Macroautophagy, hereafter referred to as autophagy, is a key regulated process occurring in cells allowing destruction and recycling of cellular components. It results in the sequestration of cytosol portions into specific vesicles, called autophagosomes, that are delivered into lysosomes for degradation and recycling . Numerous studies relate the impairment of autophagy in many diseases . Some of these reports describe that HIV-1 alters the regulation of autophagy and apoptosis both in uninfected and infected cells, for reviews see [3,4]. Autophagy level increases after HIV-1 entry, expending a large antiretroviral effect in infected CD4+ T cell and resulting in Tat degradation triggering , but the virus strikes back. To avoid its destruction, the virus itself inhibits autophagy  via Vif and Env proteins [7–9]. Actual autophagy during untreated infection, associated with a detectable plasma viral load and high HIV-DNA , is the result of these two opposite forces. Moreover, CD4+ T lymphocytes, whether infected or not, are largely depleted, notably by apoptosis  and autophagy [8,12]. During the latter, expression of envelope glycoproteins on infected cells triggers autophagy in uninfected CD4+ T cells, inducing apoptosis. This process is inhibited in HIV-infected cells [8,12]. But upon efficient cART (combined antiretroviral therapy), plasma viral load becomes undetectable and total HIV DNA diminishes , whereas CD4+ T lymphocyte restoration is evidenced . Little is known about autophagy in long-term cART-treated patients. However, a study shows a significant decrease of autophagy in CD4+ T cells of such individuals when compared with HIV-negative donors  which could be related to chronic immunological defects that remain in treated HIV infection  and/or to therapeutic treatment.
Autophagy, whether at basal level or stress-induced, involves a set of core autophagy-related (ATG) proteins . Among them are ATG6/Beclin-1 encoded by BECN1, which are essential to autophagosome initiation and the ATG8 family with LC3B (light chain 3B), encoded by MAP1LC3B (microtubule-associated protein 1A/1B light chain 3B), GABARAP (GABAA-receptor-associated protein), GABARAPL1 and GABARAPL2 (GABARAP like proteins 1 and 2), which is crucial for autophagosome elongation and maturation. Also important are the ATG9 proteins that are responsible for phagophore expansion by bringing part of membranes to the site of autophagosome formation. In addition, autophagy requires the participation of non-ATG proteins such as p62/SQSTM1 (Sequestosome-1), an adapter that can drive selected substrates to the autophagosomes . Linked to inflammatory responses  and programmed cell death pathways , autophagy also plays a major role in immune responses against pathogens . Evidences of dysfunction of autophagy genes in human diseases has recently emerged , but it should be noted that it is not always known if such inefficiency directly contributes to pathogenesis of human diseases.
The mRNA of the atypical GALIG (GALectin-3 Internal Gene) gene – as it is embedded in another gene termed galectin-3 gene (LGALS3)  – contains two distinct and overlapping open reading frames that translate into mitogaligin and cytogaligin . To date, the mechanism of GALIG's action is not fully understood. Recently, we pointed out via protein-fragment complementation assays  that GALIG's second protein, cytogaligin, can interact, on one hand, with α-synuclein , a protein which compromises both microautophagy and macroautophagy [23,24] and on the other hand, with LC3B/MAP1LC3B, p62/SQSTM1 and GABARAP, three proteins belonging to the macroautophagy pathway (A. Legrand, personal communication). Previously, our team has shown that Mitogaligin interacts with cardiolipin, the particular mitochondrial membrane phospholipid, causing membrane destabilization, aggregation and content leakage . GALIG has been initially described as a cell death gene [26,27] whose role in initiating apoptosis has been shown to be counteracted by overexpression of BCL-XL and MCL1, two antiapoptotic members of the BCL-2 family, whereas expression of BCL2 has no effect [26,28].
In the following pilot study, we investigate if GALIG and seven key ATG gene transcription is altered in cART-treated HIV+ patients when compared with HIV-negative individuals. Standard statistical methods were applied as well as classification machine learning algorithms. Thus, the goal of the analyses presented herein is to compare mRNA from peripheral blood mononuclear cell (PBMC) from efficiently cART-treated patients, that is after immune restoration, with those of uninfected individuals, that is representing the reference level, to determine if transcription alterations of eight selected genes (GALIG and seven ATG) in PBMC from HIV+ patients with efficient therapy can be highlighted.
Methods (including ethical and statistical information)
Study populations, blood samples and RNA isolation
A pilot cohort study was initiated to analyse the expression of seven ATG genes and GALIG, MCL1 and LGALS3 in 27 HIV-positive patients (18 men, nine women) under suppressive cART for at least 4 years. Peripheral blood samples were obtained from patients at the Centre Hospitalier Régional d’Orléans-La Source (CHRO). The institutional ethics committee of the CHRO approved the study protocol (CE 2004.04 – 12/10/2014). All written informed consents followed the ethical guidelines that the 1964 Declaration of Helsinki had set. Main clinical characteristics are given in Table 1.
Forty healthy HIV-negative volunteers (33 men, seven women – age: 55 years old [median] [interquartile range (IQR): 49–61]) were included in the control cohort, after obtaining their written informed consent. Blood was collected by Etablissement Français du Sang Centre Atlantique, the unique French blood transfusion operator. They fulfil the criteria for blood donation (weight: more than 50 kg, age: between 18 and 70 years old, no transfusion, no graft, good health, travel restriction, negative for many viruses, including HIV and hepatitis B and C etc).
PBMC from HIV-positive and HIV-negative individuals were purified after Ficoll-hypaque separation (GE Healthcare Life Sciences, Little Chalfont, Buckinghamshire, UK). Total RNAs were extracted using a NucleoSpin RNA II kit according to the manufacturer's instructions (Macherey-Nagel, Düren, Germany). RNA concentration was measured using a spectrophotometer (NanoDrop One; Thermo Scientific Waltham, Massachusetts, USA).
Real-time reverse transcription-quantitative PCR
Complementary DNA (cDNA) synthesis was performed with 1 μg of RNA and the Maxima 1st Strand cDNA Synthesis Kit (Thermo Scientific) and quantitative PCRs (qPCRs) on 100 ng of cDNA with a Maxima SYBR Green/Fluorescein qPCR kit, according to the manufacturer's instructions (Thermo Scientific). Specific primers for GALIG and MCL1 quantitative PCR were the same as previously listed . LGALS3 transcripts were evaluated with the following primers GGCAGACAATTTTTCGCTC and AGGTTATAAGGCACAATCAGTG. PCRs were performed on a MyiQ Single-Color Real-Time PCR Detection System (Bio-Rad Hercules, Ca, USA) as previously described . All analyses were conducted with Bio-Rad IQ5 2.0 standard edition software. Levels of mRNA were calculated according to the Pfaffl method  using the expression of HPRT1 for sample normalization.
Droplet digital PCR
cDNA synthesis was performed with 500 ng of RNA and the Maxima 1st Strand cDNA Synthesis Kit (Thermo Scientific). The concentration was then adjusted to 5 ng/μl.
The droplet digital PCR (ddPCR) mix consisted of: 10 μl 2 × QX200 ddPCR EvaGreen supermix (Bio-Rad), 100 nmol/l of primers, 0.1–20 ng of the cDNA template into a final volume of 20 μl. GALIG primers were the same as for qRT-PCR . The seven ATG gene primers were as followed: BECN1, CAAGATCCTGGACCGTGTCA and TGGCACTTTCTGTGGACATCA; MAP1LC3B, AGCAGCATCCAACCAAAATC and TTGAGCTGTAAGCGCCTTCT; ATG9a, CTCATCGGGGAGATCTTTGA and GACTTGAGCAGGCAAAAAGG; P62/SQSTM1, AAGAACGTTGGGGAGAGTGTG and GACTCCAAGGCGATCTTCCTC; GABARAP, ATGTCATTCCACCCACCAGT and CAGCAGCTTCACAGACCGTA; GABARAPL1, ATCCGGAAGAGAATCCACCT and TCTTCCTCATGATTGTCCTCA and GABARAPL2, AAATATCCCGACAGGGTTCC and CAGGAAGATCGCCTTTTCAG. The total mix was placed into the eight channel cartridge; 70 μl of droplet generating oil was added and droplets were formed in the QX200 droplet generator (Bio-Rad). Droplets in oil suspensions were transferred to an Eppendorf 96-well plate and placed into a conventional PCR machine. Cycling conditions were as follows (ramp rate of 2 °C/s): 95 °C for 5 min, followed by 40 cycles of 95 °C for 30 s, 55 °C for 30 s and 72 °C for 30 s and a terminal extension cycle 72 °C for 5 min. After an ultimate stabilization step (4 °C for 5 min, and 90 °C for 5 min), the droplets were subsequently read automatically by the QX200 droplet reader (Bio-Rad), and the data were analysed with the QuantaSoft analysis software (Bio-Rad). The copy number of each cDNA was normalized with the amount of RNA used for reverse transcription .
Statistical analyses were performed using StatEL software (www.adscience.eu). We used the nonparametric Mann–Whitney U test (criterion of significance: 5%). The nonparametric measure of statistical dependence between two variables was analysed by the Spearman's link analysis test, using the Spearman's rank correlation coefficient. The results were considered statistically significant at P value less than 0.05.
Learning algorithm system and cross-validation method
Data set encompassed the normalized expression level of eight different genes (GALIG, BECN1, MAP1LC3B, GABARAP, GABARAPL1, GABARAPL2, ATG9a, P62/SQSTM1) in 27 HIV-positive cART-treated patients and 40 HIV-negative donors. We have used the software Scikit-learn, a Python module integrating state-of-the-art machine learning engines and support vector machine (SVM) with linear kernel as classifier. To examine the performance of SVM prediction engines, the five-fold cross-validation method with feature normalization was used and the predictive accuracy of each prediction was calculated as follow. Using the confusion matrix, the accuracy is the ratio of all correct predictions (true positive + true negative) to the total number of evaluated cases (true positive + true negative + false positive + false negative).
Proapoptotic GALIG and antiapoptotic MCL1 mRNA are overexpressed in peripheral blood mononuclear cell of HIV-infected patients on combined antiretroviral therapy
Twenty-seven HIV-infected patients, under suppressive cART for at least 4 years, were enrolled. Their main clinical characteristics are summarized in Table 1. They were diagnosed with HIV infection with a median of 14.5 years ago (IQR: 9–19) and have received cART for a median of 11 years (IQR: 8–18). Four of them were on bi-therapy, 22 on tri-therapy and one on tetra-therapy. Patients reached HIV plasma viral load undetectability 8.4 (median) years ago [IQR: 7.5–10.5]. As reference values of the CHRO biology laboratory are median normal, CD4+ T-cell count 791 cell/μl [IQR: 743–1016] and median CD4+/CD8+ ratio 1.98 [IQR: 1.34–2.63], patients have quasi-normal CD4+ T-cell count [median (IQR): 702 cells/μl (560–875)] and standard CD4+/CD8+ ratio [median (IQR): 1.1 (0.73–1.49)].
To measure GALIG transcripts in uninfected donors and HIV-positive patients, classical quantitative RT-PCR as well as RT-ddPCR, more accurate for low-level transcript quantities, were performed. Differences seen on ddPCR were also confirmed by qPCR. As indicated in Fig. 1a and b, regardless of which method is carried out, GALIG mRNA in cART-treated HIV-positive PBMC are present in significantly higher amounts than in HIV-negative donors (P < 0.000066 and 0.00001, respectively). GALIG expression measured by RT-qPCR scales positively with GALIG expression via RT-ddPCR (P < 0.0068, r = 0.428). These results suggest that, even after at least a median of 8.4 years of suppressive cART (Table 1), the GALIG transcript expression stays elevated and does not come down to the uninfected donor level.
As GALIG is embedded within the LGALS3 gene, we wondered if GALIG and LGALS3 transcription efficiencies are correlated. Figure 1c shows that the LGALS3 mRNA level is similar in cART-treated HIV-positive patients and in HIV-negative controls. These data confirmed that GALIG and LGALS3 are two independent genes with two self-regulated promotors .
Along with the activation of the proapoptotic GALIG gene, there must be some mechanism preventing cell death. Concurrently with GALIG mRNA analyses, we also monitored the expression of MCL1 (Fig. 1d), an important inhibitor of GALIG-mediated cell death . The increase in GALIG transcripts occurs in conjunction with a significant rise of MCL1 mRNA (P < 0.0096).
In brief, PBMC of HIV-infected patients demonstrate a concomitant increase of both the proapototic GALIG and the antiapoptotic MCL1 gene expressions when compared with PBMC from uninfected donors.
Dysregulation of autophagic gene expression in HIV-positive treated patients
We then inquired whether the expression of seven main autophagy genes were disrupted in PBMC of HIV-infected patients under suppressive cART. All mRNA levels were quantified by ddPCR (Table 2). The level of BECN1 mRNA, driving production of Beclin-1, involved in autophagosomes initiation, is significantly underexpressed in HIV-positive patients (P < 0.0058). Two members of the ATG8 family required for autophagosome elongation and maturation are also dysregulated: MAP1LC3B and GABARAPL1 transcripts are present in significantly greater amounts in HIV-positive patients than in uninfected donors with P less than 0.00011 and 0.00001, respectively (Table 2). No differences are observed for ATG9a, p62/SQSTM1, GABARAP and GABARAPL2 between patients and uninfected donors.
Hence, at least three main genes, BECN1, MAP1LC3B and GABARAPL1, involved in basal and infection-induced autophagy are disrupted in HIV-infected patients under suppressive antiretrovirals.
GABARAPL1 and GALIG mRNA levels individually discriminate combined antiretroviral therapy-treated HIV-infected patients from uninfected donors
As classification learning algorithms allow to segregate samples into distinct populations, we applied them to determine if some pattern of tested transcript expression could distinguish cART-treated HIV-infected patients from uninfected donors. To sum up, when analysing mRNA expression levels one-by-one with SVM algorithm (Table 3, line ‘one gene tested’), we show that GAPARAPL1 mRNA expression can classify a sample as HIV-negative donor or HIV-positive patient without detectable viral load, with 91% of certainty. Accuracies of two other engines, logistic regression and random forest give comparable results whichever algorithm was used, with SVM showing a slightly higher accuracy (refer to Table, Supplemental Digital Content 1, https://links.lww.com/QAD/B271). The area under the receiver operating characteristic (ROC) curve (AUC) measures discrimination , which is the ability of expression level to correctly classify those with and without HIV infection. With AUC = 0.977, amongst all the measured mRNA levels, sensitivity and specificity are optimal with GABARAPL1 (refer to Figure, Supplemental Digital Content 2, https://links.lww.com/QAD/B271).
SVM algorithm can segregate these two groups using GALIG and LC3b mRNA level can as well, although less efficiently (respectively, 81 and 73%). On the contrary, the other ATG gene transcripts, and age and sex are close to 50%, that is random level (Table 3). Thus, as a unique classifier out of the eight tested transcript amounts, GABARAPL1 mRNA expression classify HIV positive efficiently treated patients.
Out of the eight tested genes, the combination of GABARAPL1 and ATG9a transcript levels separate combined antiretroviral therapy-treated HIV-infected patients from uninfected donors
Even though the predictive performance for GABARAPL1 transcript level is high in the one-by-one gene tests, a second round of machine learning analyses was run to check if any linear model with more than one transcript expression level could be more efficient. When GABARAPL1 and ATG9a mRNA levels are combined, the predictive performance climbs to 94.5% (Table 3, line ‘GABARAPL1+another gene’ for SVM data and refer to Table, Supplemental Digital Content 1, https://links.lww.com/QAD/B271 for SVM, logistic regression and random forest results). The three gene analyses show no mean accuracy and confidence interval increase (refer to Table, line ‘GABARAPL1+ATG9a+another gene’, Supplemental Digital Content 3, https://links.lww.com/QAD/B271). Confusion matrix, obtained with the five-fold cross-validation method used to calculate SVM accuracies, allows to build ROC curve combining GABARAPL1 and ATG9a mRNA level . Significantly, the AUC with GABARAPL1 alone, 0.977, climbs to 0.990 for GABARAPL1/ATG9a combination (refer to Figure, Supplemental Digital Material 2B, https://links.lww.com/QAD/B271). This result, so close to 1 which represents the perfect test, shows the improvement associated with ATG9a mRNA addition to the model. As expected, AUC of all gene combination falls to 0.977 again, showing no betterment of this whole gene model (refer to Figure, Supplemental Digital Material 2B, https://links.lww.com/QAD/B271).
At first sight, the selection of ATG9a is unexpected: when ATG9a expression levels is tested by classical Mann–Whitney U test, there is no significant difference between the HIV-positive patients and HIV-negative donors (Table 2). However, when analysing effective distribution (Fig. 2a), we noticed that all individuals who have a GABARAPL1 transcript amount less than 300 copies/ng of total RNA are uninfected and all individuals who have more than 1000 copies/ng of total RNA are HIV-positive individuals. But to determine which the best second gene to add is, machine learning algorithms focus on unseparated individuals, here individuals whose GABARAPL1 mRNA measure is between 300 and 1000 copies/ng. Interestingly for individuals who have a GABARAPL1 mRNA quantity in the range 300–1000 copies/ng, ATG9a transcripts in HIV-positive cART-treated patients are present in significantly lower amounts than uninfected donors with P less than 0.0027 (Fig. 2c).
Consequently, the prediction power of the three machine learning algorithms we ran is the highest, 94.5%, when using the two variables, GABARAPL1 and ATG9a mRNA levels. In conclusion, HIV-positive patients, who exhibit autophagy defects , also exhibit abnormal levels of ATG genes transcripts. This GABARAPL1 and ATG9a mRNA separator suggests that the basal level of autophagy in efficiently treated HIV-positive patients can be disrupted.
Discussion (including a conclusion)
Although HIV-infected and efficiently treated patients have quasi-normal CD4+ T-cell count, standard CD4+/CD8+ ratio and no detectable plasma viral load, this ‘quasi’ is sometimes linked to a blunted immune recovery which leads to dysfunctions of their immune system. As prolonged efficient treatments lead to new HIV-associated complications, resulting in chronic inflammatory status , autophagy dysregulation could play a role in the chronic inflammation observed in cART-treated HIV-infected patients . Furthermore, basal autophagy allows selective engulfment of damaged mitochondria . The inhibition of such phenomenon could be linked to accelerated ageing of the immune system evidenced in long-term cART-treated HIV-infected patients. Therefore, in this pilot study, we interrogated the transcript level of GALIG and seven main ATG genes and applied classical statistical and machine learning algorithm analyses to search for any differences between HIV-positive and HIV-negative PBMC.
Herein, we show that PBMCs of cART-treated HIV-infected patients demonstrate some mRNA level impairments. First, we describe a concomitant increase in the expression of both the proapoptotic GALIG gene and the antiapoptotic MCL-1. As already reported in experiments based on culture cell transfections , this MCL1 overexpression could explain why PBMC of cART-treated HIV-infected patients show no evidence of cell death despite their high level of GALIG transcripts. Whether these data are related to the fact that apoptotic features are reduced when HIV-infected patients are efficiently treated, remains to be investigated.
Our nonparametric statistical tests show that the mRNA levels of GABARAP, GABARAPL2, ATG9a and p62/SQSTM1 do not exhibit significant differences, whereas the mRNA levels of BECN1, MAP1LC3B and GABARAPL1 are disrupted in HIV-positive patients without plasma viral load. DNA methylation and histone posttranslational modifications affect the level of chromatin compaction, and therefore control gene expression via the regulation of promoter accessibility to transcriptional factors . It will be interesting to check if modulations of GALIG and ATG gene transcripts in cART-treated patients is associated with such epigenetic modifications. Moreover, even if one could expect that, if protein levels follow mRNA amounts, the increased level of MAP1LC3B and GABARAPL1 mRNA will lead to an abundance of autophagosome and increased basal autophagy, the underexpression of BECN1 mRNA could result in a low level of autophagosome initiation and consequently an inhibition of basal autophagy. But as the disrupted mRNAs include both positive and negative regulators, it is even possible that the basal autophagy steady-state is maintained. In this pilot study, protein production and functional tests had not been not performed and investigating if these discrepancies have a direct impact on the already reported autophagy dysfunctions  in cART-treated HIV-positive patients will be a crucial step to be done.
When analysing the eight genes, we selected one-by-one, the machine learning approach enables us to rank genes by importance to classification. Significantly, GABARAPL1 expression level can classify a sample as HIV-positive or HIV-negative with the best accuracy (91%), followed by GALIG mRNA (81%). Of note AUC is optimal for GABARAPL1 mRNA (0.977) and high for GALIG mRNA (0.910) as well. Then algorithms were used to build a two-gene model, from the eight mRNA-tested measures, namely GABARAPL1 and ATG9a mRNA levels, that could classified patients with or without HIV-infection with 94% overall accuracy and an AUC = 0.990. As AUC does not take negative rates into account (true negative and false negative rates), this slight difference is logical. Thus, even though ATG9a mRNA measure alone is not statistically different between the whole population of HIV-positive and HIV-negative samples, we have shown that the expression levels of GABARAPL1 and ATG9a are surrogate markers to segregate PBMC according to HIV status. Focusing on these markers, functional tests, both on constitutive and induced autophagy will definitely be the next step.
Like all pilot studies, this study has some limitations. It will be ideal to validate these findings in a larger and independent cohort. In the future, we will sort PBMC subsets to determine which types of blood cells display this dysregulated GALIG and autophagy alteration pattern. As thymidine reverse transcriptase inhibitors inhibit autophagy in hepatocytes , the cART composition should be evaluated and their impact on PBMC in vitro will also be monitored. Moreover, in the field of de-escalation strategies, early cART has been shown to have impressive results as far as HIV control, viral reservoirs and immune restoration are concerned [37,38]. In that case of posttreatment controllers, will the GABARAPL1 and ATG9a mRNA levels still be accurate? A pilot study should be built with such patients to answer this question.
We are grateful to uninfected volunteers and patients who have contributed to this study. We thank Dr Valerie Berthelier from 4Publication for critically reading, proofreading this article and providing English editorial assistance. This work was supported by grants (GALIGPATH) from the Conseil Départemental du Loiret (CD45).
Contributions: Assay development: A.S. Performed the experiments: A.S., S.E.H., F.M., C.R. Data analyses: A.S., S.E.H., L.M., F.M. Contribution to experimental analyses: F.B., S.C., T.N., A.L. Discussion of results: all authors. Supervised the study: L.M., L.H. Coordination of patient inclusion: L.H. Help with subject recruitment: T.P. Management of biological patient analyses: E.L. Daft writing: A.S.,L.M. Contribution to reviewing the article and article corrections: all authors.
Conflicts of interest
There are no conflicts of interest.
1. Lamb CA, Yoshimori T, Tooze SA. The autophagosome: origins unknown, biogenesis complex
. Nat Rev Mol Cell Biol
2. Jiang P, Mizushima N. Autophagy and human diseases
. Cell Res
3. Liu Z, Xiao Y, Torresilla C, Rassart É, Barbeau B. Implication of different HIV-1 genes in the modulation of autophagy
2017; 9:398doi:10 3390/v9120389.
4. Cossarizza A. Apoptosis and HIV infection: about molecules and genes
. Curr Pharm Des
5. Sagnier S, Daussy CF, Borel S, Robert-Hebmann V, Faure M, Blanchet FP, et al. Autophagy restricts HIV-1 infection by selectively degrading Tat in CD4+ T lymphocytes
. J Virol
6. Zhou D, Spector SA. Human immunodeficiency virus type-1 infection inhibits autophagy
7. Borel S, Robert-Hebmann V, Alfaisal J, Jain A, Faure M, Espert L, et al. HIV-1 viral infectivity factor interacts with microtubule-associated protein light chain 3 and inhibits autophagy
8. Espert L, Biard-Piechaczyk M. Autophagy in HIV-induced T cell death
. Curr Top Microbiol Immunol
9. Dinkins C, Pilli M, Kehrl JH. Roles of autophagy in HIV infection
. Immunol Cell Biol
10. Avettand-Fènoël V, Hocqueloux L, Ghosn J, Cheret A, Frange P, Melard A, et al. Total HIV-1 DNA, a marker of viral reservoir dynamics with clinical implications
. Clin Microbiol Rev
11. Gougeon ML, Lecoeur H, Dulioust A, Enouf MG, Crouvoiser M, Goujard C, et al. Programmed cell death in peripheral lymphocytes from HIV-infected persons: increased susceptibility to apoptosis of CD4 and CD8 T cells correlates with lymphocyte activation and with disease progression
. J Immunol
12. Denizot M, Varbanov M, Espert L, Robert-Hebmann V, Sagnier S, Garcia E, et al. HIV-1 gp41 fusogenic function triggers autophagy in uninfected cells
13. Autran B, Carcelain G, Li TS, Blanc C, Mathez D, Tubiana R, et al. Positive effects of combined antiretroviral therapy on CD4+ T cell homeostasis and function in advanced HIV disease
14. Gómez-Mora E, Robert-Hebmann V, García E, Massanella M, Clotet B, Cabrera C, et al. Brief report: impaired CD4 T-cell response to autophagy in treated HIV-1-infected individuals
. J Acquir Immune Defic Syndr
15. Johansen T, Lamark T. Selective autophagy mediated by autophagic adapter proteins
16. Netea-Maier RT, Plantinga TS, van de Veerdonk FL, Smit JW, Netea MG. Modulation of inflammation by autophagy: consequences for human disease
17. Maiuri MC, Zalckvar E, Kimchi A, Kroemer G. Self-eating and self-killing: crosstalk between autophagy and apoptosis
. Nat Rev Mol Cell Biol
18. Levine B, Deretic V. Unveiling the roles of autophagy in innate and adaptive immunity
. Nat Rev Immunol
19. Raimond J, Rouleux F, Monsigny M, Legrand A. The second intron of the human galectin-3 gene has a strong promoter activity down-regulated by p53
. FEBS Lett
20. Guittaut M, Charpentier S, Normand T, Dubois M, Raimond J, Legrand A. Identification of an internal gene to the human Galectin-3 gene with two different overlapping reading frames that do not encode Galectin-3
. J Biol Chem
21. Remy I, Michnick SW. A highly sensitive protein-protein interaction assay based on Gaussia luciferase
. Nat Methods
22. El Haddad S, Serrano A, Normand T, Robin C, Dubois M, Brulé-Morabito F, et al. Interaction of Alpha-synuclein with Cytogaligin, a protein encoded by the proapoptotic gene GALIG
. Biochem Biophys Res Commun
23. Chinta SJ, Mallajosyula JK, Rane A, Andersen JK. Mitochondrial α-synuclein accumulation impairs complex I function in dopaminergic neurons and results in increased mitophagy in vivo
. Neurosci Lett
24. Winslow AR, Chen C-W, Corrochano S, Acevedo-Arozena A, Gordon DE, Peden AA, et al. α-Synuclein impairs macroautophagy: implications for Parkinson's disease
. J Cell Biol
25. Gonzalez P, Duneau M, Charpentier S, Normand T, Mollet L, Dubois M, et al. Destabilization of membranes containing cardiolipin or its precursors by peptides derived from mitogaligin, a cell death protein
26. Duneau M, Boyer-Guittaut M, Gonzalez P, Charpentier S, Normand T, Dubois M, et al. Galig, a novel cell death gene that encodes a mitochondrial protein promoting cytochrome c release
. Exp Cell Res
27. Gonzalez P, Robinet P, Charpentier S, Mollet L, Normand T, Dubois M, et al. Apoptotic activity of a nuclear form of mitogaligin, a cell death protein
. Biochem Biophys Res Commun
28. Mollet L, Robinet P, Dubois M, Aurouet A, Normand T, Charpentier S, et al. Opposing Mcl-1, the GALIG proapoptotic gene is upregulated as neutrophils die and underexpressed in acute myeloid leukemia cells
. Mol Immunol
29. Pfaffl MW. A new mathematical model for relative quantification in real-time RT-PCR
. Nucleic Acids Res
30. Hindson BJ, Ness KD, Masquelier DA, Belgrader P, Heredia NJ, Makarewicz AJ, et al. High-throughput droplet digital PCR system for absolute quantitation of DNA copy number
. Anal Chem
31. Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: machine learning in python
. J Mach Learn Res
32. Metz CE. Basic principles of ROC analysis
. Semin Nucl Med
33. Deeks SG, Lewin SR, Havlir DV. The end of AIDS: HIV infection as a chronic disease
34. Ipp H, Zemlin A. The paradox of the immune response in HIV infection: when inflammation becomes harmful
. Clin Chim Acta
35. Guil S, Esteller M. DNA methylomes, histone codes and miRNAs: tying it all together
. Int J Biochem Cell Biol
36. Stankov MV, Panayotova-Dimitrova D, Leverkus M, Vondran FWR, Bauerfeind R, Binz A, et al. Autophagy inhibition due to thymidine analogues as novel mechanism leading to hepatocyte dysfunction and lipid accumulation
37. Sáez-Cirión A, Bacchus C, Hocqueloux L, Avettand-Fenoel V, Girault I, Lecuroux C, et al. Post-treatment HIV-1 controllers with a long-term virological remission after the interruption of early initiated antiretroviral therapy ANRS VISCONTI study
. PLoS Pathog
38. Hocqueloux L, Avettand-Fènoël V, Jacquot S, Prazuck T, Legac E, Mélard A, et al. Long-term antiretroviral therapy initiated during primary HIV-1 infection is key to achieving both low HIV reservoirs and normal T cell counts
. J Antimicrob Chemother