HIV influences clustering and intracellular replication of hepatitis C virus

Abstract

Background & Aims.

HCV and HIV coinfection is common and HIV leads to increased HCV viremia and accelerated disease progression. However, the biological basis of this interaction remains poorly understood and little is known about the impact of HIV on HCV replication at the cellular level.

Methods.

We analyzed HCV, based on single cell laser capture microdissection, in liver biopsies from monoinfected (n=4) and HCV/HIV coinfected (n=5) participants. HCV RNA was assayed in 3200 hepatocytes with information of spatial position. We compared HCV RNA levels and clustering properties of infection between mono- and coinfected participants, and developed a mathematical model of infection.

Results.

Although the median plasma HCV RNA level and the fraction of infected cells was comparable in monoinfected (7.0 log₁₀IU/mL and ~30%) and coinfected (7.3 log₁₀IU/mL and ~40%) participants, the median HCV RNA per infected hepatocyte in monoinfected (2.8IU) was significantly lower than in coinfected (8.2IU) participants (p=0.03). Clustering of infected cells was more prominent in monoinfected participants (91% of samples) than in coinfected participants (~48%), p=0.0045, suggesting that spatial spread may be influenced by HIV coinfection. Interestingly, when clustering does occur, the size of clusters is similar in both types of infection. A mathematical model of infection suggested that HIV allows higher intracellular accumulation of HCV RNA by impeding the export of HCV RNA.

Conclusions.

Our observations show that HIV coinfection impacts intracellular accumulation of HCV RNA and the clustering of HCV infected cells, but to a less extent the fraction of HCV infected cells.

Hepatitis C virus (HCV) is highly prevalent with almost 71 million infections globally¹. Coinfection with human immunodeficiency virus (HIV) is also common, with approximately 2.3 million coinfections worldwide², as these two viruses share transmission routes³. Highly effective treatment regimens involving combinations of direct-acting antiviral agents (DAAs) are available for HCV monoinfected and HCV-HIV coinfected individuals⁴; however, there is still no vaccine for either virus.

HIV coinfection is known to enhance replication of HCV and sub-genomic replicons in tissue culture^5,6 and to accelerate the progression of HCV disease, amplifying the chances of cirrhosis and mortality^7,8. It is important to better understand the in vivo impact of HIV on HCV replication in order to devise new intervention strategies³. Previously, we used single cell laser capture microdissection (scLCM) to quantify HCV RNA in single cells and to map the spatial spread of HCV infection in monoinfected participants^9,10. Here, we extend our observations to study the intrahepatic burden and spread of HCV in the setting of HIV coinfection compared to monoinfection. We investigate the impact of HIV on HCV replication by analyzing novel liver biopsy data using spatial statistical techniques and mathematical modeling.

Materials and Methods

Participant characteristics and data collection procedure

One liver biopsy, with sufficient parenchyma to obtain four to five spatial grids of 10×10 adjacent hepatocytes, was performed on each of five coinfected participants¹⁰. Similarly, one liver biopsy with sufficient parenchyma for three to four spatial (2D) grids of 10×10 adjacent hepatocytes was performed on each of four monoinfected participants⁹. Full details on the recruitment of participants, including informed consent, liver biopsies and single cell laser capture microdissection (scLCM) were provided before^9,10. All subjects were in the chronic stage of HIV and/or HCV and all had received <24 months of ART over their entire lives and none within 6 months of this study. For each spatial grid, we used scLCM to measure by PCR the HCV RNA content of each cell (details in 9). As discussed previously¹¹, some hepatocytes had detectable HCV RNA, but they could not be properly quantified due to missing expression of 7SL, a small ribosome-associated RNA, that was used as a normalizing factor. We refer to these cells with unquantifiable HCV RNA as unquantifiable hepatocytes. The unquantifiable hepatocyte was considered uninfected if and only if all neighboring cells were HCV RNA negative (or, uninfected), otherwise they were deemed HCV positive. We also note that the average percentage of infected hepatocytes with unquantifiable HCV was higher in monoinfected participants (9%) than in coinfected participants (4%), but this was not significant (p=0.38). In any case, to test the robustness of our results, we also repeated the clustering analyses ignoring the 4 grids (out of 41) that had more than 10% unquantifiable hepatocytes. This made no difference to the overall clustering tendency. We also imputed the amount of HCV RNA in these unquantifiable hepatocytes as the mean of the HCV RNA value in their neighboring cells. The summary statistics (Table 1) largely remained unchanged even when we assumed that these unquantifiable hepatocytes are uninfected demonstrating the robustness of our results.

Table 1:

Summary statistics of infection parameters for monoinfected and coinfected participants.

	(%) Infected cells (Total cells analyzed)	Mean HCV RNA per hepatocyte (IU)	Mean HCV RNA per infected hepatocyte (IU)	Maximum HCV RNA in a single hepatocyte (IU)	log10(HCV plasma viremia) (IU/mL)	ρ (/day)	ρR+ (virions/day/infected cell)
Monoinfected participants
P6	37.3% (300)	1.1	3.0	22.8	7.4	61.2	183.6
P7	23.0% (300)	1.2	5.2	50.3	6.9	16.9	87.9
P8	44.0% (300)	2.1	4.7	25.9	7.0	13.7	64.4
P9	8.7% (200)	0.7	5.2	20.0	6.9	46.0	239.2
Median	30.2% (300)	1.2	2.8	24.4	7.0	31.5	135.7
Coinfected participants
P1	42.8% (500)	3.2	7.6	54.3	7.3	11.8	89.7
P2	41.8% (400)	3.2	7.6	41.1	7.3	12.1	92.0
P3	52.2% (400)	10.5	20.0	96.6	6.8	1.3	26.0
P4	6.2% (400)	4.5	72.6	1383.0	6.1	0.5	36.3
P5	34.0% (400)	2.7	8.0	42.0	7.3	12.9	103.2
Median	41.8% (400)	3.2	8.2	54.3	7.3	11.8	89.7

Tests for clustering tendency and cluster properties

We analyzed clustering tendency in 2D by the Hopkins statistic (k), which quantifies how far from an uniform distribution the data is spread¹². On a 10×10 grid with H_G infected cells out of a total of N cells, the Hopkins statistics is calculated as follows:

1. Choose a small positive integer n = maximum(3,0.05×H_G), which should be much smaller than H_G but must be at least 3 since a cluster by definition must include at least 3 infected cells^12,13.

2. Randomly choose a set S₁ of n infected cells. Calculate the distance between each infected cell in S₁ and its nearest infected neighbor cell, call these distances w_i, i=1, …, n.

3. Similarly, randomly choose a set S₂ of n cells from all the N cells in the grid (not just the infected ones). Calculate the distance between each cell in S₂ and its nearest infected neighbor cell, call these distances q_i, i=1, …, n.

4. The Hopkins statistic (k) is then calculated as¹²

$$k=\frac{{\sum _{i=1}^n{q}_i}^d}{\sum _{i=1}^n{q}_i^d+\sum _{i=1}^n{w}_i^d}$$

where d=2 is the dimension. A schematic representation of how to implement this algorithm is given in Fig S1. If k≤0.5, then the distribution of infected cells approximates either a uniform or random distribution. However, as k increases from 0.5, the clustering tendency of infected cells increases. We calculated $k$ 1000 times by picking n different infected cells and n different random cells each time, and then affirming clustering of HCV infected cells on a grid if the 0.025 quantile of the distribution of k is greater than 0.5.

We also employed an alternative approach, based on quadrat analysis to confirm the clustering of HCV infected cells^14,15. In this method a region is sampled using a set of similar sized regions (quadrats) and counting the number of infected cells in each. The method we use is as follows:

1. A 10×10 grid is first divided into ‘m’ equally-sized non-overlapping regions, where m is the perfect square integer, i.e., 1, 4, 9, 16, 25, etc., closest to 2A/H_G and A is the area (=100 square arbitrary units).

2. Then the distribution of the number of infected cells per region, and the corresponding mean, variance and variance-to-mean ratio (VMR) are calculated.

If VMR≤1, then the HCV infected cells on the grid tend to be regularly or randomly distributed; otherwise the infected cells are clustered, with the clustering tendency increasing as VMR increases from 1. A schematic representation of the quadrat analysis is given in Fig S2.

We note that these methods do not account for edge effects, but when we compared the median percentage of infected cells at the edge of a grid in monoinfected vs. coinfected participants, these were not significantly different (median 5% vs 10% respectively; p=0.23 mixed-effects model).

Neither the Hopkins statistic nor the quadrat analysis have the capability to determine cluster properties. Therefore, we employ a commonly used density-based cluster algorithm, i.e., the DBSCAN algorithm, to determine cluster properties such as the number of clusters on a grid and the number of infected cells in a cluster, i.e., the cluster size. We prefer DBSCAN over k-means and agglomerative hierarchical clustering¹⁶, because it analyzes the data without any prior assumptions¹⁷. We assumed that on a grid, hepatocytes have 8 neighbors except for hepatocytes at the edges and corners of a grid, which have 5 and 3 neighbors, respectively⁹. To implement the DBSCAN algorithm, we chose infected hepatocytes as nodes and drew an edge between every two infected neighboring cells. Any two infected cells are identified as connected if there exists a path of ≥1 continuous edges between them. The DBSCAN algorithm then identifies clusters of at least 3 connected infected cells. The schematic representation of the DBSCAN algorithm is given in Fig S3. This algorithm does not determine the clustering tendency¹³, therefore we determine cluster properties, i.e., employ the DBSCAN algorithm, only on grids that show clustering by either the Hopkins statistic or the quadrat analysis.

Statistical Procedure

Assuming that the cell with the maximum HCV RNA in a cluster is the “center” of that cluster¹¹, we calculated the dimension of a cluster as the Euclidean distance between the center cell and the farthest infected cell in that cluster, assuming each cell occupies a square of 1×1 arbitrary units. To study the relationship between different variables of interest, we employed mixed-effects models. In this approach, the response (or dependent) variable is related linearly with the predictor (or independent) variable, allowing for a random effect grouped by participant (P_ID) and, when looking at clustering properties, grid (G_ID) nested within participant¹⁸. For example, to determine the relationship between the dimension of a cluster, D_c and HCV RNA content of the center cell, R_c we used the mixed-effects model represented in MATLAB R2017b by D_c ~ R_c + (1| P_ID:G_ID) fitted using the function ‘fitlme’. We also employed Fisher’s exact test, Wilcoxon rank sum test and Kolmogorov-Smirnov test to determine if infection characteristics are significantly different between monoinfected and coinfected participants at the 5% significance level.

Mathematical model

To explore the mechanisms responsible for the differences in HCV liver distribution between monoinfected and coinfected participants, we used the following model, which is adapted from^11,19,20.

$$\frac{dI}{dt}=\beta V(H-I)-\delta I$$

$$\frac{d{R}_{+}}{dt}=\alpha -\rho R_{+}-\mu R_{+}$$ (1)

$$\frac{dV}{dt}=\rho R_{+}I-cV$$

Here I, R₊ and V represent the concentration of infected hepatocytes, the mean level of intracellular HCV RNA (+) copies per infected hepatocyte and the HCV RNA (virus) concentration in plasma, respectively. To simulate the model, we fixed the following parameters H=8.16×10⁶ cells/mL is the equivalent-concentration of hepatocytes in an uninfected adult liver^21,22, μ=1.46/day is the degradation rate of R₊²⁰; c=22.3/day²⁰ and c=16.1/day²³ represent the clearance rate of HCV from the circulation in monoinfected and coinfected participants, respectively; while δ=0.14/day^11,24 and δ=0.21/day^23,25 represent the death rate of infected cells in monoinfected and coinfected participants, respectively. We have three unknown parameters in the model namely, (i) β, the virus infectivity, (ii) α, the synthesis rate of R₊ in infected cells, and (iii) ρ, the export rate of R₊ into virions.

With the assumption that HCV infection is in steady-state at the time of biopsy, at the measured values of I*, R₊* and V* from the data (Table 1), we compute the value of the three unknown parameters for each participant as

1. $\beta =\frac{\delta I^{\mathrm{*}}}{V^{\mathrm{*}}(H-I^{\mathrm{*}})}$ (from the equation $\frac{dI}{dt}=0$),

2. $\alpha =(\rho +\mu )R_{+}^{*}$ (from the equation $\frac{d{R}_{+}}{dt}=0$).

3. $\rho =\frac{c{V}^{\mathrm{*}}}{I^{\mathrm{*}}{R}_{+}^{\mathrm{*}}}$ (from the equation $\frac{dV}{dt}=0$),

For conversion between IU and HCV RNA copies, we employ 1 IU=1.96 HCV RNA copies¹¹.

Results

The median HCV RNA per hepatocyte is ~3 fold lower in monoinfected participants compared to coinfected participants

A total of 1100 and 2100 hepatocytes were analyzed for the presence of HCV RNA from 4 monoinfected and 5 coinfected participants, respectively (see Table S1 for the clinical characteristics of these participants). The quantification of HCV RNA in the 1090 infected hepatocytes yielded that infected hepatocytes contained between 1 and 100 IU of HCV RNA in both monoinfected and coinfected participants, except for one hepatocyte, from coinfected participant P4, which had 1383 IU (Figure 1).

Figure 1.

Relative frequency of HCV RNA levels in infected hepatocytes on all grids for monoinfected participants (Upper Panel) and for coinfected participants (Lower Panel). A total of 339 infected cells on 11 grids and 751 infected cells on 21 grids were analyzed for monoinfected and coinfected participants, respectively. For clarity, in the plot of the coinfected participants, one outlier cell with 1383 IU is not represented.

In Figure 1, we show the distribution of the HCV RNA level per infected hepatocyte. Over all grids, the mean (median) of HCV RNA in infected hepatocytes in monoinfected participants, 4.3 IU (2.8 IU), was much lower than in coinfected participants, 13.3 IU (8.2 IU) (p=0.03, mixed-effects model). Moreover, the maximum HCV RNA in an infected cell was ~28 fold lower in monoinfected participants (50.3 IU) than coinfected participants (1383 IU). This cell with 1383 IU was an outlier. If we ignore it, then the maximum HCV RNA in an infected cell was ~2 fold lower in monoinfected participants (50.3 IU) than coinfected participants (96.6 IU). We also calculated the mean HCV RNA per hepatocyte (infected or not), as in previous studies²⁶, which in monoinfected participants (1.3 IU) was also ~3.5 fold lower than in coinfected participants (4.7 IU). Thus, the presence of HIV infection seems to enhance intracellular HCV infection levels.

We next asked whether this increase in intracellular HCV RNA burden was also reflected in a larger fraction of infected cells in coinfected participants.

The distribution of the number of infected cells in monoinfected and coinfected participants is not significantly different

The distribution of the number of infected hepatocytes per grid is shown in Figure 2 for monoinfected and coinfected participants. The fraction of infected hepatocytes per grid is not significantly different for monoinfected and coinfected participants, i.e., median proportion of infected cells and [min, max] was 30% [9%, 61%] and 40% [0%, 62%], respectively (p=0.42, mixed-effects model). All grids (11/11) from monoinfected participants had at least nine infected hepatocytes, while for coinfected participants one out of 21 grids (~5%) had no infected cells and an additional 2/21 grids (~10%) had less than nine infected hepatocytes.

Figure 2.

Distribution of the number of infected cells on each grid for HCV monoinfected and HCV-HIV coinfected participants. A total of 11 and 21 grids were analyzed for monoinfected and coinfected participants, respectively.

Overall, the presence of HIV does not alter the distribution of the number of infected hepatocytes per grid (p=0.45, Kolmogorov-Smirnov test). So, we next investigated the clustering properties of infected cells in these two infection settings.

HCV infected cells are more clustered in monoinfected participants than in coinfected participants

To determine the clustering tendency of HCV infected cells on the 10×10 grids of hepatocytes, we used the Hopkins statistic (k) (see Methods and¹²). We found that 10/11 grids (grids 2–11) from monoinfected participants showed clustering; whereas only 1/21 grids (grid number 6) from coinfected participants showed clustering (Fig S4). To assess the robustness of this surprising finding, we also used a different technique for clustering tendency: the quadrat analysis, which is a somewhat more permissive analysis. Using it, we confirmed that clustering was indeed more common in monoinfected participants, who showed evidence of clustering in all grids in contrast to coinfected participants, with only 10/21 grids showing evidence of clustering (Fig S5). Thus, at least 10/11 grids (91%) from monoinfected participants show clustered HCV infection, while only at most 10/21 grids (48%) from coinfected participants indicate clustering and this difference is significant (p=0.0045, Fisher’s exact test).

We next defined and studied the properties of the clusters. To maximize the data to be analyzed, we defined “clustered grid” as any grid that showed such tendency by either of the methods used. Thus 11/11 grids and 10/21 grids from monoinfected and coinfected participants, respectively, are classified as clustered grids.

The distribution of the number of clusters on grids is different in monoinfected and coinfected participants, although the cluster sizes are comparable

To define the clusters on grids that showed a clustering tendency, as described in the previous section, we employed the DBSCAN algorithm. The DBSCAN analysis yielded the number and location of clusters on each grid (shown for coinfected participant P1 in Figure 3, and for the remaining participants in Supporting Information, Figs S6–S8).

Figure 3.

Identification of clusters using the DBSCAN algorithm on four clustered grids obtained from coinfected participant P1. In bold, we indicate the participant ID and G1-G4 as the grid number. Each square represents a hepatocyte on the 10×10 grid. The squares in white denote uninfected hepatocytes. The number of different colors (except black) employed to fill the squares indicates the number of clusters on that grid and the squares of one color represent infected cells belonging to one particular cluster. Hepatocytes marked with black represent infected cells that were not part of any cluster.

The distribution of the number of clusters on grids of monoinfected and coinfected patients was different (p=0.02, Kolmogorov-Smirnov test, Figure 4, Top panel). However, the median (mean) number of clusters across all grids was not statistically different in monoinfected participants 2.0 (1.9) and in coinfected participants 0 (1.3) (p=0.23, mixed-effects model) (Table S2).

Figure 4. Analysis of grids from monoinfected and coinfected participants.

(Top Panel) Distribution of number of clusters on a grid, and (Bottom Panel) Distribution of cluster sizes (i.e., the number of infected cells).

The cluster size, defined as the number of infected cells in a cluster, was similar in both groups with median [min, max] of 8 [3, 61] cells for monoinfected and 7 [3, 47] cells for coinfected participants (p=0.46, mixed-effects model) (Figure 4, Bottom Panel).

Altogether, these results indicate that clustering of HCV infected cells occurs less frequently in coinfected participants than in monoinfected participants, which could be due to different HCV spread dynamics. However, in grids with clusters, their number and size are similar in monoinfected and coinfected participants (Table S2). Therefore, we asked whether there were differences between grids with and without clusters, irrespective of the coinfection status.

There were no significant differences in the fraction of infected cells (p=0.83) in grids with and without clusters, nor in the maximum level of HCV RNA in infected hepatocytes (p=0.12). Interestingly, however, the median HCV RNA per hepatocyte on grids with clusters was lower (1.8) than on grids without clusters (4.7) (p=0.01). This is consistent with the observation that monoinfected participants have lower intrahepatic HCV RNA levels than coinfected participants, but higher percentage of clustered grids, as presented above.

We proceeded to analyze the properties of clusters in these participants, because these properties may provide information about the dynamics of HCV spread in the liver^9,11.

Maximum radius and cluster size are strongly correlated with the HCV RNA content in the central cell of the cluster for coinfected participants but not for monoinfected participants

We define the dimension of a cluster as the Euclidean distance between the cell with the highest content of HCV RNA (the central cell) and the farthest infected cell in that cluster. Using the mixed-effects model (see Statistical Procedure), the HCV RNA level in the center cell was unrelated with the dimension of the cluster in monoinfected participants (p=0.92), but this relationship was significant in coinfected participants (p=0.007). The coefficient of determination (r²), which indicates the proportion of variation in the dimension of the cluster explained by HCV RNA level in the center cell, was 5.4×10⁻⁴ and 0.25 for monoinfected and coinfected participants, respectively. Consistent with this, the correlation (r) between the HCV RNA in the center cell and cluster dimension was r=0.023 for monoinfected participants, but much larger for coinfected participants, r=0.50 (Figure 5(A)–(B)).

Figure 5. Correlation between the dimension of a cluster (A, B) or the number of infected cells in a cluster (C, D) and the HCV RNA content of the cell in the cluster with the maximum HCV RNA.

The dimension in (A) is the Euclidean distance between the cell with the highest content of HCV RNA (the central cell) and the farthest infected cell in that cluster. Each individual is represented by a different symbol, as indicated at the top of the figure.

Next, we analyzed the relationship between the HCV RNA level in the center cell and the number of cells in the cluster (i.e., its size). HCV RNA level in the center cell was unrelated with cluster size in monoinfected participants (p=0.28), but this relation was significant in coinfected participants (p=0.004). The coefficient of determination (r²) was 0.06 and 0.27 for monoinfected and coinfected participants, respectively, corresponding to a correlation r=0.24 for monoinfected participants, and r=0.52 for coinfected participants (Figure 5(C)–(D)).

Previously¹¹, we analyzed the profile of decay in the level of HCV RNA in infected cells from the center of the cluster to its periphery. Similarly, we analyzed here the relative difference in the HCV RNA content between the central cell of a cluster and the neighbor with the next highest HCV RNA level, and found that the decay was smaller in in monoinfected participants, 42% (95% CI: [29%, 54%]), than in coinfected participants, 54% (95% CI: [44%, 63%]), but this difference did not reach significance (p=0.16, mixed-effects model) (Figure 6(A) and Fig S9).

Figure 6.

Boxplot of (A) the percentage decrease in the HCV RNA content between the center cell in a cluster and the neighbor cell with the next highest HCV RNA content (or, the first generation cell) for monoinfected vs. coinfected participants, and (B) the export rate of R₊(/day) based on the mathematical model in the main text. Dotted horizontal line represents median while the black diamond represents mean and the bottom and top edges of the box indicate the 25th and 75th percentiles, respectively.

HIV promotes intracellular RNA accumulation by impeding export of HCV virions

We employed a model of HCV replication and spread (see Model (1) in Methods) to analyze the replication processes that could contribute to the observed differences in HCV RNA content in infected cells between monoinfected and coinfected participants as reported in Table 1. The model includes parameters values fixed from previous studies (see Methods) and free parameters that were determined by the data. These are the synthesis rate of intracellular HCV RNA, the infection rate of hepatocytes by virus, and the virion export rate (see Table S3). We found that only the median export rate per HCV RNA was significantly different between monoinfected (31.5/day) and coinfected (11.8/day) participants (Table 1 and Figure 6(B)) (p=0.02, Wilcoxon rank sum test). Despite a significant difference in the export rate, the viral loads and fraction of infected cells are similar in the two groups, which suggested that the higher intracellular levels of HCV RNA probably compensate for the lower export rate in coinfected participants. Thus, we calculated the total HCV RNA exported per day as virions (that is, the viral production per day per infected cell, ρR₊) and found no statistical difference (p=0.42, Wilcoxon rank sum test), even though the median viral production per day per infected cell was still higher in monoinfected participants (135.7 virions/day) than in coinfected participants (89.7 virions/day).

To make sure this finding is not due to the assumptions made for other parameters in the model, which were based on previous studies (as explained in “Mathematical Model” in Methods), we repeated our calculations assuming that the clearance rate of free virus and the loss rate of infected cells is the same in both coinfected and monoinfected (c=20/day and δ=0.14/day). We found that our results are robust, and the median export rate ρ is still higher in monoinfected participants (28.2/day) than in coinfected participants (14.7/day), and the other parameters were again not different.

To further test the robustness of this result, we also used an extended model (SI, section SI3, Fig S10 and Table S4 and references^11,27) including more details of the intracellular replication of HCV. The objective was to test whether the increase in the intracellular accumulation of HCV RNA in coinfected participants could be due to an increase in viral production not captured in the simpler model (see (+)RNA synthesis rate, γ, in Table S4). In this model, again the only parameter that was consistently different was the export rate ρ (p=0.016), as found in the simpler model. Thus, it is possible that HIV promotes intracellular HCV RNA accumulation by impeding export of HCV RNA(+).

DISCUSSION

The study of HCV infection at the cellular and extracellular level has improved our understanding of the HCV life cycle^9,11,28–34. It is well known that HIV is an important comorbidity of HCV infection^5–8, but the direct and indirect impacts of HIV on HCV replication are less-known and are the subject of interest³. This study contributes to fill this gap in knowledge by comparing HCV infection at the spatial and cellular level in monoinfected and coinfected participants.

Our analysis revealed that the number of HCV infected cells and viremia levels were comparable in both groups. The percentage of HCV infected cells in monoinfected participants (median: 42%, range: 8.7–44.0%) and in coinfected participants (median: 30%, range: 6.3–52%) was similar between the two groups in agreement with previous studies^30,33,35. We also found that monoinfected participants had lower median HCV RNA(+) per cell (1.2 IU) compared to coinfected participants (3.2 IU), which was in line with previous studies reporting lower intracellular HCV RNA in the setting of monoinfection than coinfection²⁶. In contrast with previous studies, here we could also analyze the HCV RNA content per infected hepatocyte and found the median to be 2.8 IU and 8.2 IU, in monoinfected and coinfected participants, respectively. This may indicate that HIV enhances (directly or indirectly) intracellular HCV RNA accumulation, consistent with previous results^5,6,36. Our modelling results also support the idea that HIV inhibits export of HCV leading to a higher accumulation of HCV RNA(+) in infected cells. The mechanisms underlying these processes are not fully understood and need to be investigated further.

The most striking difference between monoinfected and coinfected participants in our study was that there was a substantial difference in infected cell clustering tendency in the two groups. All biopsy grids from monoinfected participants showed evidence of this clustering, whereas this was true only for about half of the grids from coinfected participants. Importantly, the average number of infected cells on grids with and without clusters was not significantly different, and thus the level of infection can’t be responsible for masking any clustering effect. Interestingly, analyzing only grids exhibiting clusters, we found that several properties including the number of clusters per grid, the average number of infected cells in a cluster and the maximum number of infected cells in a cluster were comparable between monoinfected and coinfected participants. However, grids with evidence of clustering showed significantly lower levels of intracellular HCV RNA per hepatocyte than grids without clusters. How higher levels of intracellular HCV RNA directly or indirectly affect clustering (or vice-versa) needs to be investigated in the future. A question one can ask is why so many hepatocytes are uninfected, even though they are proximal to infected hepatocytes. The mechanism(s) for such co-existence in close proximity are unclear and the possibilities include the dynamic nature of the spread of HCV infection or the protected state of a subset of hepatocytes against HCV infection. Another possibility is that 3D spatial factors, such as collagen deposition and the sinusoids, may make the clusters asymmetric.

It is important to consider why HIV lowers the clustering tendency of HCV infected cells, despite comparable percentage of HCV infected cells in monoinfection and coinfection. This is consistent with the possibility of more efficient cell-to-cell spread in monoinfected participants compared to coinfected participants³³. However, it still needs to be investigated if this is enough to explain the lower clustering tendency in coinfected participants, or whether other host and/or viral processes are also crucial for the way HCV infection spreads in coinfected participants³⁰. One possible contributing factor, given the two processes by which HCV can infect a cell, directly from plasma or cell-to-cell, is if there is a greater barrier to plasma infection in monoinfection^37–39. As previous studies have shown that the level of HCV neutralizing antibodies is lower in coinfection than in monoinfection^37–39, perhaps with HIV coinfection (proportionally) more plasma seeding events occur than in monoinfection. It is also possible that other factors related to immune responses, which are impaired in HIV coinfection^38,40, contribute to these differences in the two groups⁶. For example, HIV gp120 protein can induce IL-8 production in hepatocytes which in turn leads to the attenuation of the anti-HCV action of IFN-$\gamma$³⁶, and thereby, could support a higher intracellular HCV RNA levels in HIV coinfection.

In this study, we did not analyze immune responses, and thus it is difficult to explore specific mechanisms responsible for the differences observed. In addition, although the number of single hepatocytes analyzed is unprecedented, 300–500 hepatocytes per person is a small sample compared to the total number of hepatocytes in the liver. We should also note that we had fewer grids sampled in monoinfected patients, but this is likely to have a minor impact, because our results were very consistent across all grids in both monoinfected and coinfected patients.

In conclusion, in this study we characterize clustering of HCV-infected hepatocytes in HCV monoinfection and HIV/HCV coinfection: coinfected participants had less evidence of clustering than monoinfected participants, despite similarities in the size of infected cell clusters, when present. In addition, intracellular HCV RNA was markedly higher in coinfection compared to monoinfection. These results may contribute to understanding the differences in HCV replication and liver disease progression in people who have HIV/HCV co-infection compared to people with HCV mono-infection.

Abbreviations

HCV

hepatitis C virus

HIV

human immunodeficiency virus

IU

international units

scLCM

single cell laser capture microdissection

VMR

variance-to-mean ratio