Machine learning enables detection of early-stage colorectal cancer by whole-genome sequencing of plasma cell-free DNA.

BACKGROUND: Blood-based methods using cell-free DNA (cfDNA) are under development as an alternative to existing screening tests. However, early-stage detection of cancer using tumor-derived cfDNA has proven challenging because of the small proportion of cfDNA derived from tumor tissue in early-stage disease. A machine learning approach to discover signatures in cfDNA, potentially reflective of both tumor and non-tumor contributions, may represent a promising direction for the early detection of cancer. METHODS: Whole-genome sequencing was performed on cfDNA extracted from plasma samples (N = 546 colorectal cancer and 271 non-cancer controls). Reads aligning to protein-coding gene bodies were extracted, and read counts were normalized. cfDNA tumor fraction was estimated using IchorCNA. Machine learning models were trained using k-fold cross-validation and confounder-based cross-validations to assess generalization performance. RESULTS: In a colorectal cancer cohort heavily weighted towards early-stage cancer (80% stage I/II), we achieved a mean AUC of 0.92 (95% CI 0.91-0.93) with a mean sensitivity of 85% (95% CI 83-86%) at 85% specificity. Sensitivity generally increased with tumor stage and increasing tumor fraction. Stratification by age, sequencing batch, and institution demonstrated the impact of these confounders and provided a more accurate assessment of generalization performance. CONCLUSIONS: A machine learning approach using cfDNA achieved high sensitivity and specificity in a large, predominantly early-stage, colorectal cancer cohort. The possibility of systematic technical and institution-specific biases warrants similar confounder analyses in other studies. Prospective validation of this machine learning method and evaluation of a multi-analyte approach are underway.

Multiple genetic pathways involving amino acid position 143 of HIV-1 integrase are preferentially associated with specific secondary amino acid substitutions and confer resistance to raltegravir and cross-resistance to elvitegravir.

Y143C,R substitutions in HIV-1 integrase define one of three primary raltegravir (RAL) resistance pathways. Here we describe clinical isolates with alternative substitutions at position 143 (Y143A, Y143G, Y143H, and Y143S [Y143A,G,H,S]) that emerge less frequently, and we compare the genotypic and phenotypic profiles of these viruses to Y143C,R viruses to reconcile the preferential selection of Y143C,R variants during RAL treatment. Integrase amino acid sequences and RAL susceptibility were characterized in 117 patient isolates submitted for drug resistance testing and contained Y143 amino acid changes. The influence of specific Y143 substitutions on RAL susceptibility and their preferential association with particular secondary substitutions were further defined by evaluating the composition of patient virus populations along with a large panel of site-directed mutants. Our observations demonstrate that the RAL resistance profiles of Y143A,G,H,S viruses and their association with specific secondary substitutions are similar to the well-established Y143C profile but distinct from the Y143R profile. Y143R viruses differ from Y143A,C,G,H,S viruses in that Y143R confers a greater reduction in RAL susceptibility as a single substitution, consistent with a lower resistance barrier. Among Y143A,C,G,H,S viruses, the higher prevalence of Y143C viruses is the result of a lower genetic barrier than that of the Y143A,G,S viruses and a lower resistance barrier than that of the Y143H viruses. In addition, Y143A,C,G,H,S viruses require multiple secondary substitutions to develop large reductions in RAL susceptibility. Patient-derived viruses containing Y143 substitutions exhibit cross-resistance to elvitegravir.

Substitutions at amino acid positions 143, 148, and 155 of HIV-1 integrase define distinct genetic barriers to raltegravir resistance in vivo.

Mutations at amino acids 143, 148, and 155 in HIV-1 integrase (IN) define primary resistance pathways in subjects failing raltegravir (RAL)-containing treatments. Although each pathway appears to be genetically distinct, shifts in the predominant resistant virus population have been reported under continued drug pressure. To better understand this dynamic, we characterized the RAL susceptibility of 200 resistant viruses, and we performed sequential clonal analysis for selected cases. Patient viruses containing Y143R, Q148R, or Q148H mutations consistently exhibited larger reductions in RAL susceptibility than patient viruses containing N155H mutations. Sequential analyses of virus populations from three subjects revealed temporal shifts in subpopulations representing N155H, Y143R, or Q148H escape pathways. Evaluation of molecular clones isolated from different time points demonstrated that Y143R and Q148H variants exhibited larger reductions in RAL susceptibility and higher IN-mediated replication capacity (RC) than N155H variants within the same subject. Furthermore, shifts from the N155H pathway to either the Q148R or H pathway or the Y143R pathway were dependent on the amino acid substitution at position 148 and the secondary mutations in Y143R- or Q148R- or H-containing variants and correlated with reductions in RAL susceptibility and restorations in RC. Our observations in patient viruses were confirmed by analyzing site-directed mutations. In summary, viruses that acquire mutations defining the 143 or 148 escape pathways are less susceptible to RAL and exhibit greater RC than viruses containing 155 pathway mutations. These selective pressures result in the displacement of N155H variants by 143 or 148 variants under continued drug exposure.

Differential impact of test performance characteristics on burden-to-benefit tradeoffs for blood-based colorectal cancer screening: A microsimulation analysis.

OBJECTIVES: To inform the development and evaluation of new blood-based colorectal cancer (CRC) screening tests satisfying minimum United States (US) coverage criteria, we estimated the impact of the different test performance characteristics on long-term testing benefits and burdens. METHODS: A novel CRC-Microsimulation of Adenoma Progression and Screening (CRC-MAPS) model was developed, validated, then used to assess different screening tests for CRC. We compared multiple, hypothetical blood-based CRC screening tests satisfying minimum coverage criteria of 74% CRC sensitivity and 90% specificity, to measure how changes in a test's CRC sensitivity, specificity, and adenoma sensitivity (sizes 1-5 mm, 6-9 mm, ≥10 mm) affect total number of colonoscopies (COL), CRC incidence reduction (IR), CRC mortality reduction (MR), and burden-to-benefit ratios (incremental COLs per percentage-point increase in IR or MR). RESULTS: A blood test meeting minimum US coverage criteria for performance characteristics resulted in 1576 lifetime COLs per 1000 individuals, 46.7% IR and 59.2% MR compared to no screening. Tests with increased CRC sensitivity of 99% ( + 25%) vs. increased ≥10 mm adenoma sensitivity of 13.6% ( + 3.6%) both yielded the same MR, 62.7%. Test benefits improved the most with increases in all-size adenoma sensitivity, then size-specific adenoma sensitivities, then specificity and CRC sensitivity, while increases in specificity or ≥10 mm adenoma sensitivity resulted in the most favorable burden-to-benefit tradeoffs (ratios <11.5). CONCLUSIONS: Burden-to-benefit ratios for blood-based CRC screening tests differ by performance characteristic, with the most favorable tradeoffs resulting from improvements in specificity and ≥10 mm adenoma sensitivity.

Connection domain mutations in HIV-1 reverse transcriptase do not impact etravirine susceptibility and virologic responses to etravirine-containing regimens.

Connection domain mutations (CDMs) in HIV-1 reverse transcriptase (RT) alter susceptibility to some nucleoside/nonnucleoside RT inhibitors (NRTIs/NNRTIs). Their effects on susceptibility and virologic responses to etravirine were analyzed. Seventeen CDMs were evaluated: L283I, E312Q, G333D, G333E, G335C, G335D, N348I, A360I, A360T, A360V, V365I, T369I, A371V, A376S, I393L, E399D, and E399G. CDM prevalence and effects on virologic responses were analyzed retrospectively using clinical data. The effects on etravirine susceptibility were assessed in clinical samples and confirmed using site-directed mutants. The most prevalent CDMs (>10%) were A371V, E399D, A376S, N348I, A360T, G333E, and L283I. CDM presence was positively correlated with thymidine analogue-associated mutations, but not with NNRTI resistance-associated mutations (RAMs). The presence or number of CDMs did not significantly reduce etravirine susceptibility, although small reductions were seen in samples with G333D, N348I, A360V, T369I, and A376S. N348I, E399G, and N348I/T369I were associated with reduced etravirine susceptibility when present with K103N, L100I, or Y181C. N348I or T369I was associated with reduced etravirine susceptibility when present with K101P or K103R/V179D. Virologic responses to an etravirine-containing regimen were slightly diminished when G333D, G335D, or A376S was present, but this was not confirmed in subgroups with higher baseline resistance or without etravirine RAMs. CDMs alone do not confer substantial reductions in etravirine susceptibility but can further reduce etravirine susceptibility in combination with certain NNRTI mutations. Since virologic responses to etravirine were not affected by CDMs, the clinical impacts of these mutations on etravirine susceptibility appear to be minimal.

Performance of the Abbott RealTime HIV-1 viral load assay is not impacted by integrase inhibitor resistance-associated mutations.

The Abbott RealTime HIV-1 viral load assay uses primers and probes targeted to integrase, which is also the target of integrase inhibitors such as raltegravir. Viral loads of 42 raltegravir-susceptible and 40 raltegravir-resistant specimens were determined using RealTime HIV-1 and Roche Monitor (v1.5). The differences in viral load measurements between assays were comparable in the two groups, demonstrating that the RealTime HIV-1 assay can tolerate raltegravir-selected mutations.

Combinations of mutations in the connection domain of human immunodeficiency virus type 1 reverse transcriptase: assessing the impact on nucleoside and nonnucleoside reverse transcriptase inhibitor resistance.

Recent reports have described the effect of mutations in the connection and RNase H domains of reverse transcriptase (RT) on nucleoside and nonnucleoside reverse transcriptase inhibitor (NRTI and NNRTI, respectively) resistance in the presence of thymidine analog resistance mutations (TAMs) and NNRTI mutations (J. H. Brehm, D. Koontz, J. D. Meteer, V. Pathak, N. Sluis-Cremer, and J. W. Mellors, J. Virol. 81:7852-7859, 2007; K. A. Delviks-Frankenberry, G. N. Nikolenko, R. Barr, and V. K. Pathak, J. Virol. 81:6837-6845, 2007; G. N. Nikolenko, K. A. Delviks-Frankenberry, S. Palmer, F. Maldarelli, M. J. Fivash, Jr., J. M. Coffin, and V. K. Pathak, Proc. Natl. Acad. Sci. U. S. A. 104:317-322, 2007; G. N. Nikolenko, S. Palmer, F. Maldarelli, J. W. Mellors, J. M. Coffin, and V. K. Pathak, Proc. Natl. Acad. Sci. U. S. A. 102:2093-2098, 2005; and S. H. Yap, C. W. Sheen, J. Fahey, M. Zanin, D. Tyssen, V. D. Lima, B. Wynhoven, M. Kuiper, N. Sluis-Cremer, P. R. Harrigan, and G. Tachedjian, PLoS Med. 4:e335, 2007). In the present study, novel mutations in the connection domain of RT (T369I/V), first identified in patient-derived viruses, were characterized, and their effects on NNRTI and NNRTI susceptibility were determined. Furthermore, the effect of N348I on NRTI and NNRTI resistance was confirmed. HIV-1 with either N348I or T369I/V demonstrated reduced susceptibility to nevirapine (NVP), efavirenz (EFV), delaviridine (DLV), and zidovudine (ZDV) compared to wild-type HIV-1. However, HIV-1 with T369I and N348I demonstrated 10- to 60-fold resistance to these same drugs. In clinical samples, these two connection domain RT mutations were predominantly observed in viruses containing TAMs and NNRTI mutations and did not alter the susceptible-resistant classifications of these samples. Introduction of T369I, N348I, or T369I/N348I also reduced replication capacity (RC). These observations suggest that it may be of scientific interest to test these mutations against new NNRTI candidates.

Loss of raltegravir susceptibility by human immunodeficiency virus type 1 is conferred via multiple nonoverlapping genetic pathways.

The human immunodeficiency virus type 1 (HIV-1) integrase mutations N155H and Q148R(H)(K) that reduce susceptibility to the integrase inhibitor raltegravir have been identified in patients failing treatment regimens containing raltegravir. Whether these resistance mutations occur individually or in combination within a single virus genome has not been defined, nor do we fully understand the impact of these primary mutations and other secondary mutations on raltegravir susceptibility and viral replication capacity. To address these important questions, we investigated the raltegravir susceptibility and replication capacity of viruses containing mutations at positions 155 and 148 separately or in combination with secondary mutations selected in subjects failing treatment regimens containing raltegravir. Clonal analysis demonstrated that N155H and Q148R(H)(K) occur independently, not in combination. Viruses containing a Q148R(H)(K) mutation generally displayed larger reductions in raltegravir susceptibility than viruses with an N155H mutation. Analysis of site-directed mutants indicated that E92Q in combination with N155H resulted in a higher level of resistance to raltegravir than N155H alone. Viruses containing a Q148R(H) mutation together with a G140S mutation were more resistant to raltegravir than viruses containing a Q148R(H) mutation alone; however, viruses containing G140S and Q148K were more susceptible to raltegravir than viruses containing a Q148K mutation alone. Both N155H and Q148R(H)(K) mutations reduced the replication capacity, while the addition of secondary mutations either improved or reduced the replication capacity depending on the primary mutation. This study demonstrates distinct genetic pathways to resistance in subjects failing raltegravir regimens and defines the effects of primary and secondary resistance mutations on raltegravir susceptibility and replication capacity.

Longitudinal analysis of raltegravir susceptibility and integrase replication capacity of human immunodeficiency virus type 1 during virologic failure.

We characterized the raltegravir (RAL) susceptibility and the integrase (IN)-mediated replication capacity (RC) of RAL-resistant human immunodeficiency virus type 1 from two patients experiencing virologic failure during continuous RAL salvage therapy. The following two distinct outcomes were observed: (i) the selective outgrowth of virus with high-level RAL resistance and high IN-mediated RC leading to significant viral load rebound and (ii) the selection of virus with a slight reduction in RAL susceptibility and low IN-mediated RC resulting in sustained low-level viremia.

Coreceptor tropism can be influenced by amino acid substitutions in the gp41 transmembrane subunit of human immunodeficiency virus type 1 envelope protein.

Many studies have demonstrated that the third variable region (V3) of the human immunodeficiency virus type 1 (HIV-1) envelope protein (Env) is a major determinant of coreceptor tropism. Other regions in the surface gp120 subunit of Env can modulate coreceptor tropism in a manner that is not fully understood. In this study, we evaluated the effect of env determinants outside of V3 on coreceptor usage through the analysis of (i) patient-derived env clones that differ in coreceptor tropism, (ii) chimeric env sequences, and (iii) site-directed mutants. The introduction of distinct V3 sequences from CXCR4-using clones into an R5-tropic env backbone conferred the inefficient use of CXCR4 in some but not all cases. Conversely, in many cases, X4- and dual-tropic env backbones containing the V3 sequences of R5-tropic clones retained the ability to use CXCR4, suggesting that sequences outside of the V3 regions of these CXCR4-using clones were responsible for CXCR4 use. The determinants of CXCR4 use in a set of dual-tropic env sequences with V3 sequences identical to those of R5-tropic clones mapped to the gp41 transmembrane (TM) subunit. In one case, a single-amino-acid substitution in the fusion peptide of TM was able to confer CXCR4 use; however, TM substitutions associated with CXCR4 use varied among different env sequences. These results demonstrate that sequences in TM can modulate coreceptor specificity and that env sequences other than that of V3 may facilitate efficient CXCR4-mediated entry. We hypothesize that the latter plays an important role in the transition from CCR5 to CXCR4 coreceptor use.

Suppression of dualtropic human immunodeficiency virus type 1 by the CXCR4 antagonist AMD3100 is associated with efficiency of CXCR4 use and baseline virus composition.

In a phase I/II evaluation of the CXCR4 antagonist AMD3100, human immunodeficiency virus RNA levels were significantly reduced in a single study subject who harbored CXCR4 (X4)-tropic virus, but not in subjects who harbored either dual/mixed (DM)-tropic or CCR5 (R5)-tropic virus (C. W. Hendrix et al., J. Acquir. Immune Defic. Syndr. 37:1253-1262, 2004). In this study, we analyzed the envelope clones of DM-tropic virus in baseline and treated virus populations from 14 subjects. Ten subjects exhibited significant reductions in CXCR4-mediated infectivity after 10 days of AMD3100 therapy relative to baseline (X4 suppressor group), while four subjects had no reduction of CXCR4-mediated infectivity (X4 nonsuppressor group). The baseline viruses of the X4 suppressor group infected CXCR4-expressing cells less efficiently than those of the X4 nonsuppressor group. Clonal analysis indicated that the baseline viruses from the X4 suppressor group contained a higher proportion of R5-tropic variants mixed with CXCR4-using variants, while the X4 nonsuppressor group was enriched for CXCR4-using variants. AMD3100 suppressed X4-tropic variants in all subjects studied, but not all dualtropic variants. Furthermore, dualtropic variants that used CXCR4 efficiently were suppressed by AMD3100, while dualtropic variants that used CXCR4 poorly were not. This study demonstrated that AMD3100 has the ability to suppress both X4-tropic and certain dualtropic variants in vivo. The suppression of CXCR4-using variants by AMD3100 is dependent on both the tropism composition of the virus population and the efficiency of CXCR4 usage of individual variants.

Minimal sequence variability of the region of HIV-1 integrase targeted by the Abbott RealTime HIV-1 viral load assay in clinical specimens with reduced susceptibility to raltegravir.

The Abbott RealTime (ART) HIV-1 assay targets the integrase region and is designed to tolerate mismatches. Variability in integrase sequences comprising the assay target regions from >1000 clinical specimens submitted for phenotypic and genotypic raltegravir resistance testing were analyzed. In this large collection of sequences from clinical specimens, the number and location of raltegravir resistance associated mutations did not differ from those tested previously and shown not to result in under-estimation of viral loads.

Mutational pathways and genetic barriers to CXCR4-mediated entry by human immunodeficiency virus type 1.

To examine mutational pathways that lead to CXCR4 use of HIV-1, we analyzed the genotypic and phenotypic characteristics of envelope sequences from a large panel of patient virus populations and individual clones containing different V3 mutations. Basic amino acid substitutions at position 11 were strong determinants of CXCR4-mediated entry but required multiple compensatory mutations to overcome associated reductions in infectivity. In contrast, basic amino acid substitutions at position 25, or substitutions at positions 6-8 resulting in the loss of a potential N-linked glycosylation site, contributed to CXCR4-mediated entry but required additional substitutions acting cooperatively to confer efficient CXCR4 use. Our assumptions, based upon examination of patient viruses, were largely confirmed by characterizing the coreceptor utilization of five distinct panels of isogenic envelope sequences containing V3 amino acid substitutions introduced by site-directed mutagenesis. These results further define the mutational pathways leading to CXCR4 use and their associated genetic barriers.

Transmission of integrase strand-transfer inhibitor multidrug-resistant HIV-1: case report and response to raltegravir-containing antiretroviral therapy.

We report the case of an integrase strand-transfer inhibitor (INI)-resistant and four-drug-class-resistant HIV-1 variant infecting an antiretroviral therapy-naive man. The virus harboured INI drug resistance substitutions (Q148H and G140S) along with multiple reverse transcriptase and protease inhibitor resistance mutations. This case illustrates an emerging need to consider the possibility of acquired INI resistance among newly diagnosed treatment-naive individuals harbouring multidrug-resistant HIV-1.

Evolution of integrase resistance during failure of integrase inhibitor-based antiretroviral therapy.

BACKGROUND: Although integrase inhibitors are highly effective in the management of drug-resistant HIV, some patients fail to achieve durable viral suppression. The long-term consequences of integrase inhibitor failure have not been well defined. METHODS: We identified 29 individuals who exhibited evidence of incomplete viral suppression on a regimen containing an integrase inhibitor (23 raltegravir, 6 elvitegravir). Before initiating the integrase inhibitor-based regimen, the median CD4 T-cell count and plasma HIV RNA levels were 62 cells/mm and 4.65 log10 copies/mL, respectively. RESULTS: At the first failure time-point, the most common integrase resistance pattern for subjects taking raltegravir was wild-type, followed in order of frequency by Q148H/K/R+G140S, N155H, and Y143R/H/C. The most common resistance pattern for subjects taking elvitegravir was E92Q. Long-term failure was associated with continued viral evolution, emergence of high-level phenotypic resistance, and a decrease in replicative capacity. CONCLUSIONS: Although wild-type failure during early integrase inhibitor failure is common, most patients eventually develop high-level phenotypic drug resistance. This resistance evolution is gradual and associated with declines in replicative capacity.

Characterization of human immunodeficiency virus type 1 populations containing CXCR4-using variants from recently infected individuals.

We screened 150 individuals from two recent seroconverter cohorts and found that six (4%) had CXCR4-using viruses. Clonal analysis of these six individuals, along with a seventh individual identified during clinical care as a recent seroconverter, revealed the presence of both X4- and dual-tropic variants in these recently infected adults. The ability of individual CXCR4-using variants to infect cells expressing CD4/CXCR4 or CD4/CCR5 varied dramatically. These data demonstrate that virus populations in some newly infected individuals can consist of either heterogeneous populations containing both CXCR4-using and CCR5-tropic viruses, or homogeneous populations containing only CXCR4-using viruses. The presence of CXCR4-using viruses at early stages of infection suggests that testing for viral tropism before using CCR5 antagonists may be important even in persons with known recent infection. The presence of CXCR4-using viruses in a subset of newly infected individuals could impact the efficacies of vaccine and microbicide strategies that target CCR5-tropic viruses.

Development and characterization of a novel single-cycle recombinant-virus assay to determine human immunodeficiency virus type 1 coreceptor tropism.

Most human immunodeficiency virus type 1 (HIV-1) strains require either the CXCR4 or CCR5 chemokine receptor to efficiently enter cells. Blocking viral binding to these coreceptors is an attractive therapeutic target. Currently, several coreceptor antagonists are being evaluated in clinical trials that require characterization of coreceptor tropism for enrollment. In this report, we describe the development of an automated and accurate procedure for determining HIV-1 coreceptor tropism (Trofile) and its validation for routine laboratory testing. HIV-1 pseudoviruses are generated using full-length env genes derived from patient virus populations. Coreceptor tropism is determined by measuring the abilities of these pseudovirus populations to efficiently infect CD4+/U87 cells expressing either the CXCR4 or CCR5 coreceptor. Viruses exclusively and efficiently infecting CXCR4+/CD4+/U87 cells are designated X4-tropic. Conversely, viruses exclusively and efficiently infecting CCR5+/CD4+/U87 cells are designated R5-tropic. Viruses capable of infecting both CXCR4+/CD4+/U87 and CCR5+/CD4+/U87 cells are designated dual/mixed-tropic. Assay accuracy and reproducibility were established by evaluating the tropisms of well-characterized viruses and the variability among replicate results from samples tested repeatedly. The viral subtype, hepatitis B virus or hepatitis C virus coinfection, and the plasma viral load did not affect assay performance. Minority subpopulations with alternate tropisms were reliably detected when present at 5 to 10%. The plasma viral load above which samples can be amplified efficiently in the Trofile assay is 1,000 copies per ml of plasma. Trofile has been automated for high-throughput use; it can be used to identify patients most likely to benefit from treatment regimens that include a coreceptor inhibitor and to monitor patients on treatment for the emergence of resistant virus populations that switch coreceptor tropism.

Coreceptor tropism in human immunodeficiency virus type 1 subtype D: high prevalence of CXCR4 tropism and heterogeneous composition of viral populations.

In human immunodeficiency virus type 1 (HIV-1) subtype B, CXCR4 coreceptor use ranges from approximately 20% in early infection to approximately 50% in advanced disease. Coreceptor use by non-subtype B HIV is less well characterized. We studied coreceptor tropism of subtype A and D HIV-1 collected from 68 pregnant, antiretroviral drug-naive Ugandan women (HIVNET 012 trial). None of 33 subtype A or 10 A/D-recombinant viruses used the CXCR4 coreceptor. In contrast, nine (36%) of 25 subtype D viruses used both CXCR4 and CCR5 coreceptors. Clonal analyses of the nine subtype D samples with dual or mixed tropism revealed heterogeneous viral populations comprised of X4-, R5-, and dual-tropic HIV-1 variants. In five of the six samples with dual-tropic strains, V3 loop sequences of dual-tropic clones were identical to those of cocirculating R5-tropic clones, indicating the presence of CXCR4 tropism determinants outside of the V3 loop. These dual-tropic variants with R5-tropic-like V3 loops, which we designated "dual-R," use CCR5 much more efficiently than CXCR4, in contrast to dual-tropic clones with X4-tropic-like V3 loops ("dual-X"). These observations have implications for pathogenesis and treatment of subtype D-infected individuals, for the association between V3 sequence and coreceptor tropism phenotype, and for understanding potential mechanisms of evolution from exclusive CCR5 use to efficient CXCR4 use by subtype D HIV-1.