Antimicrobial Therapy in the Context of the Damage-Response Framework: the Prospect of Optimizing Therapy by Reducing Host Damage

By design, antimicrobial agents act directly on microbial targets. These drugs aim to eliminate microbes and are remarkably effective against susceptible organisms. Nonetheless, some patients succumb to infectious diseases despite appropriate antimicrobial therapy. Today, with very few exceptions, physicians select antimicrobial therapy based on its activity against the targeted organism without consideration of how the regimen affects patients’ immune responses.

ABSTRACT

By design, antimicrobial agents act directly on microbial targets. These drugs aim to eliminate microbes and are remarkably effective against susceptible organisms. Nonetheless, some patients succumb to infectious diseases despite appropriate antimicrobial therapy. Today, with very few exceptions, physicians select antimicrobial therapy based on its activity against the targeted organism without consideration of how the regimen affects patients’ immune responses. An important concept to emerge in the past few decades is that immune responses to microbes can be detrimental by enhancing host damage, which can translate into clinical disease. A central tenet of the damage-response framework (DRF) of microbial pathogenesis is that the relevant outcome of host-microbe interaction is the damage that occurs in the host, which can be due to microbial factors, host factors, or both. Given that host damage can make patients sick, reducing it should be a goal of treating infectious diseases. Inflammation and damage that stem from the host response to an infectious disease can increase during therapy with some antimicrobial agents and decrease during therapy with others. When a patient cannot eliminate a microbe with their own immune response, antimicrobial therapy is essential for microbial elimination, and yet it can affect the inflammatory response. In this essay, we discuss antimicrobial therapy in the context of the DRF and propose that consideration of the DRF may help tailor therapy to a patient’s need to augment or reduce inflammation.

INTRODUCTION

Since the advent of effective antimicrobial therapy in the mid-20th century, the field of infectious diseases has focused primarily on therapeutic strategies that kill the microbe. Nonetheless, historically, the original form of antimicrobial therapy was specific immunoglobulin, a host product that enhanced host immunity, and today, host-directed therapies are garnering increased attention (1). The rationale for host-directed approaches to antimicrobial therapy is 2-fold: (i) they can augment the host immune response and enhance microbial clearance, and (ii) they can help quell the rising tide of resistance to antimicrobial agents, which act solely on the microbe. Another important way in which host-directed agents can improve the treatment of infectious diseases is by beneficial modulation of the host inflammatory response. This may involve augmentation of the immune response by (i) enhancing microbial phagocytosis and/or killing or (ii) dampening of damaging host responses.

At present, there are very limited data on the ability of either microbe-directed or host-directed therapies to control inflammation stemming from infectious diseases. Steroids are the main agent for which controlled clinical trials have been undertaken. Most of these studies were performed to determine whether adjunctive steroids, used with antibiotics, improved the prognosis of infectious diseases associated with detrimental inflammation. Although some studies were equivocal, others demonstrated a benefit of steroids for reducing short-term mortality in tuberculous meningitis and hearing loss and neurological sequelae in acute bacterial meningitis (2–4). It should be noted that the use of steroids in meningitis stems from landmark preclinical and clinical studies in children (5, 6) demonstrating that reduced inflammation improved prognosis. Steroids also provided a benefit in HIV-infected patients with Pneumocystis pneumonia, although similar results were not obtained in HIV-uninfected patients (7–10). Studies in which steroids were combined with antibiotics for community acquired pneumonia generally revealed a benefit, particularly in severe pneumonia, although steroids may also be associated with a higher rate of readmission (11–14). On the other hand, a recent review of multiple studies concluded there is insufficient evidence to support a benefit of steroids in influenza (15). Older trials with other nonspecific agents, including monoclonal antibodies (MAbs) to tumor necrosis factor (TNF) or activated protein C, intended to dampen the inflammatory response in sepsis, failed to demonstrate prolongation of survival (16), although a recent meta-analysis suggested treatment with a monoclonal antibody to TNF prolonged survival in severe sepsis (17). Numerous uncontrolled, observational studies demonstrate associations between the use of steroids or statins and improved prognosis in patients with influenza and Ebola virus disease (18–20).

The rationale for the use of adjunctive steroids in treating infectious diseases is that they can dampen inflammation, which is a by-product of the normal host response to many microbes. This makes sense and is consistent with a wealth of experimental evidence associating certain inflammatory responses with host damage and clinical disease. The difficulty in obtaining objective evidence for a benefit of these agents in patients is likely to reflect the complex and heterogeneous nature of patients’ inflammatory responses and the possibility that, in some instances, they may inhibit microbial clearance. Historically and today, treatment for most infectious diseases remains focused solely on microbial elimination, even though there is ample evidence that host immunity influences the success of antimicrobial therapy. This became clear at the beginning of the AIDS epidemic, when infectious diseases for which powerful antimicrobial agents were available, such as toxoplasmosis, cryptococcosis, histoplasmosis, and pneumocystis pneumonia were essentially incurable in the setting of the severe immunosuppression of advanced HIV infection. Until the advent of antiretroviral (ARV) therapy, persons with HIV/AIDS required lifelong antimicrobial therapy to control diseases that were curable or that rarely, if ever, occurred in persons with normal immunity. ARVs reduce the HIV viral load and restore immune function by reducing damage to CD4 T cells. However, in some persons, ARV-mediated immune restoration triggers immune reconstitution inflammatory syndrome (IRIS), which is marked by inflammation and tissue damage due to an exuberant immune response to microbial antigens (21, 22).

The occurrence of IRIS, as well as non-ARV-treatment-associated clinical worsening in patients with Pneumocystis pneumonia and those receiving immunosuppressants, highlights the complexities of microbial elimination strategies and their impact on host immunity. Paradoxical clinical worsening and inflammation can also occur after the initiation of appropriate antimicrobial therapy for Gram-negative bacteremia due to lipopolysaccharide (LPS) release, as well as syphilis, tuberculosis, typhoid fever, and malaria in healthy, apparently normal persons (23–27). Clinicians have long considered the possibility of paradoxical worsening, namely, Herxheimer reactions, in treatment of syphilis (27). In addition, since its discovery, the possibility of HIV-associated IRIS is considered in determining when to initiate ARVs in HIV-infected patients treated for cryptococcal meningitis and tuberculosis (28–30). Clinical trials of steroid therapy to prevent IRIS in HIV-infected patients demonstrated a benefit in preventing tuberculosis-associated IRIS (31). Notably, steroid therapy reduced inflammation in tuberculosis-associated IRIS in HIV-infected patients with and without a polymorphism associated with severe IRIS. On the other hand, steroids were not beneficial in HIV-associated cryptococcus-associated IRIS (32), in part because they may have impaired fungal clearance. These examples underscore the dichotomous effects that steroids may sometimes have, at times suppressing microbial clearance, at times suppressing the inflammatory response, at times to the benefit, and at times to the detriment of the host. The host inflammatory response is also likely to affect the efficacy of antimicrobial therapy for other infectious diseases. However, this possibility is seldom considered in therapeutic selection or assessment of the clinical course of patients on antimicrobial therapy. A major roadblock to these considerations is the absence of practical means to incorporate the role of the host into the outcome of therapy.

DAMAGE-RESPONSE FRAMEWORK

The damage-response framework (DRF) was first proposed in 1999 (33) and was further developed in a series of subsequent studies (34–37). It is a theory that accounts for the contributions of both the host and the microbe to the outcome of host-microbe interactions in infectious diseases. In the 2 decades since it was first proposed, the DRF has provided a flexible and dynamic construct to understand microbial pathogenesis and infectious diseases. In the DRF, host-microbe interaction is depicted by a parabola that plots host damage on the y axis as a function of the host immune response on the x-axis. This simple schema demystifies myriad outcomes of complex infectious diseases. For example, the DRF parabola makes it possible to understand how the same microbe, such as Cryptococcus neoformans, could cause disease in a setting of overly exuberant, as well as severely impaired, immune responses (38). It can also account for why certain influenza strains may cause severe disease in young, immunocompetent persons with robust inflammatory responses and why certain infectious diseases occur in patients with certain immune defects (39). This problem underscores the need to consider the role of host inflammation in the course of antimicrobial therapy for infectious diseases. The DRF provides a theoretical construct to analyze the choice and efficacy of antimicrobial therapy in a new light from the usual considerations, which are usually limited to drug coverage and microbial susceptibility.

ANTIMICROBIAL THERAPY IN THE CONTEXT OF THE DRF

Antimicrobial therapy in individuals with serious life-threatening infectious diseases is lifesaving because it eliminates (or controls) the microbe. For many infectious diseases, a reduction in the microbial burden leads to a reduction in host damage. Interestingly, powerful antimicrobial agents often fail to make patients feel better as quickly as they eliminate their microbial targets because fever, leukocytosis, and symptoms stemming from host inflammation and/or microbial damage may resolve slowly. Persistent fever and/or leukocytosis can prompt escalation of antimicrobial therapy and contribute to antibiotic overuse and misuse, fueling resistance. In some instances, antimicrobial agents can induce or increase an already overexuberant inflammatory response. For example, penicillin can induce inflammation and damage via microbial lysis and the release of TLR2 ligands (40). In fact, bacterial killing and lysis are linked to excessive inflammation via release of proinflammatory molecules, including LPS, and more recently to inflammasome activation (41). The risk of exacerbating inflammatory responses led to consideration of the use of immune globulin, which may reduce inflammation by Fc receptor-mediated interactions, in the treatment of streptococcal toxic shock (42–44). However, it has been difficult to establish a clear benefit of immune globulin in clinical trials, perhaps because patients’ inflammatory responses fall on different parts of the DRF parabola, and the amount of “beneficial” immunoglobulin needed to stem the inflammatory response is unknown. Another approach to dampening antibiotic-induced inflammation is the use of combination therapy with a cell wall and protein synthesis inhibitor (see below). Treating infectious diseases in patients with inflammatory conditions, sepsis, and immunocompromised states could be optimized by knowing where individual patients’ immune responses are located on the DRF parabola before and during antimicrobial therapy and the status of their microbial burden (Fig. 1).

FIG 1.

Effects of antimicrobial therapy on the immune response of patients with weak, strong, or no perturbation in their immune response prior to therapy. Antimicrobial therapy can increase or decrease host damage by increasing or decreasing the host response. The point that depicts a patient’s position on the DRF parabola can be concordant (increase in damage or increase in immune response) or discordant (increase in damage, decrease in immune response, or decrease in damage, increase in immune response) depending on the effect of antimicrobial therapy on the patient’s pretreatment immune status. Vertical green lines signify a reduction in damage. Horizontal green lines signify a decrease in the immune response. Vertical (or diagonal) red lines signify an increase in damage. Horizontal red lines signify an increase in the immune response. The orange squares denote patients with weak (W), strong (S), or no perturbation (NP) in their immune responses prior to the initiation of antimicrobial therapy. These points reflect the damage stemming from individual patients’ immune responses to infectious disease-causing microbes.

ANTIBIOTICS AS IMMUNE MODULATORS

Although antimicrobial drugs are not usually viewed as immune modulating agents, some affect the host immune response. For example, amphotericin B can activate Toll-like receptors (TLRs) (45), and macrolides can inhibit cytokine responses (46), resulting in pro- and anti-inflammatory responses, respectively. Other agents have indirect effects on the immune response that result from microbial lysis and the release of components that interact with the immune system. This occurs with penicillin (40) and other antibiotics that inhibit cell wall synthesis (41) and induce bacterial lysis. On the other hand, clindamycin and other protein synthesis inhibitors mediate their antimicrobial effect through interference with prokaryotic protein synthesis and do not mediate lysis or the release of microbial compounds. The efficacy of beta-lactams can be reduced by the Eagle effect (47), which occurs with drugs that inhibit cell wall synthesis in the setting of a large bacterial burden and/or stationary-phase growth. The fact that the Eagle effect does not occur with protein synthesis inhibitors (47), which also have the potential to inhibit toxin/enzyme production, led to the current recommendation to use a beta-lactam and a protein synthesis inhibitor for group A streptococcal infection (47, 48). Notably, a comparison of clindamycin to beta-lactams and aminoglycosides revealed no discernible effects on cell-mediated or humoral immunity (49), although clindamycin can reduce the efficacy of phagocytosis and was detrimental in experimental feline toxoplasmosis via effects on immune function (50). In addition, some have called for caution and more study when combination therapy, rather than monotherapy, is used for skin and soft tissue infections (51). An Eagle-like effect with echinocandins has been described (52).

Macrolides are antimicrobial agents with the potential to reduce the immune response. In addition to their direct antimicrobial effect, macrolides also have nonantimicrobial activity that dampens the inflammatory response, most likely by blocking the activation of nuclear factor kappa-light chain enhancer of activated B cells (NF-κB), and can ameliorate clinical symptoms in asthma (53). Thus, macrolides reduce inflammation (46) and host damage by shifting the x axis of the DRF parabola to the left, resulting in reduced damage on the y axis (Fig. 1). Therefore, they may reduce the microbial burden due to antimicrobial activity and reduce inflammation due to anti-inflammatory effects. These actions may produce a therapeutic effect that owes in part to reduced host damage. The DRF predicts that these drugs would function best in individuals with strong or chronic immune responses (Fig. 1), which is consistent with their use in chronic obstructive pulmonary disease and cystic fibrosis. However, it may make sense to avoid them in those in whom a reduction in the inflammatory response may be detrimental, e.g., those with neutropenia or impaired phagocyte activity. In contrast to macrolides, polyenes can increase inflammatory responses by binding to TLRs and triggering inflammatory cascades. This could explain observations of deterioration in patients soon after initiation of antibiotics that lyse bacteria (2, 25, 54) and the notorious infusion related side effects observed with intravenous administration of amphotericin B. Hence, a reduced microbial burden from direct microbial killing may be accompanied by an enhanced inflammatory response and tissue damage. Thus, as shown in Fig. 1, it makes sense that patients’ overall responses are likely to reflect a combination of the initial host response, direct drug action on the microbe, and direct and/or indirect drug effects on the host response, making the “net” effect of antimicrobial therapy equal to a combination of drug-against-bug activity and host immune response.

The adjunctive use of immune dampening agents such as steroids is largely pragmatic, perhaps supported by meta-analyses, as described above, and population data. In many of these instances, the microbial burden is presumed to be high. However, quantitative microbial cultures are rarely performed, the microbial burden is usually unknown, and there is a paucity of research to identify mechanisms by which immune dampening or enhancement may confer a therapeutic benefit. We submit that the DRF provides a theoretical framework to determine how antimicrobial therapy works, including its effect(s) on host inflammation, and that this can help design new clinical trials and identify patients that could benefit by interventions that modify their immune responses. More data on the effects of different types of drugs on the immune system can make the selection of antimicrobial therapy more rational. This would provide much needed help to clinicians in their efforts to optimize and make antimicrobial therapy shorter, safer, and less costly. Importantly, such data may also help clinicians refrain from escalating antibiotic regimens when patients continue to exhibit inflammatory manifestations while on appropriate therapy by arming them with the understanding that certain agents can increase the inflammatory response.

Although more than 7 decades of experience show that antimicrobial agent-induced microbial killing usually cures infectious diseases, it is not known how much the microbial burden needs to decrease for antimicrobial therapy to be effective or whether and how the host response affects therapeutic efficacy. This is important because antimicrobial agents are often given for prolonged periods, even after presumed microbial clearance. The movement to shorten antimicrobial therapy is gaining traction (55), further underscoring the gap in our knowledge of how much of an antimicrobial agent is needed to achieve a clinical cure. On the other hand, even when antimicrobial agents are highly active and eliminate their targets, some patients experience significant morbidity and mortality that often reflects irreversible host damage. Hence, there is considerable room for improving current antimicrobial therapies by interventions that maximally reduce host damage by manipulating the position of the patient on the DRF parabola. This is a logical, yet underdeveloped and underutilized option for which some tools may already be available.

USING THE DRF TO GUIDE ANTIMICROBIAL THERAPY

While measures of host response are not part of routine clinical care, certain such measures, including C-reactive protein (CRP) and procalcitonin, have informed decision making in clinical trials on length of therapy, intravenous to oral therapy switch in osteomyelitis, and antibiotic deescalation for pneumonia, although the evidence for the latter is conflicting (56–59). The benefit of steroid therapy, administered before or with initiation of antibiotic therapy for acute bacterial meningitis, is believed to associate with a reduction in cerebrospinal fluid (CSF) proinflammatory cytokines; however, large-scale data to support this assertion are lacking and cytokines are not measured in routine clinical care. To this point, a recent study showed that combination therapy with a beta-lactam and protein synthesis inhibitors reduced cochlear damage in experimental pneumococcal meningitis, and another suggested such combinations are superior to adjunctive steroids (60, 61), highlighting the potential utility of measures of inflammation, such as serum and/or central nervous system (CNS) cytokines/chemokines, in clinical care. Along the same lines, a recent study linked changes in T cell phenotypes, also not part of routine care, to the response to therapy for tuberculosis (80), providing more evidence of the complex relationship between antimicrobial therapy and host immunity.

For meningitis, pneumonia, and other infectious diseases where clinical signs and symptoms owe to inflammation, choosing or using drugs that dampen the immune response makes sense. Consistent with this idea, an anti-inflammatory cytokine, interleukin-10, improved the efficacy of ceftriaxone treatment of experimental pneumococcal pneumonia (62). Although the foregoing is an older study, it is important. Mortality in the first 5 days of pneumococcal pneumonia is ∼50%, largely unchanged from the preantibiotic era, despite powerful antibiotics and advances in supportive care (63–65). Pneumococcal pneumonia is a disease for which the inflammatory response is a major contributor to lung damage, and the beta-lactam drugs used to treat this disease are known to lead to the release of bacterial inflammatory mediators (41). The addition of a macrolide to beta-lactam therapy can improve the prognosis of pneumococcal pneumonia (66), and several meta-analyses showed either a mortality benefit or decreased length of hospital stay in patients with severe pneumonia who received steroids (13, 14, 67). Therapies that reduce host inflammatory damage may be needed to further improve the prognosis of severe community-acquired pneumonia. A prominent editorial endorsed a practice change for hospitalized patients with this condition but warned that steroids can also be associated with treatment failure and suggested the path forward requires more information on the host inflammatory response (68).

Returning to the example provided by cryptococcosis, a disease for which inflammation has been extensively studied, the failure of adjunctive steroids to ameliorate IRIS in cryptococcal meningitis was linked to reduced CNS proinflammatory cytokine levels and impaired fungal clearance (32, 69). Although at present there is no evidence-based way to identify patients in whom steroids will have this effect, a machine learning approach could reveal a constellation of host features that associate with safe use of steroids or cytokines. In fact, associations between CSF cytokines and fungal clearance in HIV-associated cryptococcal meningitis (69, 70) support this approach. The DRF illustrates how the best therapy for patients with cryptococcosis could be selected (38, 71) (Fig. 1). For example, amphotericin B may be most effective in the setting of an insufficient inflammatory response but may be deleterious in patients with robust responses, e.g., immunocompetent patients. Hence, the current recommendation of amphotericin B plus 5-fluorocytosine may be best for severely immunosuppressed patients, but this may enhance the risk of inflammation in immunologically intact patients who can suffer from intractable CNS inflammation (72). Other antifungal agents can also affect the immune response. Different lipid amphotericin B formulations have independent anti- or proinflammatory effects, azoles can enhance the activity of phagocytes, and echinocandins can enhance antifungal activity by unmasking microbial targets on fungal cell walls (73, 74). Hence, it may be that antifungal drugs with pro- and anti-inflammatory (or neutral) properties are best for cryptococcosis in the setting of impaired or competent immunity, respectively. This hypothesis could be tested with a machine learning approach. Data obtained could be used to design new clinical trials of pro- and anti-inflammatory agents in the treatment of cryptococcal and other fungal diseases that occur in those with strong or weak inflammatory responses.

Identification of novel biomarkers of inflammation will require further investigative efforts. Currently available data platforms, such as machine learning models, make it possible to link known clinical measures (e.g., blood pressure, pulse, and fever), biomarkers (e.g., CRP, procalcitonin, serum, CSF, and tissue cytokines), and routine measures (e.g., hematocrit, platelets, white blood cell count, and albumin) to the efficacy of antimicrobial agents, adjunctive steroids, or other interventions to known clinical measures. Use of machine learning to gain insight from large data sets may identify a constellation of findings to guide clinical decision-making and study design. Pragmatic clinical trials could then be designed to identify links between antibiotics, adjunctive steroids, and clinical endpoints (e.g., mortality, length of stay, and days of fever) for patients with pneumonia and meningitis. This may be even more important for influenza, which can threaten large numbers of people during an epidemic.

Despite the promise of machine learning and precision medicine for tailoring antimicrobial and adjunctive pro- or anti-inflammatory therapy to individual patients by shifting the position of the patient along the x axis of the DRF parabola (Fig. 1), non-microbe-specific agents such as steroids and cytokines will always associate with the risk of untoward effects. Furthermore, they remain understudied for infectious diseases. Although cogent arguments have been made to use such agents in epidemics such as influenza or Ebola virus disease (18–20, 75), it is unlikely they will be subjected to large-scale trials or that their debits can be predicted or overcome. In contrast, microbe-specific antibodies hold significant promise for dampening or augmenting host inflammation in the course of antimicrobial therapy. Defined microbe-specific MAbs exerted independent effects on microbial elimination and host inflammation in experimental pneumococcal and Acinetobacter infection models (76–78), and a recent study showed that combination therapy with specific antibody and an antimicrobial agent reduced inflammation and improved therapy of experimental pneumocystis in mice (79). A similar benefit of antibodies on the efficacy of antimicrobial therapy for pneumococcus has also been described (81). For pneumococcus, distinct MAbs that reduce the early lung cytokine response without affecting the bacterial burden and those that mediate early bacterial clearance in mice have each been identified (76). The efficacy of the latter depends on presence of the inhibitory Fc receptor, which regulates inflammation (77). This underscores the potential of antibody engineering to develop molecules with predictable effects on inflammation via Fc receptor binding. Although preclinical data suggest this approach holds promise, it has not been tested in humans.

We conclude this essay by pointing out that when physicians evaluate a patient, they usually do not know the coordinates of their patient’s condition on the DRF parabola. This leads to a continuation of empirical approaches to the use of steroids in clinical situations where it is presumed that inflammation is contributing to a lack of clinical improvement, particularly in cases of pneumonia and meningitis. Hence, there is a need for new diagnostic tools that provide information on the damage incurred by the host and the magnitude of the inflammatory/immune response in real time to assess where individuals with microbial diseases are in the parabola. Such information can lead to new paradigms for use of adjunctive immune modulators in treatment of infectious diseases. As discussed above, the use of machine learning to assemble a profile of clinical and laboratory measures that associate with beneficial and deleterious outcomes when steroids are used in pneumonia, meningitis, and other infectious diseases can generate hypotheses to test in pragmatic trials. This can be done today. In parallel, informed by such data, further research on microbial pathogenesis and the inflammatory response is needed to overcome the dogma that antimicrobial therapy has a singular effect on the microbe without affecting the host. Rational antimicrobial therapy should require knowledge of the immunological and inflammatory state of the host. This will optimize antimicrobial therapy and improve outcomes by focusing on clinical improvement as a complex outcome of host-microbe, microbe-antimicrobial agent, and host-antimicrobial therapy interaction.