Acquired immunological imbalance after surgery with cardiopulmonary bypass due to epigenetic over-activation of PU.1/M-CSF

Systemic inflammatory response syndrome (SIRS) after cardiopulmonary bypass procedure (CPB) broadly resembles the inflammatory responses seen in critical care illnesses and the postoperative period of extensive surgeries [1‐6]. Moreover, the increase in immunoaberrance related morbidities has been shown to persist for many years after the initial insult, suggesting the acquisition of altered long-term immunostasis during severe insults [6‐9].

Monocyte (MO) abnormalities are linked to susceptibility to various infections, the resurgence of opportunistic diseases, accelerated atherosclerosis, and premature failure of grafted coronary vessels during recovery from critical illness [1‐3, 10‐14]. Some attribute the aberrant MO performance to the persevered acquisition of new characteristics, but the evidence regarding the duration of these characteristics are missing. Moreover, heterogeneity of human MO is difficult to interpret in clinical settings due to lack of absolute flow cytometric markers, diversity in function and differentiation, and interspecies differences [15, 16]. A variety of MO subpopulations have been described such as steady-state, naïve, activated, patrolling, tolerogenic, alternatively activated, and atypical [11, 13, 17‐20]. Functional plasticity allows MO to become M1 macrophages (MΦ) [typically CD16^(high), MRP^(high)], or anti-inflammatory M2 cells [typically CD206^(high)] [11, 12]. Alternatively, a small yet critical number of MO-derived dendritic cells (DC) emerge (MO → DC) quickly in critical care illnesses and facilitate the conversion of initial innate responses into acquired immunity and healing [17, 19, 21]. The emergence of MO subtypes depends on environmental influences, kinases (e.g., suppressor of cytokine signaling [SOCS]), and transcription factors (e.g., PU.1) [11‐13, 15, 16]. The MO heterogeneity is the key to their ability to adapt and optimally respond to the challenges. An imbalance of emerging MO subtypes immediately post-insult skews the inflammatory response and increases morbidity [8, 14, 19‐27]. For example, the inadequate emergence of DC from naïve MO results in the predominance of MΦ and overwhelming inflammatory response in trauma and burn patients [20‐22, 25‐27]. Conversely, the excessive emergence of M2, tolerogenic MO, or atypical MO leads to immunosuppression and increases the chance of infections, including cytomegalovirus (CMV) [13, 20, 21]. In patients undergoing cardiac surgery with CPB, the frequency of phagocytic MO is elevated shortly afterward, but an increase in MΦ-like MO may persist [CD64^(high) and CD32^(high)] for 7 days [28]. This suggests that MO can be overcommitted developmentally [29]. Similar findings were observed in patients suffering from critical care insults, trauma, and burns but not a clear timeline for restoration of the immune system to pre-insult was ever defined. The excessive secretion of M-CSF is observed after acute inflammation, and prolonged M-CSF influence is related to MΦ predominance and other immune-aberrancies [13, 30, 31]. Continued M-CSF secretion can be sustained via a positive feedback loop involving either M-CSF and the M-CSF receptor (M-CSFR) or the activation of PU.1 [31, 32]. Consequently, we set up on investigating the activation of PU.1 3 months after surgery as the measure of recovery.

The major goal of this pilot study was to investigate the long-term characteristic of MO in the aftermath of CPB. We hypothesized that (1) CPB would result in the emergence of a new immunological balance in MO characteristics, (2) post-CPB MO abnormalities would be sustained by M-CSF secretion, secondary to PU.1 activation, and (3) these MO abnormalities might be associated with surrogates of unfavorable clinical outcomes.

In the first step, we evaluated characteristics of peripheral blood MO using flow cytometric markers and functional assays to define their developmental characteristics 3 months post-CPB. We found a significant increase in inflammatory CD16^(high) MO, but only immediately after surgery (Fig. 1a). At 3 months, the frequency of CD14^(high)CD16^(high) MO returned to pre-CPB levels (Fig. 1a). We also found that the MO population had increased surface density of CD206, CD163 and TGFβ/LAP while expression of CD86 was diminished (Table 2). A significant increase in SOCS3 positive cells was observed while SOCS1 remained more closely to the pre-CPB level at 3 months post CPB (Fig. 1c, d, Table 2). There was also a significant increase in the phagocytic capacity of the peripheral blood MO obtained 3 months after surgery (which were pre-incubated with zymosan) as compared to the performance of MO before CPB (Fig. 2a). Moreover, the ability of peripheral blood MO to stimulate allogeneic T cells was severely diminished at 3 months after surgery (Fig. 2b). Concomitantly, we noticed that there was a significant decline in the emergence of CD1a⁽⁺⁾ (immature DC marker) and CD83⁽⁺⁾ (mature DC marker) on IL-4 and GM-CSF-differentiated MO even 3 months post-surgery in the CPB cells (Fig. 2c) [12, 13, 20]. IL-4 and GM-CSF-differentiated MO had a significantly depressed ability to stimulate T cells in MLR if the MO were obtained from subject 3 months post-CPB compared to pre-CPB levels (Fig. 2d). The frequency of endogenous BDCA-3/CD11c(+) dendritic cells was also diminished at 3 months post-surgery (%BDCA-3t₀ = 2.14 ± 3.08 vs %BDCA3t_+3m = 0.55 ± .63; p < 0.05).

We found a significant increase in the secretion of M-CSF in response to LPS in supernatants of MO both immediately following and 3 months after surgery (Fig. 3a). A similar trend was seen after stimulation of MO by PMA and ION (Fig. 3b). In contrast, no persistent increase in secretion of IL-1β, IL-6, or IL-10 was seen at 3 months post-CPB after stimulation with LPS (Table 3). The elevated secretion of M-CSF is not due to an increased sensitivity to LPS signals since the surface expression of TLR4 remained unchanged before and after CPB (Fig. 3c).

M-CSF is a cytokine induced by several pathways, but most potently via activation of CD115 (M-CSFR) followed by PU.1 [30, 31]. We found that the percentage of CD115⁺ positive cells were similar at all measured time points (Fig. 4a). M-CSF-R/CD115 surface density for M-CSF was significantly increased shortly after and 3 months after surgery (Fig. 4b). We also found an increase in the frequency of PU.1 positive MO as compared to pre-CPB levels (Fig. 5a–c). Finally, we found that there was a significant decrease in the methylation of the SPI1 promoter fragment flanking the potential transcription start site of the three transcripts (main PU.1 transcript and two other transcripts) at 3 months post-CPB (Fig. 5d).

No clinically-relevant complications (e.g., incidence of sepsis, pneumonia, sternal wound infection, or acute kidney injury) could be appreciated in our study due to the low frequency and pilot nature of the study. Therefore, we focused on surrogates of abnormal function in the immune system. C-reactive protein (CRP), but not other inflammatory markers, were elevated at 3 months but the heterogeneity of the response was apparent (Table 4). Interestingly, the M-CSF production by MO upon stimulation with LPS correlated with serum level of CRP (r ²  = 0.82; p = 0.0062) and SAP (r ²  = 0.61; p = 0.026). Furthermore, the titer of IgG αCMV was increased in our patient population 3 months after surgery (IgG αCMV_t0 = 2.54 ± 1.88 vs IgG αCMV_t+3m = 5.27 ± 2.56; Δx = − 2.73; CI _95% [− 5,32; − 0,13]; p = 0.041; d = − 1.71). The titer of IgM αCMV was unchanged at 3 months after surgery as compared to the baseline (data not shown since the test is qualitative).

This pilot study aimed to provide evidence of a remodeled post-CPB MO milieu 3 months after surgery and elucidate the potential role of the M-CSF/PU.1 system its persistence. Our study shows acquisition of a new MO balance, which was characterized by an increase in MRP8^(high), TGFβ/LAP^(high), CD115^(high), and SOCS3^(high), as well as augmented phagocytic capabilities for at least 3 months after CPB. Decreases in the expression of CD86, the ability of MO to become DC, and T-cell stimulatory capacity were also observed. HLA-DR expression on MO was recovering already at 7 days after CPB but these data are of pilot nature (data not shown). We concluded that 3 months after CPB, peripheral blood MO resemble both MΦ and M2 cells [11, 12, 22]. These characteristics are different from classical or nonclassical MO. We expanded prior research by looking much further after the initial insult and seeking the potential mechanisms [20, 22‐24]. Furthermore, we explored the possible mechanisms involved in sustaining the post-CPB MO milieu. Post-CPB MO produced a large amount of M-CSF, whereas the secretion of other cytokines returned to baseline levels. We demonstrated that positive feedback between M-CSF secretion and M-CSFR is present. Most notably, the overexpression of PU.1, a master factor in the stimulation of M-CSF, was demonstrated at 3 months after CPB [31]. The methylation of the promoter region encoding PU.1 was decreased, suggesting a potential mechanism of post-CPB persistence of acquired characteristics in MO. Finally, we attempted to link the emergence of clinically relevant surrogates of morbidity in our patients despite study not being intended for looking for clinical variables. We noticed signs of immuno-incompetency (increase in serum αCMV IgG) and smoldering inflammation (increase in CRP in some subjects) [33‐35]. Elevation in CRP was highly correlating with the M-CSF production by MO. The small sample size and observed heterogeneity of the inflammatory markers 3 months post-CPB, warrants a larger and more focused study.

The process of immune system response has distinctive time evolution. The immediate innate response is modulated by circulatory MO (among others) [5, 6, 36]. At later stages, acquired immunity becomes dominant, and DC are pivotal regulators [14, 19, 21]. Inflammation is then extinguished via the compensatory anti-inflammatory response syndrome. Along with this trajectory, the MO population characteristics change over time, and their plasticity is critical. Under optimal conditions, the inflammatory MO emerge quickly from peripheral blood MO and are replaced by tissue repair MΦ and other components of acquired immunity, but the exact duration of this process is unclear in subjects recovering from surgery or critical care illness [6, 11, 13]. In our study, we attempted to characterize the MO population at 3 months after surgery at the time when most patients are in recovery stages. We noticed that flow cytometric and functional features of circulating MO resembled deactivated MO, MΦ, and M2 cells [11, 13]. Persistent expression of SOCS3 and SOCS1 at 3 months suggests persistence of both the M2 and M1 phenotype respectively [11, 12, 32, 36]. An increased ability to phagocytose, a decreased ability to become dendritic cells, and the poor ability to circulate MO to stimulate T cells highly suggest that circulating MO resemble deactivated MΦ more than activated MO [11, 13, 22]. For future research, we into to employ RNA-seq to more definitely establish MO features considering that transcriptional markers are more definitive, though more costly than flow cytometry markers [12, 31, 32]. We demonstrated that certain newly-acquired characteristics of MO persisted over a 3-month period, which is well in the recovery period, in contrast to other short-term studies [2, 24, 26].

At 3 months acute inflammation should resolve, and MO should participate in tissue healing. Cytokines, especially M-CSF, modulate MO differentiation and function [11‐13]. In contrast to other immune-inhibitory cytokines, M-CSF secretion is supported by a positive feedback loop [30, 37‐39]. This positive feedback loop relies on PU.1 activation by secreted M-CSF and can reduce the ability of MO to stimulate T cells, trigger increases in atherosclerotic markers, inhibit the MO → DC process, augment the MO → MΦ process, and directly induce immune anergy [22, 30, 34, 36, 40]. Thus, M-CSF is one potential mechanism for long-term, post-critical care insult immunosuppression [6, 36]. We observed increased secretion of M-CSF but no other cytokines 3 months after surgery, suggesting a reprogramming of cells in their response to pathogen pattern recognition [6, 7, 36]. This cellular reprogramming was not related to LPS sensitivity, as the surface expression of TLR4 was unaltered. Moreover, M-CSF secretion increased even after stimulation with agents directly affecting post-receptor mechanisms, suggesting that the process does not involve the aberration of membrane receptors.

PU.1 is the main promoter involved in M-CSF activation [31, 32]. We found an elevation of PU.1 associated with several MΦ characteristics in MO obtained at 3 months after CPB. PU.1 is a critical regulator of MO function in the bone marrow and periphery under resting and stress conditions [31]. Under resting conditions, PU.1 supports hematopoiesis while having anti-inflammatory properties in the cells outside the bone marrow [30, 37]. It also encourages the differentiation of MO into MAC while inhibiting the emergence of DC [21, 22]. These effects are mostly achieved via M-CSF [21, 30, 37]. Interestingly, demethylation of the PU.1-encoding SPI1 promoter was observed, and it was associated with increased expression of the protein level of PU.1. Epigenetic modulation of PU.1 activity can easily sustain the secretion of M-CSF, because the SPI1 expression is no longer inhibited.

Though our study was not powered to look for statistical differences, we showed some increase in markers of opportunistic infection [1, 33, 34]. Elevation of CMV IgG titer was present and is most likely secondary to re-activation of latent infection not the acquisition of a new infection [7, 36]. However, small population size and heterogeneity of response caution against any conclusion but it revealed a potential direction to further develop this study. Also, the deactivation of a viral infection can be secondary to the dysfunction of other leukocyte populations, emergence of myeloid-derived suppressor cells, or may occur independently of MO dysfunction. A similar cautionary approach should be used to observed elevation of CRP suggesting smoldering inflammation and increase in atherosclerotic risk [35, 40, 41].

Experimental design of our study was aimed at increased robustness despite its pilot nature [4, 5, 36]. Utilizing CPB to study the long-term consequences of severe surgery allowed for an assessment of the pre-stress immunological make-up. We also significantly reduced the effect of high inter-individual heterogeneity inherent to the study of immune system function [42]. Despite the small sample size, the longitudinal design of this study further added to its quality. Our standardized protocol minimized hospital practice-related variations. There was > 90% post-CPB compliance with medications that can affect the immune system, including statins, aspirin, and narcotics.

However, there were several limitations as well. First, the pre-operative use of statins, tobacco, and hypo/hyperglycemic agents was frequent in our study population; these agents are known to affect the performance of the immune system under stress. Second, the small sample size limited the measurement of certain clinical outcomes, such as pneumonia and kidney failure. Third, we did not account for MO population shift in depth. The depression of T cells proliferation in response to deficient MO could account for an emergence of alternatively activated MO, myeloid-derived suppressor cells or monocyte anergy [43]. Fourth, although CPB is a good example of severe critical care stress, other clinical illnesses have distinct features which may play a post-insult role. It is unclear whether the CPB-related stress was more severe than the cardiac surgery-related, anesthesia-related stress or the observed finding are typical to any several enough insult to the immune system. Finally, despite the designed our study in longitudinal fashion it remains to be seen if the increase expression of MO-secreted M-CSF is incited by changes in methylation level of SPII, or the acquisition of these features is part of broader tendency to acquire macrophage characteristic after severe insult.

In summary, we demonstrated the persistence of post-CPB MO alterations as late as 3 months after heart surgery [1, 6, 11, 12]. Second, we found a concomitant over-activation of M-CSF/PU.1 with epigenetic changes in the critical promoter region of a PU.1-encoding gene in recovering patients. The clinical significance of the study is that despite apparent recovery from heart surgery, the immunological aftermath continues for at least 3 months and potentially longer. Though our sample size was not powered to uncover the clinical correlates, we found evidence of newly-acquired immunosuppression and elevation of markers in atherosclerosis concomitant with the PU.1/MCSF activation pathway.

The next step is to characterize the persistence of post-surgical immune aberrancies beyond 3 months, investigate other epigenetic mechanisms involved in maintenance of the newly-acquired immunostasis following CPB, and selectively targeting M-CSF/PU.1 activity to see whether the recovery of pre-insult MO function can be achieved. Finally, investigating the long-term activation pattern of MO and M-CSF/PU.1 in other clinical scenarios will define if the described mechanism is specific to CPB or of more universal nature.