significance (P = 0.046). These data may suggest that viral escape through antigen-processing mutations located outside of defined epitopes may be more common than previously appreciated (23, 47, 66).
Reversions preferentially arise within the first 6 months after infection.Our previous study showed that nearly 20% of the mutations arising over 1 to 5 years of HIV-1 infection represented reversions (2). In addition, we had observed in one individual, from whom samples taken at multiple time points had been sequenced, that mutations arising in the first year after infection were dominated by reversions (2). Few other studies have examined the extent to which reversions contribute to the early evolution of HIV-1 (3, 28, 29, 42). Most of the NS mutations observed in this study are those that evolved either from a clade B consensus residue to a nonconsensus residue (forward mutations) or from a nonconsensus residue back to a clade B consensus residue (reversions). Of the 280 such mutations we observed, 118 (42%) represented reversions toward the clade B consensus sequence, a substantially greater percentage than we previously observed. Therefore, by using the four patients within our cohort that were sampled three times at approximately 0, 6, and 12 months (Table 1), we specifically evaluated the dynamics of both forward mutations and reversions. Amino acid substitutions were divided into those appearing within the first 6 months of infection (termed early/fast mutations) and those only arising after 6 months (termed late/slow mutations) (Fig. 3). Strikingly, more than half of the early mutations detected represented reversions (62%), while most of the forward mutations (73%) occurred late in the second half of the year after infection. Overall, reversions occurred significantly earlier than forward mutations (P = 0.000004) (Fig. 4). We also repeated the same analysis individually for each of these four patients. Again, most of the early mutations represented reversions, while most of the late mutations represented forward mutations in all but one patient, PS2019 (Fig. 3). Even for these intrapatient analyses with much smaller sample sizes, three achieved statistical significance (P = 0.01 for PS3002, P = 0.007 for PS2008, and P = 0.02 for PS2016) and the other patient also followed the same trend (P = 0.16 for PS2019). Notably, when we mapped the reversions onto known CTL epitopes, nearly 30% were located within defined epitopes that are not restricted by contemporary hosts, suggesting that prior CTL pressure may have selected the mutations in an HLA-disparate donor. Taken together, the early events of HIV-1 evolution following transmission appear to be dominated by amino acid substitutions reverting toward the consensus sequence, while most of the mutations arising later during infection were predominantly forward mutations.
Distribution of forward and reverting mutations identified across the HIV-1 genome. Locations of identified amino acid mutations were determined in each of the seven longitudinally sampled subjects. NS amino acid mutations were categorized as forward mutations or reversions in reference to the clade B consensus sequence (2002) from the LANL HIV Sequence Database. The numbers of forward mutations and reversions of each patient are shown. For the four patients (PS3002, PS2008, PS2016, and PS2019) with samples from three time points, mutations were stratified according to either a fast mutation if they occurred within the first 6 months after infection or a slow mutation if they occurred during the second half of the year.
Reversions dominate early HIV-1 evolution. The rates at which forward mutations (F) and reversions (R) arose were assessed in subjects PS3002, PS2008, PS2016, and PS2019. The numbers of forward mutations and reversions were stratified by time as shown, with reversions occurring significantly faster than forward mutations (P = 0.000004).
Reversions arise significantly faster in conserved regions.To explore the factors contributing to the interesting observation that a large portion of early mutations are reflective of reversions, the inherent stability of residues at which these reversions occurred was evaluated. Entropy scores for each residue were calculated by using clade B sequences from the LANL HIV Sequence Database. When reversions were stratified according to time (early/fast versus late/slow), the mean entropy score of residues within which fast reversions arose was found to be significantly lower than for residues upon which slow reversions arose (P = 0.0014) (Fig. 5). Thus, reversions arose substantially faster within more conserved residues. The same trend was also observed when the reversions were mapped onto individual proteins (Fig. 6). Gag and Pol, the two most conserved proteins in HIV-1 (68), reflected the highest ratios of fast-to-slow reversions, with fast reversions occurring nearly four times more often than slow reversions in the genes coding for these proteins. Conversely, slow reversions were seen as frequently as, or even more frequently than, fast reversions in Env and nonstructural proteins, such as Vif, Tat, and Nef (Fig. 6). The greater rate of reversion in conserved regions of residues signifies that the structural and functional constraints of viral proteins influence the tolerance for sequence changes at different regions across the genome.
Reversions occur rapidly at conserved residues. The entropy score of each residue was determined by using clade B sequences from the LANL HIV Sequence Database. Fast reversions, those occurring within 0 to 6 months, and slow reversions, those occurring between 7 and 12 months, were identified in subjects PS3002, PS2008, PS2016, and PS2019. Entropy scores of residues where fast and slow reversions arose were then plotted, with fast reversions predominately occurring at conserved residues (P = 0.0014).
Mutations in Gag and Pol revert rapidly. The numbers of fast and slow reversions in each protein were determined, and ratios of fast reversions to slow reversions in different proteins are illustrated.
In our earlier report on CTL response-driven HIV-1 evolution, we also observed the preferential selection for escape mutations to arise within the most variable residue of a CTL epitope during viral escape, suggesting that purifying selection pressures and functional constraints were also influencing viral escape (2). Here, of the 18 forward mutations that were mapped onto host HLA-restricted CTL epitopes, half (50%) again mapped onto one of the two most variable residues within the epitopes (data not shown), with an additional four mutations (22%) arising within poorly conserved positions (the upper quartile of the most variable residues across the genome), again supportive of structural and functional constraints impacting the development of forward CTL escape mutations.
DISCUSSION
In this study, we have characterized the intrahost evolution of HIV-1 during the first year after infection, at a time when many of the key virus-host interactions are taking place. When compared to two recent studies that focused on viral evolution within the first month and 3 to 5 years after HIV-1 infection (2, 10), these data support the notion that a sizable proportion of mutations arise within the first few months after HIV-1 infection. Moreover, this early evolution appears to be significantly driven by reverting amino acid substitutions that preferentially arise within the first 6 months after infection. The observation that rapidly reverting mutations preferentially arise within highly conserved residues is consistent with recent studies documenting the fitness costs associated with particular CTL escape mutations (27, 44, 46, 54) and the propensity of some of these mutations to revert upon transmission (3, 27, 42). Taken together, these data suggest that a substantial degree of early sequence evolution across the entire HIV-1 genome is driven by the reversion of transmitted mutations. These data are in line with a recent publication by Herbeck et al. illustrating that HIV-1 Env also evolves toward an ancestral, or consensus-like, form upon transmission to a new host (34).
Consistent with other studies (2) and the overall sequence variation of HIV-1 clade B sequences in the LANL HIV Sequence Database (68), our data showed that the overall highest mutation frequencies were found in Env and nonstructural proteins (Fig. 2). Structural proteins and functionally conserved regions are expected to poorly accommodate residue changes due to a requirement to preserve protein structure and function. It is thus not surprising that more mutations were observed in Env and nonstructural proteins, which are less constrained and may have a lower cost to viral fitness. The same trend was also seen with a subset of the forward mutations that were located within CTL epitopes. Most of these CTL-associated mutations were mapped onto the most variable residues within the epitopes, suggestive of the influence of viral constraints at the CTL epitope level. HIV-1 evolution is thus confined by structural and functional constraints from the very early stage of infection.
The observation that a considerable portion of the evolving amino acid substitutions arise within the first few months after infection is likely influenced by peak viremia, which is typically observed shortly after infection and prior to the establishment of a viral set point. In addition, during the first few months after infection, many CTL responses first arise and likely exert their strongest selective forces prior to CD4 depletion and immune dysfunction (14, 40, 43, 57). However, the observation that a majority of reversions arise within the first 6 months after infection provides another possible or contributing explanation. This high rate of evolution during the first few months of infection may represent the combined influence of both forward and reverting mutations, while during the later years this rate slows because most of the reversions have already taken place. Therefore, this early phase of HIV evolution may represent a particularly critical stage for the virus to adapt to the new host.
CTL immune responses have been considered a major force in driving HIV evolution, with more than 50% of the sequence variations across the genome having been attributed to CTL pressure in SIV and HIV-1 (2, 51). CTL-driven escape mutations have been well documented in individuals, and HLA-associated sequence polymorphisms have also been reported at the population level (12, 33, 38, 48, 55). Of the 162 forward mutations identified within 1 year of HIV-1 infection, 24 (15%) were mapped onto host-restricted CTL epitopes or flanking residues, where HLA-specific antigen-processing CTL escape mutations have been identified (3, 23, 47, 66). That a significant number of forward mutations were associated with described CTL epitopes (P = 0.046) supports the influential impact of CTL-adapted immune pressure on early HIV-1 evolution. One trend observed was that over time an increasing proportion of mutations were associated with described host-restricted CTL epitopes, with less than 10% seen within the first year but more than 50% seen within 5 years (2). This difference may be partially attributed to the fact that the study of Allen et al. specifically screened many identified mutations for CTL responses (2), which undoubtedly contributed to the considerably higher ratio of CTL-associated mutations. In the present study, we did not specifically undertake a thorough assessment of detectable CTL responses because of a lack of PBMC samples. However, this trend may also be influenced by a decrease in reversions contributing to HIV-1 evolution over the course of infection, thereby leaving immune selection pressures as a dominant driving force during chronic HIV evolution.
Recent studies have started to reveal that CTL escape mutations may exact a cost to viral replicative capacity or fitness (27, 44, 46, 54). Therefore, clearly the dominant quasispecies represents a compromise between evading host immune responses and harboring mutations diminishing replicative capacity in the contemporary host. Upon transmission to a new host expressing different HLA alleles, the CTL escape mutations may lose their benefit and instead carry solely a fitness disadvantage (27). The observation that the rate at which a residue reverts correlates with the overall conservation of that residue is strongly supportive of structural and functional constraints driving the reversion of transmitted mutations. While few studies have systematically begun to address this issue, Friedrich et al. showed that mutations introduced into SIVmac239 Gag, Tat, and Nef reverted in relation to the relative conservation of these proteins (27). Therefore, the rates at which forward mutations and reversions arise may even begin to serve as surrogate markers for the degree of selective pressure applied by particular CTL responses and, conversely, the impact of mutations on viral fitness (24). The significant temporal difference between the occurrence of forward mutations and that of reversions (Fig. 4) strongly suggests that the predominant impact of viral fitness costs is exerted within the first 6 months of infection, while CTL immune pressures are still weak until these responses predominate and have matured a few months into infection (8, 17, 19, 52) and perhaps prior to significant CD4 depletion and immune dysfunction (14, 43). The critical role of viral fitness in early HIV-1 infection was supported by the dominant presence of early/fast reverting mutations that occurred at conserved residues (Fig. 5) and regions such as Gag and Pol (Fig. 6). Similarly, the previous observation by Jones et al. that escaping epitopes were associated with the earliest and strongest CTL responses (36) and the observation that escaping residues preferentially arise at the most polymorphic residues (2) suggest a competitive balance between these two opposing forces. Therefore, both immune selection pressures and viral fitness costs should be considered when selecting targets for HIV vaccines, potentially focusing on regions that induce strong and persistent immune responses, as well as incur a high fitness cost if escape mutations do occur.
In conclusion, reverting mutations play a dominant role in early HIV-1 evolution, supporting the significant role of viral fitness in shaping the landscape of HIV-1 evolution, especially during early infection. The dominance of rapid reversions reveals a substantial impact of viral fitness upon early HIV-1 evolution and may argue against the continued accumulation of immune pressure-driven mutations in the population, which may function to preserve critical CTL epitopes. These data warrant further investigation into the specific impact of particular CTL escape mutations on viral fitness and their correlation with immune control in the design of an effective HIV-1 vaccine.
ACKNOWLEDGMENTS
We thank the subjects who participated in this study. We also thank Francine E. McCutchan and Sodsai Tovanabutra at the Henry M. Jackson Foundation, Rockville, MD, for kind help in viral sequencing and Kate Merlin for assistance with storage and archiving of samples.
This study was supported by National Institutes of Health grants R01-AI054178 and R21-AI067078 (T.M.A.) and the Acute Infection and Early Disease Research Program (AIEDRP) (M.A., J.M.K., D.A.C., A.D.K., and T.M.A.). The National Center in HIV Epidemiology and Clinical Research is supported by the Australian Commonwealth Department of Health and Aging.
FOOTNOTES
REFERENCES
- American Society for Microbiology
