|Home | About | Journals | Submit | Contact Us | Français|
A.C. and J.B.d.S. conducted the experiments. A.C. and E.R. wrote the manuscript. All authors contributed to the conception and planning of the experiments, and to critical revision of the manuscript. R-M.W. and E.R. supervised the project.
In sub-Saharan Africa, invasive non-Typhoid Salmonella (NTS) is a common and often fatal complication of Plasmodium falciparum infection. Induction of heme oxygenase-1 (HO-1) mediates tolerance to the cytotoxic effects of heme during malarial hemolysis but might impair resistance to NTS by limiting production of bactericidal reactive oxygen species. We show that co-infection of mice with Plasmodium yoelii 17XNL (Py17XNL) and S. typhimurium causes acute, fatal bacteremia with increased bacterial load; features reproduced by phenylhydrazine hemolysis or hemin administration. S. typhimurium localized predominantly in granulocytes. Py17XNL, phenylhydrazine and hemin caused premature mobilization of granulocytes from bone marrow with a quantitative defect in the oxidative burst. Inhibition of HO by tin protoporphyrin abrogated the impairment of resistance to S. typhimurium by hemolysis. Thus a mechanism of tolerance to one infection, malaria, impairs resistance to another, NTS. Furthermore, HO inhibitors may be useful adjunctive therapy for NTS infection in the context of hemolysis.
NTS bacteremia is the most common cause of community acquired bacteremia in many parts of sub-Saharan Africa1 and NTS co-infection has been associated with increased malaria mortality2. The association of NTS infection with hemolysis is well established in humans with malaria (especially severe malarial anemia)3-4 and sickle cell disease5, and in mice with hemolysis due to rodent malaria infection6-8, treatment with phenylhydrazine or anti-erythrocyte antibodies, or red blood cell enzyme defects 9-11. It has been assumed that hemolysis-induced macrophage dysfunction was responsible for this phenomenon, although there is no direct evidence that macrophages are the primary refuge of NTS in vivo in the context of hemolysis7-8,10.
Hemolysis results in liberation of heme, leading to expression of the inducible isoform of heme oxygenase-1 (HO-1)12 which degrades heme to biliverdin, carbon monoxide and iron13. Heme is pro-oxidant, induces neutrophil migration and activates the neutrophil oxidative burst14-16 but HO-1 (and its products) play an essential cytoprotective role (reviewed in 17) - dramatically demonstrated by the severe susceptibility to oxidative stress in mice and humans with HO-1 deficiency18-20. HO-1 induction has been shown to protect against infectious, inflammatory and hypoxic-ischemic insults in mice (reviewed in 21) and has been linked to modulation of malarial pathogenesis22 and sickle cell disease23. Recently, in mice, induction of HO-1 has been proposed as a tolerance mechanism in severe malaria24-26 and polymicrobial sepsis27: HO-1 reduces heme-mediated tissue damage and enhances survival without reducing pathogen load. An important cytoprotective effect of HO-1, and thus a likely explanation for its ability to confer tolerance, is its ability to limit the production of damaging reactive oxygen species (ROS, reviewed in 17). However, ROS are important for resistance to certain pathogens, including Salmonella28, and this raises the intriguing possibility that tolerance of one pathogen may sometimes come at the price of loss of resistance to another. We hypothesized that liberation of heme by intravascular hemolysis may lead to HO-1 induction and impairment of resistance to NTS, with increased bacterial replication and mortality.
To determine whether heme liberated by hemolysis impairs resistance to NTS infection, we compared survival and bacterial loads following intraperitoneal infection with green fluorescent protein (GFP)-expressing S. enterica serovar Typhimurium 12023 (hereafter referred to as S. typhimurium) in C57BL/6 mice with or without preceding Plasmodium yoelii 17XNL (Py17XNL) co-infection, phenylhydrazine (PHZ) or hemin treatment. Py17XNL infection of C57BL/6 mice causes a self-resolving infection; parasitemia peaks at 20-30% and is accompanied by progressive hemolytic anemia (Fig. 1a). By contrast, PHZ treatment causes acute hemolysis (Fig. 1b). In both cases, plasma heme concentrations are markedly increased and similar to concentrations achieved 12 hours after injection of hemin (Fig. 1c), but without depletion of haptoglobin or hemopexin (Supplementary Fig. 1a,b). Survival of Salmonella-infected mice was dramatically shortened by prior Py17XNL infection, PHZ or hemin treatment (Fig. 1d), and was significantly shorter in PHZ- and hemin-treated mice (16 h) than in Py17XNL co-infected mice (18 h) (P < 0.01, Log Rank Mantel Cox test). PHZ, hemin and Py17XNL did not cause any mortality in the absence of S. typhimurium infection.
Decreased survival of malaria-infected, PHZ- or hemin-treated mice after S. typhimurium infection was accompanied by increased bacterial loads in whole blood, spleen, liver and bone marrow (Fig. 1e) and bacteremia was much more pronounced: immediately before death (i.e. 16-18 h after infection in Py17XNL-infected, PHZ- or hemin-treated mice and 72 h after infection in control mice) bacterial loads in the blood of infected or treated mice were proportionately higher, and bacterial loads in liver and spleen proportionately lower (Fig. 1e) than in control mice (Fig. 1f).
By flow cytometry, we identified GFP+ (Salmonella-containing) cells in blood, spleen and bone marrow. In the blood of Py17NXL-infected mice, and of PHZ- or hemin-treated mice, just before death, Salmonellae were found predominantly in Gr-1Hi cells (Fig. 2a), and were enriched in this cell population compared to control (PBS-treated) mice (Fig. 2b). The proportion of all Gr-1Hi cells in blood, spleen and bone marrow which were GFP+ (Fig. 2c) correlated with the bacterial load determined by culture (Fig. 1e,f). The Gr-1Hi cells were identified as granulocytes (Ly6G+ F4/80−CD115−; Supplementary Fig. 2a,b). Almost all GFP+ cells were CD115− (Fig. 2d); moreover, Salmonella infection caused an increase in the proportion of Gr-1LoCD115− blood leukocytes (Fig. 2e), suggesting that immature granulocytes are mobilized from the bone marrow to the peripheral blood during infection29. In support of this, blood films from Py17XNL-infected and PHZ or hemin-treated mice 18 h after Salmonella infection (Fig. 2f) revealed numerous neutrophils containing S. typhimurium and many of these neutrophils had immature nuclear morphology. In contrast, neutrophils from PBS-treated mice displayed mature nuclear morphology and did not contain Salmonella.
The accumulation of GFP+ bacteria in granulocytes was not simply due to failure of bacterial uptake by monocytes and macrophages, since the proportions of GFP+ cells in the spleen which were either monocytes or macrophages (F4/80LoCD11bHi and F4/80HiCD11bLo respectively) did not differ between Py17XNL infected- or PHZ or hemin-treated mice and those treated with PBS alone (Supplementary Fig. 2c).
Since there is no obvious defect in uptake of S. typhimurium by macrophages and monocytes following hemolysis or hemin treatment, accumulation of S. typhimurium in blood granulocytes may result from impaired bacterial killing or a more permissive intracellular environment for bacterial replication. To investigate these possibilities we isolated CD11b+ cells from blood of Py17XNL-infected and uninfected mice and compared their ability to phagocytose and kill S. typhimurium. Neither flow cytometric analysis of GFP+ cells nor quantitative culture (in a gentamicin protection assay) revealed any differences in rates of phagocytosis of S. typhimurium between neutrophils or monocytes or between cells from malaria-infected or uninfected mice (Fig. 3a,b); the intracellular location of GFP+ bacteria was confirmed by confocal microscopy (Supplementary Fig. 3a,b). However, when cells were lysed after 2 h in the gentamicin protection assay and live bacteria enumerated by culture, live bacterial recovery from cells from Py17XNL-infected mice was significantly higher than from control mice (Fig. 3b) indicating significant impairment of intracellular killing of S. typhimurium by cells from Py17XNL-infected mice.
Since HO-1 reduces the production of ROS30-33, and since phagocyte NADPH oxidase is essential for early resistance to S. typhimurium28, we investigated whether Py17XNL infection impaired the granulocyte oxidative burst. Using oxidation of dihydrorhodamine to its fluorescent derivative rhodamine to assess oxidative burst34, we observed progressive suppression of the PMA-induced oxidative burst of blood granulocytes during infection (Fig. 3c). Granulocytes were not simply refractory to PMA stimulation since PMA-induced degranulation (assessed by surface CD11b expression35) was actually enhanced 14 and 21 days after malaria infection (Fig. 3d). However, blood granulocytes isolated 24 h after PHZ or hemin treatment did not differ from those of control mice in oxidative burst capacity, degranulation, ex vivo phagocytosis or killing of S. typhimurium (Fig. 3c,d and Supplementary Fig. 3c,d).
Accumulation of heme following PHZ-hemolysis or hemin administration is faster than during Py17XNL infection. Heme directly induces neutrophil migration and ROS production16 whilst subsequent HO-1 induction in myeloid cells suppresses neutrophil maturation and ROS production30,33. Since HO-1 is induced in bone marrow by hemolysis26, we wondered whether the chronic hemolysis associated with Py17XNL infection might induce HO-1 in immature bone marrow myeloid cells, suppress their oxidative burst capacity as they mature, and allow gradual accumulation of dysfunctional cells in the circulation, whereas acute hemolysis (induced by PHZ) may both activate the oxidative burst of circulating granulocytes and mobilize functionally immature bone marrow granulocytes, resulting in heterogeneous oxidative burst activity of blood granulocytes (as suggested by Fig. 3c).
In mice, granulocyte maturation in bone marrow is characterized by increasing expression of Gr-129. Gr-1Hi cells are mature neutrophils and Gr-1Lo cells are immature granulocytes and granulocyte progenitors; the Gr-1Int (intermediate) compartment contains a mixture of cell types. Generation of an oxidative burst is restricted to a functionally mature subpopulation of cells36 in the Gr-1Hi, and to a lesser extent Gr-1Int, compartment (Supplementary Fig. 4).
Hemin and PHZ treatment, and Py17XNL infection, all caused striking depletion of Gr-1Hi cells from bone marrow (Fig. 3e and Supplementary Fig. 5). For PHZ and hemin treatment, loss of Gr-1Hi cells from bone marrow was accompanied by an increase in granulocytes in peripheral blood (Supplementary Fig. 5), confirming the effect of free heme in mobilization of granulocytes from bone marow to the periphery. Although the proportion of circulating granulocytes did not increase during Py17XNL infection, granulocyte mobilization may be obscured by an overall increase in leukocyte count or granulocyte redistribution (eg. from blood to the spleen)37. S. typhimurium infection caused granulocyte mobilization in PBS-treated mice, and markedly exacerbated the granulocyte mobilization in Py17XNL-infected and PHZ- or hemin-treated mice (Fig. 3e and Supplementary Fig. 5), consistent with the presence of immature granulocytes in blood (Fig 2e). To confirm that hemolysis and bacterial challenge did indeed result in granulocytes with reduced oxidative burst activity entering the circulation, the oxidative burst of circulating granulocytes was assessed. Eight hours after S. typhimurium infection the oxidative burst response to PMA was enhanced in PBS-treated mice (presumably due to priming38) but oxidative burst capacity was dramatically reduced in PHZ treated mice (Fig. 3f).
Finally, we investigated whether maturation of the oxidative burst in bone marrow granulocytes was also impaired, as predicted. On day 14 of Py17XNL infection, and 18 h after PHZ or hemin treatment, there was a clear quantitative defect in the PMA-induced oxidative burst of Gr-1Hi cells (Fig. 3g), evident as an increase in the proportion of cells with low oxidative burst capacity (Fig. 3h) and a decrease in the proportion of cells with high burst capacity, compared to the PBS control.
Together, these data indicate that intravascular heme (released during hemolysis) mobilizes granulocytes from bone marrow and simultaneously impairs development of their oxidative burst. Thus, granulocytes entering the circulation in response to subsequent infection are able to phagocytose S. typhimurium but, due to reduced oxidative burst capacity, fail to kill them, providing instead a niche for bacterial replication and dissemination.
Since the cytoprotective effects of HO-1, and of the heme degradation product, carbon monoxide, have been attributed to inhibition of ROS production17,39, we wondered whether suppression of granulocyte oxidative burst correlated with HO-1 induction during granulopoiesis. As expected25,40, PHZ treatment and Py17XNL infection led to systemic induction of HO-1 (Supplementary Fig. 6a-c). HO-1 was consistently induced in peripheral blood monocytes by Py17XNL, PHZ and hemin, but only (to a modest extent) by hemin in circulating granulocytes and by Py17XNL in the non-myeloid population (Supplementary Fig. 6d).
In bone marrow of untreated-uninfected mice, HO-1 is expressed mainly in F4/80+ cells (Fig. 4a), presumably macrophages and monocytes. However, in hemin and PHZ-treated mice there was a significant increase in the proportion of HO-1+ bone marrow cells, especially in the Gr-1Lo/−F4/80− compartment (Fig. 4a,b). There was a small but significant increase in the proportion of HO-1+ Gr-1Lo F4/80− cells in Py17XNL-infected mice (Fig. 4b) but no overall increase in bone marrow HO-1+ cells, likely due to mobilization of F4/80+ cells from bone marrow to blood and spleen37,41 (Supplementary Fig. 7). The Gr-1lo/−F4/80− compartment contains myeloid progenitors and immature myeloid cells. Since surface markers to positively identify murine myeloblasts and promyelocytes have not yet been defined, we assessed HO-1 expression in the granulocyte macrophage progenitor (GMP) population which is proximal to the myeloblast in the myeloid differentiation pathway42. Py17XNL, PHZ and hemin all caused significant induction of HO-1 even at this very early stage of development (Fig. 4c). Thus, malaria infection, hemolysis and hemin treatment all induce HO-1 expression in the earliest stages of granulocyte development and thereby impair subsequent functional maturation of these cells.
To test the hypothesis that HO-1 impairs resistance of Py17XNL-infected mice to S. typhimurium bacteremia, we pre-treated mice with the competitive HO inhibitor, tin protoporphyrin IX (SnPP). Treatment of Py17XNL-infected mice with SnPP for 48 h prior to S. typhimurium infection reduced bacterial loads in blood, spleen and liver to levels not significantly different from those in PBS treated mice (Fig. 5a). SnPP treatment had no effect on bacterial load in mice without Py17XNL infection (Fig. 5a), or on parasitemia in Py17XNL-infected mice (Fig. 5b) but very effectively prevented accumulation of GFP+ S. typhimurium within granulocytes in Py17XNL-infected mice (Fig. 5c,d). SnPP also partially restored resistance to S. typhimurium when administered 2 h prior to PHZ treatment (Fig. 5a,c,d) and prolonged the survival of Py17XNL or PHZ-treated mice after S. typhimurium infection (Fig. 5e). Cobalt protoporphyrin (CoPP), which induces HO-1 in the absence of hemolysis or free heme43, did not impair resistance to S. typhimurium 16 h post infection (Fig. 5a). Thus, both heme and HO-1 are necessary for impairment of resistance to S. typhimurium caused by Py17XNL or PHZ hemolysis, and inhibition of HO abrogates this effect. Inhibition of HO by SnPP did not restore the oxidative burst of Gr-1Hi bone marrow granulocytes in Py17XNL-infected or PHZ-treated mice (Fig. 5f), presumably due to enhanced mobilization of mature Gr-1Hi cells (Fig. 5g) as a result of greater heme accumulation (Fig. 3e, Supplementary Fig. 5). However, SnPP did reverse the accumulation of granulocytes with low level oxidative burst activity in the bone marrow of Py17XNL-infected and PHZ-treated mice (Fig. 5h), indicating that inhibition of HO restores normal development of the oxidative burst in maturing bone marrow granulocytes.
Understanding the etiology of NTS septicemia in individuals with malaria and other hemolytic disorders may lead to new strategies to reduce morbidity and mortality. To reflect the clinical association between NTS septicemia and severe malarial anemia3-4 we have used a model in which malaria infection causes progressive hemolysis, eventually resulting in severe (but non-lethal) anemia, to assess the impact of S. typhimurium co-infection on disease. We demonstrate that loss of resistance to S. typhimurium requires hemolytic release of cell-free heme and subsequent induction of HO-1, and that inhibition of HO-1 reverses this susceptibility to NTS. Thus, although HO-1 is essential for tolerance to the cytotoxic effects of free heme, reducing disease severity without altering pathogen load24-27, HO-1-mediated tolerance to malaria simultaneously impairs resistance to S. typhimurium.
We propose (Fig 6) that during acute hemolysis, heme triggers immediate mobilisation of granulocytes from bone marrow to blood and generation of ROS16, whilst simultaneously inducing HO-1 in immature myeloid cells and thereby reducing their subsequent oxidative burst capacity30,33, perhaps by limiting the availability of heme for incorporation into NADPH oxidase44. This results in mobilization of a heterogeneous population of granulocytes with varying levels of oxidative burst capacity. During malaria infection however, progressive hemolysis leads to sustained release of free heme which both impairs maturation of oxidative burst capacity of granulocytes in the bone marrow and mobilizes functionally immature granulocytes from bone marrow into the peripheral circulation. Accumulation in peripheral blood of functionally-impaired granulocytes, which phagocytose but are unable to kill bacteria, provides a new niche for bacterial replication and dissemination. In this scenario, HO-1 contributes to impaired resistance to NTS but heme also plays a direct role – either in granulocyte mobilization16 or as a substrate for HO-1. The heme degradation products carbon monoxide, biliverdin and iron may further impair resistance to NTS by reducing production of ROS21 or facilitating bacterial replication45. In contrast, non-heme induction of HO-1 (e.g. by CoPP) may limit available iron for bacterial replication and protect phagocytic cells from apoptosis45-46.
Our observation that hemolysis specifically suppresses the oxidative burst capacity of neutrophils offers a plausible explanation for the particular susceptibility to NTS bacteremia in individuals with hemolysis. Salmonella have evolved to survive and replicate inside mononuclear phagocytes47; hemolysis provides an additional niche for sustained bacterial replication in circulating neutrophils. Our results are also consistent with studies of the cytoprotective role of HO-1 in mice; indeed limitation of the granulocyte oxidative burst could be an important adaptive mechanism to reduce self-damage by ROS during hemolysis and to prevent tissue injury associated with release of heme.
Very few tolerance mechanisms have been clearly identified48, despite recent interest in their therapeutic potential 49. In mice, HO-1 confers tolerance to blood stage malaria24-25 but simultaneously diminishes resistance to malaria parasites developing the liver50. However, in Drosophila, infection-induced anorexia increases tolerance against S. typhimurium but reduces resistance against Listeria monocytogenes 51, indicating that resistance and tolerance mechanisms can be highly pathogen specific and that a mechanism of tolerance to one pathogen can diminish resistance to another. Although it is well recognized that co-infection with different pathogens can enhance disease severity, and in some cases molecular mechanisms have been elucidated52, this study provides the first direct evidence in a mammal of tolerance to one pathogen impairing resistance to another.
To conclude, our findings have a number of important implications. First, they provide an explanation for the susceptibility to NTS bacteremia in malaria and sickle cell disease patients. Second, they imply that tolerance and resistance mechanisms identified from studies of single pathogens may not easily translate to the “real world” where people are simultaneously exposed to multiple pathogens. Specifically, the concept that the cytoprotective effects of HO-1 may be harnessed by administering its products therapeutically in humans without adversely affecting host defence against infection53-54 may not be valid. Third, we have identified a potential adjunct therapy (SnPP) which might enhance resistance to NTS in patients with hemolytic diseases. SnPP has been used experimentally to prevent severe jaundice55 but optimization of treatment would be crucial to avoid impairment of tolerance to heme. The experimental system described here may be a good starting point to assess and optimize such treatments.
This work was supported by a Medical Research Council Clinical Research Training Fellowship (G0701427), and a small grant award from the European Society for Pediatric Infectious Diseases awarded to A.J.C. The authors wish thank D. Holden (Imperial College, London, UK) for provision of GFP-expressing S. typhimurium, and R. Motterlini, S. Baines, H. Kaur, L. King, C. Stanley, R. Gregory, L. McCarthy, K. Couper, J. Hafalla, E. Findlay, and D. Blount for technical advice and assistance.
Detailed methods are provided as Supplementary Methods.
We stored aliquots from a single broth culture of Salmonella enterica serovar Typhimurium 12023 (S. typhimurium), constitutively expressing GFP, in 10% glycerol at −80 °C. For in vitro experiments we opsonized Salmonella in 20% normal mouse serum at 37°C for 30 min prior to inoculation. For quantitation of bacterial loads, we plated 10 fold-dilutions of lysed organ homogenates and whole blood onto LB agar and incubated for 18 hours before counting the number of colony forming units (CFUs).
Animal experimentation conformed with UK Home Office Regulations and was approved by Institutional ethical review (London School of Hygiene and Tropical Medicine). We obtained female, 6-10 week old C57BL/6 mice from Harlan and Charles River, UK, and infected them with Plasmodium yoelii 17X non-Lethal (Py17XNL) by i.p. injection of 105 parasitised red blood cells (pRBCs). We determined parasitemia by examination of Giemsa-stained thin blood smears. We determined erythrocyte counts using a Z2 Coulter particle counter. We induced acute hemolysis by subcutaneous injection of phenylhydrazine (125 μg g−1). We administered hemin by i.p. injection (50 μmol kg−1 per dose) in two doses 12 hours apart. We initiated Salmonella infections by i.p. inoculation of 105 CFU of S. typhimurium on day 15 of Py17XNL infection or 6 h after PHZ or first dose of hemin treatment. We administered SnPP (40 μmol kg−1 per dose) by i.p. injection 48, 24 and 8 h before S. typhimurium infection of Py17XNL infected mice, 2 h before PHZ treatment, or 8 h before S. typhimurium alone. We administered CoPP (10mg kg−1 body weight) i.p. 6 h before S. typhimurium infection. We monitored S. typhimurium infected mice 6 hourly until they displayed signs of illness (clinical stage 2, see Supplementary Methods) and then every 1-2 h until they reached the humane endpoint (clinical stage 4), at which point they were euthanized. The time of euthanasia was used for survival analysis.
Chamber slides were protected from light and air-dried for 2 h before nuclear staining and mounting with DAPI dissolved in confocal matrix. To determine whether bacteria were intracellular or adherent to the cell surface, we assessed cells with overlapping GFP+ S. typhimurium by 0.5 μm interval Z-stack imaging. For light microscopy, we fixed thin blood films with methanol and air dried before staining with May-Grünwald Giemsa stain).
We assessed ex-vivo phagocytosis and killing of S. typhimurium in a gentamicin protection assay and quantified by flow cytometry (staining with APC-conjugated antibody against Gr-1 and PE-Cy7-conjugated antibody against CD11b), confocal microscopy and bacterial culture. We seeded isolated CD11b+ cells at 1×105 per well in flat bottomed 96-well plates or at 2×105 per well in 8-well chamber slides and incubated at 37°C and 5% CO2 for 20 min prior to addition of S. typhimurium at a multiplicity of infection (MOI) of 10 S. typhimurium: 1 CD11b+ cell. We determined phagocytosis at 45 min after infection and killing at 2 h after infection.
We assessed neutrophil oxidative burst using a modification of the flow cytometric assay described by Richardson et al.34 in which dihydrorhodamine 123 is converted to rhodamine. We measured neutrophil degranulation by the percentage increase in the median fluorescence intensity of surface CD11b expression in stimulated versus unstimulated samples.
We determined concentrations of hemin and hemoglobin using Quantichrom Heme and Hemoglobin assay kits. We quantified protein bound plasma heme using the spectrophotometric method of Shinowara and Waters57 and the concentration of plasma hemoglobin using the method of Khan et al.58.
We determined Hmox1 mRNA expression in liver by quantitative RT-PCR. We standardized cDNA expression for each sample using the housekeeping genes Gapdh and Tbp and calculated expression as relative fold change compared to healthy control samples. We measured HO activity in whole liver homogenates after RBC lysis, using the method of Foresti et al. 59. We measured plasma HO-1 by ELISA using a HO-1 Immunoset.
We performed statistical analysis using Graph Pad Prism 5 software. We used an alpha value of 0.05 for all pre-planned statistical analyses. We used the Log Rank Mantel Cox test for survival analysis. We analyzed continuous data which were approximately normally distributed using two-sided unpaired or paired student’s t-test for pairwise comparisons, or one-way ANOVA with post-hoc testing using Dunnett’s multiple comparison test for comparison with the control group, Tukey’s multiple comparison test for comparison between multiple groups, and Bonferroni’s multiple comparison test for comparison between two or more selected pairs. We log10-transformed all data relating to bacterial loads prior to analysis. We compared proportions between groups using Fisher’s exact test.
Declaration of competing financial interests
The authors have no competing interests as defined by Nature Publishing Group, or other interests that might be perceived to influence the results and/or discussion reported in this article.