HIV-1 resistance testing from proviral DNA Alexander Thielen AREVIR - PowerPoint PPT Presentation

HIV-1 resistance testing from proviral DNA Alexander Thielen AREVIR 2018

Resistance testing from proviral DNA? – amount of samples with low viral loads increasing – desire to switch under successful therapy

Zusammensetzung der Viruslasten 2005-2017 (jeweils Q2, Kaiserslautern) 100% 90% – amount of samples with low viral loads increasing 80% 70% – desire to switch under successful therapy 60% >1000 Kop/ml 401-1000 Kop/ml 50% 201-400 Kop/ml – studies, e.g. the LOWER study 40% 50-200 Kop/ml <50 Kop/ml 30% 20% 10% 0% Däumer, M., 2018, unpublished

Resistenzteste 2017, KL 1000 900 800 700 600 500 Plasma RNA: 913 400 Provirale DNA: 138 300 200 100 0 Plasma-RNA provirale DNA provirale DNA 2017 2010 Däumer, M., 2018, unpublished

Resistance testing from proviral DNA? – amount of samples with low viral loads increasing – desire to switch under successful therapy – studies, e.g. the LOWER study

The LOWER study “Limited Options with Extended Resistance to antiretroviral therapy: A National Survey of Triple Class Resistance” (LOWER) – headed by PD Dr. med. Christian Hoffmann – analysis of HIV patients with triple class resistance – resistance testing from proviral DNA with NGS – comparison of current resistance status with historical data

Resistance testing from proviral DNA

Resistance testing from proviral DNA – problem: not only old populations archived but also hypermutated reads – Apobec 3F/3G mutations (G to A) in DNA: • Apobec3F: GA  AA • Apobec3G: GG  AG, further preference for TGG, TGGG motifs! – so, what is really there?

How to detect Apobec mutations – typical mutations: • ATG  ATA M  I • GGY  AGY G  S • GGR  AGR G  R • TGG  TAG/TGA W  * – potential (resistance-associated) amino acid substitutions WT AA MZ AA NRT NNRTI PI INI AI EI Tropism G (Gly) S (Ser) G190S G73S G140S, G163S G357S G36S G11S G (Gly) R (Arg) G163R, G193R G11R, G25R G (Gly) E (Glu) G190E G16E G163E, G193E D (Asp) N (Asn) D67N D30N E (Glu) K (Lys) E138K E138K E25K R (Arg) Q (Gln) R (Arg) K (Lys) M (met) I (Ile) M184I M184I, M230I M36I, M46I M154I – attention: differences in codon usage between subtypes

How to deal with Apobec? 1.The political solution: sit it out – just use the data as it is – problem: M184I and M230I

How to deal with Apobec? 2.The non-believer solution: ignore it – remove all mutations that could occur from Apobec – problem: M184I and M230I sometimes occur in the viral population – really! – M184I usually emerges before M184V which then outcompetes the M184I within several weeks of viral replication

How to deal with Apobec? 2.The non-believer solution: ignore it

How to deal with Apobec? 3.The linkage solution: link & delete – idea: if several mutations possibly induced by Apobec have similar frequencies, then delete them – otherwise stay with them

How to deal with Apobec? 3.The linkage solution: link & delete – idea: if several mutation possibly induced by Apobec have similar frequencies, then delete them – otherwise stay with it mutation frequency M184I 12.4% M230I 14.6% W24* 15.7% W88* 13.3% W153* 13.3% G51R 16.4% ... ...

How to deal with Apobec? 3.The linkage solution: link & delete – idea: if several mutations possibly induced by Apobec have similar frequencies, then delete them – otherwise stay with them – often works very well, but...

How to deal with Apobec? 3.The linkage solution: link & delete

How to deal with Apobec? 3.The linkage solution: link & delete – idea: if several mutations possibly induced by Apobec have similar frequencies, then delete them – otherwise stay with them – often works very well, but... • unfortunately not always

How to deal with Apobec? 3.The linkage solution: link & delete mutation frequency M184I 15.8% V82I 20.0% – idea: if several mutation possibly induced by Apobec have V108I 12.3% similar frequencies, then delete them – otherwise stay with it G190R 14.3% G84R (PR) 19.9% W153* 19.1% – often works very well, but... W71* 14.8% W88* 14.3%

How to deal with Apobec? 3.The linkage solution: link & delete mutation frequency M184I 15.8% V82I 20.0% – idea: if several mutation possibly induced by Apobec have V108I 12.3% similar frequencies, then delete them – otherwise stay with it G190R 14.3% G84R (PR) 19.9% W153* 19.1% – often works very well, but... W71* 14.8% W88* 14.3% how is the K103N affected? will it go below 10% L90M 21.1% L210W 14.3% T215Y 12.4% do we remove them, too? M230L 14.8% ... ...

How to deal with Apobec? 3.The linkage solution: link & delete – idea: if several mutations possibly induced by Apobec have similar frequencies, then delete them – otherwise stay with them – often works very well, but... • unfortunately not always • we do not know how other mutations are affected – how do the frequencies change?

How to deal with Apobec? 4.The lazy solution: filter it automatically – create a filtering method – work on the reads themselves, not on the final result – approach: naïve Bayes classifiers (related to Reuman et al., 2010)

How to deal with Apobec? 4.The lazy solution: filter it automatically – stops: the number of stop codons – atypical aa: the number of atypical amino acid substitutions – burden: #G-to-A / #G – preference: #G-to-A / #substitutions

How to deal with Apobec? 4.The lazy solution: filter it automatically – trained on >100mio reads – tested on 523 samples (111 DNA, 412 RNA) hypermutated 5% 10% 15% 20% 30% reads DNA 19.82% 12.61% 11.71% 9.01% 4.50% RNA 15.05% 5.58% 1.46% 0.73% 0.00% – problems: • too much weight on stop-codons • less power on PR and IN

How to deal with Apobec? 4.The lazy solution: filter it automatically

Acknowledgments Kirsten Becker Elisa Danner Nina Engel Anja Förster Anna Memmer Bettina Welter Martin Däumer Bernhard Thiele