Immune responses in beta-thalassaemia: heme oxygenase 1 reduces cytokine production and bactericidal activity of human leucocytes

Patients with beta-thalassaemia increase the risk of bacterial infections, particularly Burkholderia pseudomallei (Bp), the causative agent of melioidosis in Thailand. Impaired immune cell functions may be the cause of this susceptibility, but detailed mechanisms have not been defined. In this study, we observed impaired production of IFN-gamma and IL-10 by whole blood from beta-thalassaemia patients upon stimulation with a range of bacteria-derived stimuli. In contrast, IFN-gamma response via TCR and plasma IgG specific for Bp were still intact. Importantly, mRNA expression of heme oxygenase 1 (HO-1), a potential modulator of immune function, was increased in whole blood from beta-thalassaemia patients, either with or without stimulation with Bp in vitro. Induction of HO-1 by hemin or CoPP in vitro reduced production of IFN-gamma and IL-10 from healthy human PBMCs and decreased bacterial clearance activity of whole blood from healthy controls and beta-thalassaemia, while inhibition of HO-1 by SnPP enhanced both functions in healthy controls. These results were confirmed to some extent in purified human monocytes of healthy controls. Our results suggest a mechanism that excess hemin of beta-thalassaemia patients is a significant cause of immune suppression via HO-1 induction and may underlie the susceptibility of these individuals to severe bacterial infection.

. Comparison of demographics, red blood cell (RBC) indices and white blood cell (WBC) parameters between healthy donors without thalassaemia phenotype, β-thalassaemia disease patients (β-thal), and diabetes mellitus individuals (DM). Data was shown as mean ± standard deviation. Statistical significance was determined by using one-way ANOVA with Tukey's multiple comparisons post-test. RBC red blood cell, Hb haemoglobin, HCT haematocrit, MCV mean corpuscular volume, MCH mean corpuscular haemoglobin, MCHC mean corpuscular haemoglobin concentration, ns non-significant.

Parameters
Healthy (n = 20) www.nature.com/scientificreports/ ity that uninfected HbE trait individuals may have some impairment on cytokine production upon stimulation, we compared the cytokine results from uninfected HbE trait individuals with healthy without thalassaemic phenotype; IFN-γ and IL-10 production was unimpaired in response to various microbial stimuli in whole blood samples (Fig. S1). IFN-γ production from β-thalassaemia patients after stimulation with LPS, PFA fixed Bp and FlgK are significantly lower than healthy controls (Fig. 1A). In addition, IFN-γ production from individuals with DM is also lower than healthy controls when stimulated with LPS, fixed Bp and FlgK (Fig. 1A). Reduced levels of IL-10 production from β-thalassaemia patients in response to LPS, fixed Bp and FlgK were seen when compared to healthy controls, but the comparisons between healthy controls versus diabetic individuals do not show a significant difference (Fig. 1B). Furthermore, we classified the patients phenotypically into transfusion dependent thalassaemia (TDT) and non-transfusion dependent thalassaemia (NTDT) based on their clinical severity and transfusion requirement 33 and compared their disease progression and immune (Table S1). Our data in the Table S1 show that 57% of patients with TDT presented splenectomy which was only 22% in NTDT. Moreover, increased ferritin level was demonstrated significantly in TDT compared with NTDT. These results are consistent with a previous report showing severe iron overload that starts early in TDT 34 . In addition, no difference of IFN-γ and IL-10 production was observed between TDT and NTDT patients or between splenectomized and non-splenectomized patients (Fig. S2).

Figure 1.
Reduction of IFN-γ and IL-10 produced from whole blood samples of β-thalassaemia patients upon stimulation. Whole blood samples (adjusted the number of lymphocyte plus monocyte at 1.8 × 10 5 cells) from healthy controls (open circle, n = 13), β-thalassaemia patients (filled circle, n = 15) and diabetes individuals (filled triangle, n = 10) were cultured with medium alone, 10 μg/ml of LPS, 5.4 × 10 6 CFUs PFA fixed Bp or 10 μg/ml FlgK protein for 48 h. IFN-γ (A) and IL-10 (B) production upon stimulation in supernatant were measured by ELISA, and the results were shown as scattered dot plot and line at mean with 95% confidence interval. Statistically significant was analyzed by using one-way ANOVA with Tukey's multiple comparisons post-test; *P < 0.05, **P < 0.01, ***P < 0.001, and ns non-significant. www.nature.com/scientificreports/ T cell response though T cell receptor (TCR) and plasma Bp-binding IgG antibody of β-thalassaemia patients are not reduced compared to healthy controls. We next sought to examine whether responses from T and B lymphocytes of β-thalassaemia patients are similar or different from healthy controls. Whole blood from 9 healthy controls and 7 β-thalassaemia patients were stimulated with TCR antagonists (anti-CD3 and anti-CD28 antibodies) for 48 h and IFN-γ production in the culture supernatant was measured by ELISA. The results revealed a higher IFN-γ production from β-thalassaemia patients compared to healthy controls ( Fig. 2A). Meanwhile, plasma samples from 34 healthy controls and 16 β-thalassaemia patients were measured for Bp-binding IgG antibodies by indirect ELISA. There was no significant difference between Bp-binding IgG antibody levels from healthy controls and β-thalassaemia patients (Fig. 2B). Taken together, this suggests that IFN-γ response via TCR and plasma IgG level from β-thalassaemia patients are intact as healthy controls. Therefore, the causes of alteration in reduction of cytokine production are more likely to be an effect on the other immune functions rather than the adaptive immune responses.
β-Thalassaemia patients have increased expression of HO-1 in whole blood culture with Bp. β-thalassaemia disease is a condition of impairment of β-globin chain synthesis resulting in increased unpaired α-globin chains. The accumulation of excessed α-globin chains is bound to heme which is finally degraded by heme oxygenases (HO), particularly HO-1. To test whether HO-1 expression was altered, blood samples from 7 β-thalassaemia patients and 7 healthy donors were collected for measuring expression of HO-1 mRNA by real-time PCR either with or without in vitro stimulation with PFA fixed Bp. The baseline results reveal that HO-1 expression levels of unstimulated whole blood from β-thalassaemia patients was higher than those of healthy donors (Fig. 3). After Bp stimulation, HO-1 expression was significantly enhanced in all healthy donors and β-thalassaemia patients. Furthermore, in response to stimulation with fixed Bp, whole blood of β-thalassaemia patients showed significantly increased levels of HO-1 expression compared with healthy controls (P < 0.05). These data indicate that β-thalassaemia patients present the greater up-regulation of HO-1 expression against Bp when compared with healthy controls.
Presence of hemin and induction of HO-1 decreased IFN-γ and IL-10 production in response to Bp. To address the hypothesis that HO-1 is responsible for alteration of immune responses to Bp, PBMCs were isolated from 12 healthy donors and HO-1 mRNA expression after pretreatment with chemical HO-1 inducer or inhibitor was determined by real-time PCR. Hemin (ferriprotoporphyrin IX), a derivative chemical from heme, was used as a representative of heme-induced HO-1 expression 35,36 . Application of cobalt protoporphyrin IX (CoPP) as HO-1 inducer and tin protoporphyrin IX (SnPP) as HO-1 inhibitor are widely used in previous researches 29,30,37,38 . Hemin and CoPP induced HO-1 mRNA expression as expected (Fig. S3A, B), while SnPP inhibited HO-1 mRNA expression (Fig. S3C). When cells were pretreated with vehicle control or each HO-1 modulator at various concentrations, reduction of IFN-γ and IL-10 responses were found from both HO-1 inducer, hemin and CoPP, pretreated PBMC in a dose dependent manner (Fig. 4A, B). In contrast, HO-1 inhibitor (SnPP) treatment significantly increased IFN-γ production upon stimulation in a dose dependent manner but did not alter IL-10 production (Fig. 4C). These results support our hypothesis that increasing levels www.nature.com/scientificreports/ of heme in the blood of β-thalassaemia patients leads to a reduction of IFN-γ and IL-10 production in response to infection or stimulation with various types of bacterial stimuli.

Presence of hemin and induction of HO-1 expression resulted in the reduced bacterial killing in vitro.
Next, we simulated the situation of immune responses occurring in the presence of heme or HO-1 inducer or inhibitor. We addressed the effect of the heme and HO-1 expression on the clearance and cytokines production against common causes of bacterial infection in Northeast of Thailand; B. pseudomallei, Escherichia coli, S. typhimurium and Staphylococcus aureus 2 . Whole blood samples of healthy donors or β-thalassaemia patients were pre-treated with medium, hemin, HO-1 inducer (CoPP) or HO-1 inhibitor (SnPP) before being infected with live bacteria and the numbers of surviving bacteria were counted at different time points. In the medium control, bacterial killing efficacy of whole blood from healthy control was significantly more effective than β-thalassaemia patients (Fig. 5A). Induction of HO-1 expression by hemin and CoPP significantly reduced the efficacy of bacterial killing in all bacterial infected whole blood, either from healthy donors or β-thalassaemia patients, regardless of the bacteria species used (Fig. 5B, C). Inhibition of HO-1 expression by SnPP significantly enhanced bacterial killing in whole blood from healthy controls and only slightly increased in β-thalassaemia patients (Fig. 5D). Of note, the delay of bacterial killing activity of whole blood from β-thalassaemia patients was observed, when compared to healthy controls. Meanwhile, bacterial infected whole blood of healthy donors showed significantly induced IFN-γ after treatment with SnPP for 24 h but most of β-thalassaemia patients failed to produce IFN-γ (Fig. S4). We then addressed the effect of HO-1 on bacterial killing by using purified human monocytes from healthy controls. The bacterial killing activities at 3 and 6 h post infection with live Bp plus hemin or CoPP pre-treatment were significantly lower than medium control (Fig. 6A, B), whereas SnPP pre-treated monocytes showed significantly higher killing activities (Fig. 6C). Similarly, infection with other types of bacteria also found similar results, suggesting that expression of HO-1 impairs the efficacy of bacterial killing, in both human whole blood and purified monocytes.

Discussion
β-Thalassaemia is a chronic disease with chronic anaemia caused by mutation in β-globin gene consequently reduced or absent of β-globin chain synthesis resulting to ineffective functions of red blood cells 39 . We observed anaemia in β-thalassaemia patients and DM individuals in this study, especially in β-thalassaemia patients that presented with more severe anaemia. This is consistent with reports that β-thalassaemia is a dyserythropoiesis leading to anaemia 40 , while occurrence of anaemia in DM results from impairment of erythropoietin by the peritubular fibroblasts due to chronic hyperglycemia 41 . Increasing absolute number of total white blood cells and lymphocytes were increased in β-thalassaemia, while DM shown increasing number of monocytes only. These observations are similar as previous reports 42 .
Besides DM, thalassaemia disease is one of the important risk factors of melioidosis3. In this study, the whole blood response to B. pseudomallei infection of both DM individuals and β-thalassaemia patients showed  7) and β-thalassaemia patients (filled circle, n = 7) were analyzed by real-time PCR using GADPH as internal reference gene. Unstimulated whole blood of healthy and β-thalassaemia were analyzed for ΔCt between HO-1 and GADPH before calculated for HO-1 mRNA expression by 2 −ΔCt . PFA fixed Bp stimulated whole blood of healthy and β-thalassaemia were analyzed for ΔCt between HO-1 and GADPH, then ΔΔCt comparing to medium control before calculated for HO-1 mRNA expression by 2 −ΔΔCt . Results were plotted as scattered dot plot and line at median with interquartile range. Statistical analysis was done by using Wilcoxon matched pairs signed rank test or Mann Whitney test; *P < 0.05 and **P < 0.01. www.nature.com/scientificreports/ defects in IFN-γ production. These results in DM are consistent with previously established studies by presenting the reduction of IFN-γ of whole blood from DM against B. pseudomallei in vitro 15 and in PBMC of acute melioidosis patients with DM as underlying condition 43 . These data suggest that both DM individuals and β-thalassaemia patients may have similarly impair T cell immunity in response to B. pseudomallei infection. In contrast, we found that β-thalassaemia also impaired IL-10 production when compared with heathy controls, which is not the case in DM 15 . Some beta-thalassaemia/HbE patients have increased plasma IFN-γ level which possibly imply low-grade inflammation that may be caused by an additional independent factor in determining the severity of the anaemia 44,45 . Low-grade inflammation is associated with the reduction of immune functions and proinflammatory cytokines are increasingly found in chronic diseases including type 2 diabetes mellitus (T2DM) 15,45 . Although the causes of low-grade inflammation in T2DM might differ from thalassaemia patients, the consequence of this phenomenon might be similar. Likewise, we found that restimulation of whole blood from beta-thalassaemia/HbE patients with probable low-grade inflammation presented reduced IFN-γ and IL-10 production and bacterial killing activity which may lead to increased susceptibility to melioidosis. Taking these data together, it may imply that the precise nature of the defects in immune response of β-thalassaemia patients against B. pseudomallei infection differs from DM individuals and further studies are still needed for investigating these mechanisms. Next, we studied cytokine responses upon stimulation with various type of stimuli including; LPS, a component of Gram-negative bacterial cell wall, which activates immune response via TLR4 46 , whole intact bacteria of PFA fixed Bp that activate cytokine production via pathways of innate and adaptive immune responses, finally FlgK protein (BPSL0280), a flagella hook-associated protein that can activate IFN-γ production from T cells 43 . We found a clear reduction of IFN-γ and IL-10 secretion from whole blood of β-thalassaemia patients incubated with LPS, fixed whole intact bacteria of Bp, and FlgK protein compared to healthy controls. Other reports in β-thalassaemia found impairments in innate immune functions of NK cells 21 , neutrophil 22 , monocyte/macrophage 23 , and complement fixation 24 which seems to be similar to diabetes condition 47 . Furthermore, a cohort study of Thai adult patients with thalassaemia reported higher prevalence of infection in transfusion www.nature.com/scientificreports/ dependent thalassaemia (TDT) than non-transfusion dependent thalassaemia (NTDT) 34 . Our results correlated with a previous study reporting no significant difference of IFN-γ production between splenectomized and non-splenectomized β-thalassaemia patients with TDT 19 . Furthermore, splenectomy had benefit probably in some TDT patients, but advanced age and splenectomy were the significant clinical risk factors for most of the disease-related complications in both TDT and NTDT groups 34,48 . Further experiments revealed that activation of IFN-γ production via T cell receptor (TCR) signal by using anti-CD3 and anti-CD28 antibodies stimulated on whole blood from β-thalassaemia patients is higher than healthy controls. This data is consistent with previous reports that β-thalassaemia patients have increased expression of CD3 marker on T cell surface than healthy controls 19 , and also T cell activation is not defective 42 , although dysfunction of T and B cells have been reported in conditions with high levels of ferritin 19,49 . Moreover, concentrations of anti-Bp serum IgG from β-thalassaemia patients was comparable to healthy controls, while total www.nature.com/scientificreports/ antibody levels in serum of β-thalassaemia patients was increasing 50 . Therefore, based on our results, there is no significant general impairment of immune responses through TCR and IgG antibody production against Bp from β-thalassaemia patients in comparison to healthy controls. The hemolytic consequences of β-thalassaemia lead to increasing levels of heme in blood circulation 25 which causes cell death by inducing free radical oxidative species (ROS) [51][52][53] and elicit inflammatory injuries 51,54 . HO-1 is upregulated to catalyze heme into iron, carbon monoxide and bilirubin 55 . Expression of HO-1 is not only for heme catalysis but can also adversely affect immune function. Studies using mouse derived macrophages show increasing level of HO-1 mRNA expression after infected with Bp, and leading to increase serum IL-6, TNF-α and MCP-1, but IFN-γ production was decreased 29 . Here we found increased HO-1 mRNA expression in whole blood from β-thalassaemia patients, which increased further in the presence of Bp in vitro. The presence of increased baseline levels of HO-1 in β-thalassaemia can be explained by the presence of high levels of hemin (> 50 μM) in serum from β-thalassaemia patients, but undetectable in non-thalassaemia individuals 25 . Hemin and products of hemin degradation by HO-1 can have multiple effects on cellular immune functions. CD4 T cell and monocyte coculture in presence of hemin triggers a polarization of CD4 T cells to a regulatory T cell (Treg) subset 56 . A study of carbon monoxide releasing molecule (CORM-2) treated C57BL/6 mice infected with Bp revealed increasing burden of bacteria in lung and increased bacterial load in macrophages 29 . Carbon monoxide is also able to suppress human CD4 T cell proliferation and IL-2 secretion 57 . In addition, in a mouse model HO-1 mediated iron released from catabolism of heme leads to cell damage in thalassaemic erythroblasts 58 . Taken together, all studies lead to our hypothesis that hemin induced HO-1 reaction may explain immune response impairment and increasing of susceptibility to infection.
To further test the effects of HO-1 per se on immune function, we investigated the effect of chemical induced HO-1 activation or inhibition before stimulation of leucocytes from healthy individuals with B. pseudomallei. Treatment of normal human PBMCs with HO-1 activators, hemin and CoPP, caused reduction of IFN-γ and IL-10 after stimulation with B. pseudomallei. In contrast inhibition of HO-1 by using SnPP increased IFN-γ production with no effect on IL-10. This is in comparison with studies in mice showing that HO-1 activation by CoPP impaired dendritic cell maturation, CD4 + and CD8 + T-cell proliferation and IFN-γ production but increased IL-10 in response to LPS in vitro 59 . HO-1 can also inhibit IFN-γ, decrease immune cell numbers and also induce T cell apoptosis of mice 60 .
To investigate the consequences of these changes to anti-bacterial immunity, we measured cell mediated responses induced by live pathogenic bacteria (namely B. pseudomallei, E. coli, S. typhimurium and S. aureus) which cause serious disease in patients with thalassaemia in Thailand 2 . The HO-1 activators, hemin and CoPP impaired bacterial killing activity of human whole blood and purified monocytes from healthy controls, which correlated with low production of IFN-γ, in all bacteria tested. In contrast, the HO-1 inhibitor, SnPP enhanced bacterial killing and resulted in higher levels of IFN-γ compared with medium controls. Interestingly whole blood from β-thalassaemia patients impaired bactericidal activities upon exogenous HO-1 induced by hemin or CoPP, similar to healthy controls. However HO-1 inhibition by SnPP did not enhance bactericidal activity suggesting that high levels of endogenous HO-1 might not be completely inhibited under this condition or there might be other mechanisms to impair bactericidal activity in β-thalassaemia. This issue still needs further investigation. Apart from our observation, there are more reported bacterial clearance mechanisms that HO-1 can be modulated by hemin and strongly inhibited phagocytic activity and migration of human monocyte and neutrophil, which is not correlated with heme-iron catabolism 26 . HO-1 expression can promote M. tuberculosis growth within primary human macrophage cells 30 , and also reduce bacterial killing in mice infected with B. pseudomallei 29 . Induction of HO-1 expression by CoPP combined with LPS can inhibit T cell responses and reduce IFN-γ production by impairing DC maturation and expression of MHC-II, CD40, CD80, and CD86 in mice 59 .
In conclusion, our study demonstrates defective cytokine production of β-thalassaemia patients in response to a diverse range of stimuli derived from pathogenic bacteria. We believe that HO-1, as well as hemin, is primarily responsible for these effects on production of both IFN-γ and IL-10, and also impair the bactericidal activity of human leucocytes against pathogens known to infect individuals with β-thalassaemia. Modulation of the heme/HO-1 pathway may hold promise for future host directed therapies. Protection of macrophage cell lines and human macrophages, by quinine can prevent damaging effects of increasingly endogenous heme following sepsis with gram negative bacteria and restore macrophage phagocytosis in vivo 26 . In mouse models of β-thalassaemia, inhibition of HO-1 by SnPP can decrease heme catabolism and reduce iron release including improved erythropoiesis 61 . Blocking of HO-1 expression may be an alternative therapy for septic β-thalassaemia patients to reduce heme catabolism and enhance proinflammatory cytokines for improved bacterial killing.

Methods
Subjects. This study was approved by Human Ethics Committee at Nakhon Phanom Hospital, IEC-NKP1-No.19/2558. Informed consent was obtained from all participants and in compliance with the Declaration of Helsinki. Healthy donors were defined by blood bank guidelines and had no signs of infection at the time of blood collection. β-thalassaemia patients were previously diagnosed by clinicians, and blood samples were collected before receiving blood transfusion. All donors with mean corpuscular volume (MCV) < 80 fL and dichlorophenol indophenol precipitation (DCIP) positive were confirmed by Hb typing and classified as HbE carriers 62 . Diabetes mellitus (DM) individuals were defined as type 2 DM by clinicians and had fasting blood sugar more than 126 mg/dl 63  In other experiments, PBMCs or CD14 positive cells were isolated from whole blood samples 4 and plated at 2.5 × 10 5 cells/ml, then pretreated for 3 h at 37 °C with vehicle control (0.1 M NaOH; Sigma), a substrate of HO-1 (hemin; Sigma), an HO-1 inducer (cobalt protoporphyrin IX (CoPP); Enzo Life Sciences), or an HO-1 inhibitor (tin protoporphyrin IX (SnPP); Enzo Life Sciences). Sequentially, 7.5 × 10 6 CFUs of PFA fixed Bp (ratio at 30:1) were cultured with pretreated PBMCs or monocytes for 48 h at 37 °C. After 3 h of pre-treatment at 37 °C, cell pellets were collected for HO-1 expression analysis by real-time PCR and after 48 h supernatants were collected for cytokine quantification by ELISA.
Plasma IgG antibody detection. PFA fixed Bp at 10 6 CFUs per well were coated onto 96 wells ELISA plates overnight before adding 1:300 diluted heparinized plasma in duplicate and following our published protocol 64-66 . Measurement of HO-1 mRNA expression by quantitative real-time polymerase chain reaction. Total RNA from whole blood was isolated using Tempus Spin RNA Isolation Kit (Tempus), while total RNA from cell culture pellets was isolated using RNeasy Mini Kit (Qiagen) following the manufacturer's instructions. Reverse transcription of 1 μg total RNA was done by using ImProm-II Reverse Transcription System (Promega). Quantitative real-time PCR was performed by using Rotor gene 3000, with AccuPower 2X GreenstarTM qPCR Master Mix (Bioneer). GADPH was used as internal reference gene 67 . HO-1 amplicons were detected by using forward primer 5′-GCA GAG AAT GCT GAG TTC ATG-3′ and reverse primer 5′-CAC ATC TAT GTG GCC CTG GAG GAG G-3 68 . All assays were performed in duplicate and the results were shown as cycle threshold (Ct).
Bacterial clearance assay. After pretreatment for 3 h with vehicle control, 25 μM hemin, 50 μM CoPP, or 10 μM SnPP, whole blood or monocytes were infected with 10 5 CFUs and 5 × 10 5 CFUs, respectively, of live Bp, E. coli, S. enterica serovar Typhimurium or S. aureus for 1, 3 and 6 h at 37 °C. At each time point, infected cells were lysed with 1% Triton X-100 (Biotech), then total numbers of viable bacteria were counted by colony count on LB agar plate. Data was analyzed as % Bacterial killing = ((inoculum bacteria -remaining bacteria)/inoculum bacteria) × 100. Data analysis. All statistical analysis was done by using Graphpad Prism version 6 (Graphpad software).
One-way ANOVA was applied for independent sample sets, while two-way ANOVA was applied for matched sample sets. Wilcoxon matched-pair's signed rank test was applied for comparison of non-normally distribution data, whereas Mann Whitney test was applied for comparison of independent patients. Statistically significant differences were considered at P < 0.05.