The question of when antiretroviral therapy has to be initiated remains a challenging issue. Recent studies show that the early immune response to HIV-1 infection is likely to be an important factor in determining the clinical course of disease [1]. The first weeks following HIV-1 transmission are extremely dynamic. They are associated with rapid damage to generative immune cell micro-environments and with immune responses that partially control the virus. Following HIV-1 infection, the virus first replicates locally in the mucosa and then is transported to draining lymph nodes where further amplification occurs. This initial phase of infection, until the systemic viral dissemination begins, constitutes the eclipse phase [1]. In general, there is an exponential increase in plasma viremia with a peak 21-28 days after infection. By this time, significant depletion of mucosal CD4⁺T cells has already occurred. Around the time of peak viremia, patients may become symptomatic and reservoirs of latent virus are established [1, 2].

The "window of opportunity" between the infection and peaking of viremia, prior to massive CD4⁺ T cell destruction and the establishment of viral reservoirs, seems to be a narrow but crucial period in which an antiretroviral therapy can control viral replication, prevent an extensive CD4⁺ T cell depletion from occurring and curb generalized immune activation. Thus, thwarting HIV replication by introducing HAART in the early phases of infection could have a substantial impact on the whole disease course. In particular, suggested factors that may contribute to the observed better viral control after treatment interruption in very early treated patients are [3]: i) early arrest of viral escape, leaving the virus vulnerable; ii) preservation or even enhancement of the immune response resulting from the early clearing of antigen; iii) prevention of the establishment of a pool of HIV-specific memory CD4 T cells thus leaving fewer target cells available for viral infection.

An ideal clinical model aimed at addressing such issue should compare a number of patients treated starting on different times: from very early to very late. Besides the ethical issues, it is rather difficult to collect enough patients to significantly represent the whole spectrum of possible HAART initiation timings. As a matter of fact, a practical clinical model would compare very early to late-treated patients. While informative on the overall role of HAART timing on disease course, this approach would not allow to verify if there are events in the early infection influenced by the starting time of HAART that affect directly and decisively the course of the disease.

We have already shown in [4–6] that an agent-based model of HIV-1 infection could be a valuable tool for the study of the AIDS disease progression and treatment. The computerized simulation allows us to track the effect of HAART timing on the progression of the disease.

The aim of the present work is to verify the effect of HAART timing on subsequent events. Indeed, both a clinical model and a computational simulation show that a late initiation of treatment affects HIV-1 replication control. Interestingly, the in silico model identifies a significant three-week time threshold as the "ultimate" time point beyond which the decisive HIV-induced damages already occurred, affecting the whole disease course.

In a previous work [4], we analyzed clinical data of patients initiating HAART within six months from infection (i.e., we called that early phase) and performed computer simulations to predict the differences in viral rebound at therapy interruption between those patients and subjects initiating therapy during the chronic phase (i.e., we called that late phase, corresponding to initiating six or more months after infection). Our conclusion was that early initiation of therapy does not prolong the disease-free period when compared to a treatment started during the late phase. However, other studies suggest that an earlier initiation is preferable [7–9]. This motivated us to better identify the meaning of early initiation. In the present article, we extend the analysis in [4] to get a more complete picture. We analyze clinical data of very early patients (i.e., treated before seroconvertion) against late-treated patients.

We analyze virological data from HIV patients treated during the very early, early and late phase of infection and compare them with computer simulations.

In Figure 1 clinical data of all three analyzed groups is shown altogether. In a point-by-point comparison we find no statistical difference in viral rebound between early and late-treated patients (P ≥ 0.05, Mann-Whitney U two-tailed test) confirming the results of [4]. In addition, we observe that a difference does exist for very early initiation of therapy (P < 0.05, Mann-Whitney U two-tailed test).

In the present work we extend the simulations of [4] to include the new clinical settings corresponding to a very early initiation of therapy. In particular, the very early simulation settings correspond to a beginning of the therapy within the first week whereas the late settings correspond to initiating therapy between week five and six from infection.

Figure 2 summarizes data of virological rebound (averages) after therapy interruption at different time points (4, 8 and 24 weeks) for very early and late patients for both clinical (empty boxes) and simulation data (filled boxes). Firstly, the figure shows that clinical and simulation data are in good agreement (differences are not statistically significant: P ≥ 0.05, Mann-Whitney U two-tailed test). Secondly, the difference in plasma HIV-1 RNA (copies/ml) between very early and late-treated patients decreases with increasing time from therapy interruption. Panel (a) shows a difference of about two logs with panel (b) for both clinical data and simulations. These differences vanish after 24 weeks from therapy interruption (cfr. panel (e) and (f)). The overall message is that a delay in the initiation of therapy reduces the chances of maintaining a therapy effect at discontinuation.

In order to provide a more precise estimate of the time "limit" beyond which the benefit of an early initiation of therapy vanishes, we use the simulation to investigate the influence of HAART initiation time (t _s ) on the viral rebound. The corresponding results are shown in Figure 3. The virological rebound at one week after therapy interruption as a function of t _s is presented. We observe that there are two regimens, one for t _s < 20 days and one for t _s > 30 corresponding to what clinicians call respectively best controllers (with undetectable HIV RNA levels) and rebounders (whose HIV viremia load returns, approximately, to the pre-HAART level).

where the parameter k is the carrying capacity or the upper asymptote, a is the lower asymptote, d is the growth rate, and $t_s^{*}$ is the time of maximum growth. By moving the time of the measurements beyond one week after therapy interruption, the resulting data still fit the same V (t _s ) but with a greater a, a smaller d and a greater $t_s^{*}$. In particular the limit for d going to zero, of V (t _s ) is (a + k)/2 may lead to the deceiving conclusion that there is no window of opportunity because the viral rebound is independent from t _s .

With respect to $t_s^{*}$, the value of~ 23:5 days points to the early inflammation as a critical phase of the disease. To bring into focus this facet, we compare two simulated "markers" of the inflammation state in untreated (control case), very early and lately treated simulated patients (see Figure 4). These virtual markers are given by the cell counts of active macrophages (a) and dendritic cells presenting viral proteins on class II MHC molecules (b). We observe that the late-treated case is comparable to the control case (untreated) whereas the very early stands on its own. This observation suggests that it is the activation of the immune system through the set up of an in ammatory state that has to be blamed for the increased viral rebound for t _s > $t_s^{*}$.

Figure 4 shows with clarity that very early initiation of the treatment can down-regulate the immune activation, hence limiting viral replication and spread. Interestingly, this view is supported by the observation that HIV triggers the immune activation directly (e.g., HIV gene products can induce the activation of lymphocytes and macrophages as well as the production of pro-inflammatory cytokines and chemokines [2]) or indirectly (e.g., sustained antigen-mediated immune activation occurs in HIV-1-infected patients due also to other viruses like the cytomegalovirus or the Epstein-Barr virus [2]). In both case, the result is a high level of pro-inflammatory cytokines, such as tumor necrosis factor alpha, interleukin 6 and interleukin 1 beta, right from the early stages of HIV-1 infection [2].

Recent analysis (performed by Fiebig et al. [16]) of samples from individuals that have been infected by HIV-1 has revealed that patients can be categorized into six stages on the basis of a sequential gain in positive HIV-1 clinical diagnostic assays (viral RNA measured by PCR, p24 and p31 viral antigens measured by enzyme-linked immunosorbent assay (ELISA), HIV-1-specific antibody detected by ELISA and HIV-1-specific antibodies detected by western blot, [16]). Patients progress from acute to early chronic infection at the end of stage V, approximately 100 days following infection, as the plasma viral load begins to stabilize.

With respect to the study conducted by clinical data analysis and computer simulation described so far, we identify three regimens, as highlighted in Figure 3. These can be paralleled to Fiebig et al. stages [16]. In particular, we observe that patients treated with HAART in very early stages of the infection (stage I-III) are likely to better control the viremia after treatment interruption [3]. If therapy starts in the acute phase (stage V-VI) then the action of the drug foils the immune response and, as a consequence, at the end of the therapeutic period, the virus rebounds undisturbed.

These considerations are summarized in Figure 5 where we draw a schematic picture of the importance of an early initiation of HAART with respect to the progression of HIV markers according to Fiebig's et al. stages. In particular we identified the "window of opportunity" corresponding to stages I-III, that is, the first three weeks from primary HIV-1 infection. Patients receiving therapy in this narrow period are likely to turn out to be best controllers. Probably, the massive immune activation in the early stage of the disease favors the virus, as it finds more host target cells to exploit for replication. In point of fact, the ensuing massive depletion of CD4+ T cells in mucosal lymphoid tissues, can result in the disruption of the mucosal barrier in the gut. This barrier prevents the translocation of the intestinal flora from the gut to the systemic immune system restricting it to the lamina propia and the mesenteric lymph nodes [2]. HIV-1 infection is indeed associated with a significant increase of plasma lipopolysaccharide levels that is an indicator of microbial translocation, directly correlated with measures of immune activation.

A number of studies indicate that interfering with HIV replication by starting the therapy in the early phases of the infection could have a deep impact on the whole disease course. However, HAART is costly, it is onerous for both patient and health care provider, and often brings adverse effects. Its clinical benefit must therefore be weighed against its burden.

In the present study, we resorted to a computer model to study the dynamics of the plasma viral load after prolonged treatment interruption in two groups of in silico patients: those who initiate HAART very early and those who start it lately. We evaluated the model comparing the results to clinical data. We found that an opportunity time-window exists for the initiation of HAART (roughly within three weeks before the establishment of viral reservoirs), in which the therapy can control viral replication, preventing generalized immune activation and extensive CD4+ T cell depletion.