In this study on a cohort of South African patients with DLBCL and a high HIV seropositivity rate, M1 macrophages predominated by a small margin within the TME. While M2 macrophage numbers appeared relatively higher among the people living with HIV, this finding lacked statistical significance (possibly owing to the small number of HIV-negative patients enrolled). M1 macrophage numbers did not differ according to HIV status but were significantly and independently associated with survival. These findings differ somewhat from those of Liapis et al., who also reported no significant association between macrophage numbers and HIV status, but found that macrophage numbers were not associated with survival in HIV-DLBCL [12]. In contrast to previous reports [2,3,4], survival in relation to macrophage numbers was strongly associated with low M1 and not high M2 macrophage numbers. This suggests a dominant role of the M1 pro-inflammatory anti-cancer immune response in HIV-DLBCL and would seem to downplay the negative impact of the M2-mediated anti-inflammatory effect provided sufficient numbers of M1 macrophage are present.

Surprisingly, there was no relationship between either pro-inflammatory markers (such as CRP and IL6) or HIV-related parameters and macrophage numbers of any subtype, although ferritin levels were marginally associated with both M1 (inversely so) and M2 numbers. The lower ferritin levels in cases with higher M1 macrophage numbers are perplexing, as this goes contrary to the prevailing dogma that ferritin levels are higher in M1 macrophages, which sequester iron to make it less bioavailable to pathogens and tumour cells [15]. In contrast, M2 macrophages typically release iron to the surrounding stroma/circulation through ferroportin-mediated iron egress and store less iron as ferritin [15]. The higher ferritin levels in cases with lower M1 and higher M2 macrophage numbers in this study are thus unexpected. However, ferritin expression by M2 macrophages has been described previously in the setting of breast cancer, where it was a negative prognostic marker in lymph node–negative disease [16]. Furthermore, there is some evidence that ferritin may drive M2 macrophage polarisation in the context of wound healing [17], possibly as part of the homeostatic negative feedback response to a pro-inflammatory stimulus. Escalating ferritin levels could thus result in diminished M1 macrophage numbers as they are switched to M2 macrophages. Ferritin is known to have anti-oxidant activity, potentially protecting cancer cells from reactive oxygen species and chemotherapy-related cytotoxicity [18]. Since ferritin [19] and cytokines (including IL10 [20,21,22]) are known to be produced by some tumour cells, the high ferritin and M2 macrophage numbers here could also reflect paracrine manipulation of the immune response by the tumour cells owing to the tumour biology.

Since IL10 is well known to both drive M2 macrophage polarisation and to be produced by these cells, it was not surprising that the number of M2 macrophages was significantly correlated with IL10 levels. However, while we have previously shown high IL10 levels to be significantly associated with inferior survival, M2 macrophage numbers were not. Similarly, M2 macrophage numbers were marginally correlated with ferritin levels, a factor we showed previously to also be significantly and independently associated with survival in HIV-DLBCL [14]. Seemingly, the negative impact of IL10 and ferritin is not due to M2 macrophage–mediated tumour tolerance.

Rituximab therapy is now well recognised to improve outcomes in HIV-DLBCL [23,24,25]. Despite this, rituximab has not generally been routinely used for first-line treatment of HIV-DLBCL in our setting owing to resource constraints and lingering concerns about the compounding immunosuppressive effect of this drug. Here, we showed improved survival among patients treated with rituximab either in their first- or second-line chemotherapy protocols, with no significant difference in survival between patients treated with or without rituximab upfront. This finding suggests that rituximab can potentially be safely reserved for patients with HIV-DLBCL without a complete response to CHOP-based therapy in the resource-constrained environment. We have also shown a significant relationship between the response to rituximab and the number of macrophages in the TME. This agrees with the findings of Riihijarvi et al., who reported that increased numbers of TAMs identified a sub-set of patients with particularly poor outcomes if treated with multi-agent chemotherapy alone, but with significantly better outcomes compared to patients with low TAM numbers when rituximab was included in the therapy protocol [5]. In addition, we found that this association related particularly to the number of M2 macrophages in the TME, suggesting that the efficacy of rituximab is particularly dependent on these cells. This hypothesis is supported by the findings of Leidi et al., who showed a 2–3 fold greater phagocytic capacity of rituximab-opsonised leukaemic targets by M2 macrophages as compared to M1 macrophages in vitro, which was further amplified by the addition of IL10 [26]. This beneficial effect could then explain the lack of the anticipated negative prognostic association with M2 macrophage enrichment.

Unfortunately, macrophage enumeration by automated cell scanners is plagued with difficulties owing to the irregular shape and inhomogeneous staining of these cells and is therefore best performed manually [27]. As such assessment of macrophage enrichment is laborious and somewhat subjective and is consequently not routinely performed. Working on the hypothesis that macrophages would prove to be prognostically important in HIV-DLBCL, we also measured a number of potential peripheral blood biomarkers of macrophage numbers (including several cytokines, monocyte fluorescence and HLA-DR expression, Treg numbers, IDO activity, immune cell numbers (including monocytes, lymphocytes and neutrophils), ferritin and CRP levels). Of these, the only marker which showed any relationship to M1 macrophage numbers was the serum ferritin level, which showed modest performance as a biomarker for low M1 macrophage number prediction (AUC 0.60, sensitivity 55.2%, specificity 62.8%) at a level of > 374 µg/L. Among patients with lower risk features (IPI < 4 and a CD4 count > 150 cells/µl), those with a ferritin level > 374 µg/L had significantly shorter median survival times. Moreover, since ferritin levels were positively correlated with M2 macrophage numbers, we assessed the value of the serum ferritin in predicting the response to rituximab therapy and demonstrated a significant benefit from this agent among patients with high, but not low, serum ferritin levels. Rituximab therapy could potentially be reserved for patients with high ferritin levels in resource-restrained clinical settings and could improve outcomes among patients with high ferritin levels if used upfront. These findings emphasise the value of the serum ferritin level, an inexpensive and readily available blood test, to refine prognostic stratification and therapy selection in HIV-DLBCL, incorporating a biomarker of an important aspect of the TME in a logistically feasible manner. Further studies validating these findings are however required.

Also of interest is that a number of preclinical studies have evaluated strategies to promote macrophage repolarisation to a pro-inflammatory (M1) phenotype, including the use of iron supplementation, genetic reprogramming of M2 macrophages using injectable mRNAs, oxygen-producing nanoparticles, modulation of glutamine metabolism, glycolytic inhibitors and Toll-like receptor agonists (among others) [28]. Such strategies may prove useful in boosting M1 macrophage numbers in HIV-DLBCL in future.

Important limitations to this study include the small number of HIV-negative patients included as a comparator group, the incomplete data in a proportion of the patients and the somewhat subjective nature of manual macrophage enumeration.