SERINC5 down-regulation activity correlates with CD4 down-regulation activity

The Nef-mediated SERINC5 down-regulation ability was determined for 106 Nef clones, each isolated from a unique individual in early/acute HIV-1 subtype C infection, to assess the effect of this Nef function on subsequent disease progression as well as its contribution to overall Nef function. The Nef clones used for this study were prepared in a previous study [15], which confirmed subtype C lineage as well as Nef protein expression for a subset. The down-regulation of SERINC5 was measured by the co-transfection of the patient-derived Nef clones and a SERINC5 expression plasmid into a CEM-derived CD4 + T cell line, followed by detection of cell-surface SERINC5 using a fluorescently labelled antibody and flow cytometry. The ability of each Nef clone to down-regulate SERINC5 was expressed relative to the positive control (the highly functional SF2 subtype B Nef isolate, representing 100% activity) and a negative control (the defective G2A mutant of SF2, representing no activity) (Fig. 1a). The assay showed good reproducibility with a significant correlation between the duplicate measurements (Spearman’s correlation; r = 0.93 and p < 0.0001). Overall, Nef clones varied widely in SERINC5 down-regulation ability: the median SERINC5 down-regulation activity was 85% (interquartile range [IQR], 60–94%) (Fig. 1b).

Fig. 1

SERINC5 down-regulation activities of patient-derived Nef clones. A Representative flow cytometry plots of the positive control (SF2 Nef) and negative control (SF2 Nef with the G2A mutation, rendering it inactive for SERINC5 down-regulation) are shown. The median fluorescence intensity (MFI) values indicate SERINC5 cell-surface expression in green fluorescent protein (GFP) expressing cells (representing cells transfected with Nef clones). B The SERINC5 down-regulation activities of Nef clones derived from patients from the HPP acute infection, TRAPS and Tshedimoso cohorts are shown. SERINC5 down-regulation activities of Nef clones were expressed relative to SF2 (100% down-regulation) and G2A (0% down-regulation)

The Nef clones were from three different cohorts, namely the HPP acute infection cohort (n = 32), Tshedimoso cohort (n = 27) and TRAPS cohort (n = 47) (Fig. 1b). While the sequences of the Nef clones from the different cohorts were previously shown to intermingle in a phylogenetic tree [15], there was a significant difference in the distribution of SERINC5 down-regulation values between the three cohorts (Kruskal–Wallis; p = 0.001). The highest median SERINC5 down-regulation activity was observed in the Tshedimoso cohort (95%; IQR, 74–97%) followed by 86% (IQR, 68–92%) in the TRAPS cohort and 66% (IQR, 46–91%) in the HPP acute infection cohort, where there was a significant difference specifically between the Tshedimoso and HPP acute infection cohorts (Dunn’s multiple comparisons test; p < 0.001). This cohort difference in SERINC5 down-regulation function warranted careful consideration of cohort effects in the models assessing the effect of SERINC5 down-regulation on viral load set point and rate of CD4 + T cell decline.

Previously, other Nef functions were measured for the same Nef clones [15, 16]. These included CD4 and HLA-I down-regulation, using flow cytometry-based methods similar to that for SERINC5 down-regulation [15]. Alteration of TCR signalling was also previously measured using a high throughput NFAT-based luciferase reporter T cell assay to measure the ability of each Nef clone to inhibit NFAT, a downstream molecule of TCR signalling, following TCR stimulation [16]. While there was significantly lower Nef-mediated alteration of TCR signalling activity in the Tshedimoso cohort compared to the HPP acute infection cohort, there was no difference in CD4/HLA-I down-regulation activities between the cohorts.

All the Nef functional measurements for these Nef clones are available in Additional file 1. When investigating relationships between the different Nef functions, a statistically significant correlation between SERINC5 down-regulation activity and CD4 down-regulation activity (Spearman’s correlation; r = 0.63 and p < 0.0001) was observed (Fig. 2). On the other hand, there was no correlation observed between SERINC5 down-regulation activity and HLA-I down-regulation activity (Spearman’s correlation; r = 0.13 and p = 0.16) and neither between SERINC5 down-regulation activity and alteration of TCR signalling (Spearman’s correlation; r = − 0.04 and p = 0.66) (Fig. 2).

Fig. 2

Comparison between SERINC5 down-regulation activity and the other Nef functions. A  The graph shows a statistically significant correlation (Spearman’s correlation) between SERINC5 down-regulation and CD4 down-regulation activities of patient-derived Nef clones, while graphs in panels B and C show no relationship between SERINC5 down-regulation activity and the activities of HLA-I down-regulation and alteration of TCR signalling. All Nef functions were expressed relative to that of SF2 Nef (100% activity)

SERINC5 down-regulation activity does not associate with markers of disease progression

In previous analyses of the same Nef clones, CD4 down-regulation function correlated positively with viral load set point and higher HLA-I down-regulation activity associated with a faster rate of CD4 + T cell decline, while the alteration of TCR signalling function did not correlate with either of these markers of disease progression [15, 16]. Nef-mediated SERINC5 down-regulation activity is responsible for enhancing virion infectivity, but it is not known to what extent SERINC5 down-regulation influences HIV-1 subtype C disease progression. To investigate this, the SERINC5 down-regulation activities of Nef clones derived from early infection were analysed together with subsequent viral load set point and the rate of CD4+ T cell decline using univariable and multivariable linear regression.

There was no significant effect of SERINC5 down-regulation on viral load set point in both the univariable and multivariable linear regression analyses (p = 0.59 and p = 0.28, respectively) (Table 1). Multivariable linear regression included cohort as a variable since there was a significant difference in SERINC5 down-regulation between cohorts. Interestingly, while there was no significant relationship between SERINC5 down-regulation and viral load set point overall, within the Tshedimoso cohort alone (the cohort with the highest SERINC5 down-regulation function) there was a significant positive correlation between SERINC5 down-regulation and viral load set point (Spearman’s correlation; r = 0.46 and p = 0.02) (Fig. 3).

Table 1 The association between SERINC5 down-regulation and viral load set pointFig. 3

The relationship between SERINC5 down-regulation activity and viral load set point by cohort. The SERINC5 down-regulation activity was measured for Nef clones derived from three different cohorts, namely the HPP acute infection cohort (A), the Tshedimoso cohort (B) and the TRAPS cohort (C). SERINC5 down-regulation correlated significantly with viral load set point in the Tshedimoso cohort only (Spearman’s correlation). SERINC5 down-regulation activity was expressed relative to that of SF2 Nef (100% activity)

There was no significant effect of SERINC5 down-regulation on the rate of CD4+ T cell decline in univariable or multivariable analysis, which included cohort, follow-up time, baseline CD4 + T cell count and baseline viral load (p = 0.15 and p = 0.45, respectively) (Table 2).

Table 2 The association between SERINC5 down-regulation and rate of CD4 + T cell decline

In summary, there was no significant relationship between SERINC5 down-regulation and markers of disease progression overall.

SERINC5 down-regulation contributes significantly to overall Nef function

An E value, which is a proxy for overall Nef function in vivo, has been predicted by computational modelling for each of the patient-derived Nef clones used in this study [17]. In past work, different Nef functional measurements were used as predictors of the E value with multiple linear regression to assess the contribution of each Nef function to overall Nef fitness [17]. In that study, CD4 down-regulation emerged as the strongest contributor of the Nef functions measured, however these measurements did not include SERINC5 down-regulation. Here, the contribution of SERINC5 down-regulation ability to overall Nef function was assessed using E values that were previously assigned to each of the clones and were derived from the Ising (dE0 values) and Potts models (dE90 values) [17] (dE0 and dE90 values for these clones are listed in Additional file 1). The Potts model accounts for the diversity of amino acids present at each residue. In contrast, in the Ising model, only the consensus amino acid present at each residue was modelled explicitly, and all other amino acids were treated as the same mutant type. A high dE0 or dE90 value is interpreted as corresponding to low in vivo Nef fitness (or a high fitness cost), while a low dE0 or dE90 value is interpreted as corresponding to high in vivo Nef fitness. The distribution of dE0 and dE90 values and their correlation with SERINC5 down-regulation values is shown in Fig. 4a, b (Spearman’s correlation; r = − 0.33 and p = 0.0005 for dE0 and r = − 0.39 and p < 0.0001 for dE90).

Fig. 4

The relationship between dE0/dE90 values and SERINC5 down-regulation as well as viral load set point. SERINC5 down-regulation activity of patient-derived Nef clones (expressed relative to SF2 Nef, which represents 100% activity) correlated significantly (Spearman’s correlation) with dE0 values (A) and dE90 values (B), which are proxies for overall Nef function in vivo. dE0 values were derived from the Nef fitness landscape Ising model (only the consensus amino acid present at each residue was modelled explicitly) for each Nef clone, while dE90 values were derived from the Nef fitness landscape Potts model (each amino acid present at each residue was modelled explicitly) [17]. The dE0 values (C) and dE90 values (D) also correlated significantly with viral load set point (Spearman’s correlation)

To allow for assessment of the relative contribution of each Nef function to dE0/dE90, Nef functions were standardised according to the means and the relationship between each standardised Nef function and dE0 was analysed using univariable and multivariable analysis (Table 3). In the univariable analysis, both CD4 down-regulation and SERINC5 down-regulation were significantly associated with dE0 with similar coefficients (− 2.22 and − 2.23) that were larger than those of the other Nef functions. A negative coefficient means that an increase in Nef function is associated with a decrease in dE0 value (i.e. an increase in in vivo Nef fitness). In multivariable analysis, however, these effects were not significant. We next explored whether the difference between the univariable and multivariable analysis was due to the possible impact of collinearity. This was considered due to the correlation between SERINC5 and CD4 down-regulation observed here, together with the correlation between HLA-I down-regulation and alteration of TCR signalling described previously [16]. This was, however, ruled out by assessing the variance inflation factors. A value greater than the threshold value of 5 would have suggested the influence of collinearity; however, all were below the threshold value of 5.

Table 3 The association between Nef functions and dE0 values

The relationship between each standardised Nef function and dE90 is presented in Table 4. In the univariable analysis, CD4 down-regulation and SERINC5 down-regulation were significantly associated with dE90 (p = 0.02 and p = 0.003, respectively). The greatest relative contribution to the overall Nef function was observed for SERINC5 down-regulation with a model coefficient of − 4.91 followed by CD4 down-regulation with a coefficient of − 3.64. However, in the multivariable analysis, only SERINC5 down-regulation remained statistically significant and the greatest driver of dE90, with a coefficient of − 5.32 (p = 0.003). This means that a one unit standardised increase in SERINC5 down-regulation function is associated with a 5.32 unit decrease in dE90.

Table 4 The association between Nef functions and dE90 values

Of interest, the dE0 and dE90 values were both correlated significantly with viral load set point (Spearman’s correlation; r = − 0.2 and p = 0.046, and r = − 0.22 and p = 0.03, respectively) (Fig. 4c, d), suggesting that the E value measure of overall Nef function has relevance for disease progression in vivo. However, neither dE0 (p = 0.85) or dE90 (p = 0.94) were significant predictors of the rate of CD4 + T cell decline in multivariable regression analysis that controlled for potential confounding factors (Tables 5, 6).

Table 5 The association between dE0 values and rate of CD4 + T cell declineTable 6 The association between dE90 values and rate of CD4 + T cell decline

In summary, the results suggest that CD4 down-regulation and SERINC5 down-regulation are the largest contributors of the Nef functions considered here to overall Nef function and that the contribution of SERINC5 down-regulation is the most significant. Results further suggest that overall Nef function affects viral load set point.

Sequence determinants of Nef-mediated SERINC5 down-regulation activity

To identify Nef amino acids that either increase or decrease the ability of Nef to down-regulate SERINC5, a function-sequence analysis was performed using an online tool [18] that generates codon-by-codon Mann–Whitney U tests for every Nef amino acid variant present at least 5 times in the dataset.

We identified 15 amino acid variants at 11 different codons associated with altered SERINC5 down-regulation activity (Table 7). The most statistically significant association was observed at codon number 65, where Nef clones that encoded the consensus amino glutamic acid (n = 94) displayed lower SERINC5 down-regulation activity (median 81.3%) compared with clones that did not (n = 12; median 95.1%) (p = 0.0005). The amino acid variants that were associated with the largest alterations (by more than 30%) in SERINC5 down-regulation activity were at codons 10, 11, 38 and 173. Together these results suggest that natural polymorphisms in Nef can lead to both increased and decreased ability to down-regulate SERINC5.

Table 7 Nef amino acids significantly associated with altered SERINC5 down-regulation activity

To confirm the results of the codon-by-codon analysis, four amino acid variants (10K, 135F, 173T and 176T) were selected and introduced into a consensus C Nef by site-directed mutagenesis. While mutants 10K, 135F and 176T modestly reduced SERINC5 down-regulation activity (by 11–16%, p > 0.05), the 173T mutation markedly reduced SERINC5 down-regulation ability (reduced by 92% relative to wild-type; ANOVA with Tukey post-hoc test, p < 0.001) (Fig. 5). To confirm the deleterious effect of the 173T mutation on SERINC5 down-regulation, it was also introduced into a patient-derived Nef sequence (SK446). In the SK446 sequence background, 173T was similarly associated with a substantial reduction in Nef-mediated SERINC5 down-regulation activity (by 57%; Student’s T test, p = 0.015) (Fig. 5).

Fig. 5

SERINC5 down-regulation activity of Nef mutants. A Flow cytometry plots of the consensus C Nef (WT) and the consensus C Nef 173T mutant (173T) are shown, with SERINC5 cell-surface expression on the y axis and the green fluorescent protein (GFP) expressing cells (representing cells transfected with Nef clones) on the x axis. The SERINC5 down-regulation activities of the consensus C Nef mutants (B) and SK446 Nef mutant (C) normalised to the respective WT proteins (representing 100% activity) are shown, where the data represents the means and standard deviations of three independent experiments. The SERINC5 down-regulation ability expressed relative to SF2 Nef was 80% and 91% for wild-type (WT) consensus C Nef and SK446 Nef, respectively. ANOVA with Tukey post-hoc tests was performed to assess which consensus C Nef mutants differed significantly from WT, and the Student’s T test was used to compare the down-regulation activity of the SK446 WT and SK446 173T mutant. The ANOVA/Student’s T test p values are shown and the Tukey post hoc test p value is indicated by asterisks (*** is p < 0.001)