Abstract
Infections account for relevant morbidity and mortality, especially if the cardiovascular system is affected. Clinical manifestations are often unspecific, resulting in a challenging diagnostic work-up. The use of molecular imaging methods, namely [18F]FDG PET and leukocyte scintigraphy, is increasingly recognized in recently published international guidelines. However, these 2 established methods focus on the host’s immune response to the pathogen and are therefore virtually unable to differentiate infection from inflammation. Targeting the microorganism responsible for the infection directly with novel imaging agents is a promising strategy to overcome these limitations. In this review, we discuss clinically approved [18F]FDG PET with its advantages and limitations in cardiovascular infections, followed by new PET-based approaches for the detection of cardiovascular infections by bacteria-specific molecular imaging methods. A multitude of different targeting options has already been preclinically evaluated, but most still lack clinical translation. We give an overview not only on promising tracer candidates for noninvasive molecular imaging of infections but also on issues hampering clinical translation.
Among all human diseases, infections are the third most common cause of mortality and the leading cause of morbidity (1,2). The most common bacterial pathogens of health-care–associated infections are members of the Enterobacteria family, Clostridium difficile, Staphylococcus aureus, and Pseudomonas aeruginosa (3).
Infective endocarditis (IE) and vascular graft and endograft infection (VGEI) are major, predominantly bacterial, cardiovascular infections and are associated with a high morbidity, hospital mortality, and economic burden (4–6). IE can be divided into native valve IE (NVIE), accounting for around 55% of IE patients; prosthetic valve IE (PVIE), for around 30%; and cardiac device–related IE (CDRIE), for around 10% (4).
The clinical presentation of IE and VGEI is influenced by many different factors such as the underlying pathogen, preexisting or congenital heart disease, previous interventional procedures (e.g., dental or gastrointestinal), and the presence or absence of prosthetic valves, cardiac devices, or vascular grafts (6). In IE patients, symptoms are mainly unspecific, and many different symptoms may occur at different times. Up to 90% of patients present with the unspecific symptom of fever, followed by heart murmurs in up to 85% and embolic complications in up to 25% (4,7). Laboratory testing includes elevation of markers of infection such as C-reactive protein, procalcitonin, and leukocytosis, as well as markers of end-organ dysfunction. Many of these parameters are used to assess sepsis severity; however, these are not specific for the diagnosis of IE or VGEI (6–8). Blood cultures are positive in about 80% of cases and are important to specify the infective microorganism (4,9). In IE, the most common bacteria are gram-positive staphylococci followed by oral streptococci and enterococci, whereas gram-negative bacteria account for only a minority of cases (4,10). In VGEI, most pathogens are again gram-positive bacteria, including mainly staphylococci and enterococci, whereas gram-negative bacteria account for about one third of infections (9,11). The pathogen responsible for the infection depends also on the time point after surgical intervention and the site of the vascular graft, resulting in a higher percentage of polymicrobial infections in abdominal VGEI (6,9,11).
In addition to microbiology, imaging is an important approach for the diagnosis of IE and VGEI. In IE, echocardiography is the imaging method of choice in initial work-up and therapy monitoring (7). In 2015, the European Society of Cardiology guideline recommended multislice CT, [18F]FDG PET/CT, and leukocyte scintigraphy in addition to conventional echocardiography (7). The American Heart Association and the American College of Cardiology incorporate [18F]FDG PET/CT and leukocyte scintigraphy into the clinical work-up of IE as well (12). These nuclear medicine imaging techniques have specific advantages, such as the high sensitivity of [18F]FDG PET in PVIE and CDRIE. However, they also suffer from disadvantages, such as the lower sensitivity of [18F]FDG PET/CT, especially in NVIE (13). With radioactive labeling of leukocytes, nuclear imaging becomes more specific, but imaging still shows the cellular response of the immune system rather than the presence and activity of the pathogen (14). Direct imaging of the infecting bacteria could further improve specificity and would allow differentiation between sterile inflammation and bacterial infection. In this review, we therefore focus on current advances in clinically approved [18F]FDG PET, with particular emphasis on the potential clinical need for new bacteria-specific tracers for the imaging of cardiovascular infections, and we provide an overview of potential bacteria-specific targets and corresponding tracer candidates.
CURRENTLY APPROVED IMAGING METHODS
[18F]FDG PET is a sensitive imaging method for the detection of IE and, in addition to standard echocardiography, is recommended by current guidelines (7,12). A current metaanalysis by Wang et al., including 26 studies with 1,358 patients, reported a pooled sensitivity of 74% and a specificity of 88% for [18F]FDG PET in IE. In all subgroups of IE, specificity was above 80%, but sensitivity was significantly lower in NVIE, at 31% compared with 86% and 72% in PVIE and CDRIE (Table 1) (13). Therefore, [18F]FDG PET is not routinely recommended in NVIE given the comparably high diagnostic accuracy of echocardiography (transthoracic echocardiography, NVIE: pooled sensitivity, 0.71; specificity, 0.80 (15); PVIE: pooled sensitivity, 0.29; specificity, 1.00 (16)). In a recent metaanalysis (with only 3 studies), the pooled sensitivity and specificity for leukocyte scintigraphy in IE were 86% and 97% (13). Recent metaanalyses failed to provide pooled sensitivity and specificity for leukocyte scintigraphy in different subgroups of IE, because of limited data (17).
Pooled Sensitivity and Specificity of [18F]FDG PET in Cardiovascular Infections
A major drawback of [18F]FDG PET in IE is the highly variable uptake in the adjacent myocardium, which can complicate diagnosis of NVIE or PVIE. Because glucose uptake in the myocardium is dependent on the overall metabolic state mediated by glucose transporter type 4 and the presence of ubiquitous free fatty acids, patient preparation is a crucial step in imaging IE (18). Further, false-positive results can be due to [18F]FDG uptake in postsurgical inflammation in PVIE or CRDIE. In PVIE, this unspecific uptake occurs particularly in the first months but also up to 1 y after prosthetic valve implantation, as shown in a recently published prospective study (19). The same factors might apply to CRDIE, but data are still lacking. This is a major drawback because the additional use of [18F]FDG PET is recommended in these 2 subgroups for intracardiac assessment but not in NVIE (13). False-negative results, which are more common in NVIE (13), may be related to the smaller size and highly spatially moving nature of the vegetations (20).
In contrast to [18F]FDG PET, leukocyte scintigraphy does not suffer from background myocardial uptake; however, the main limitation of leukocyte scintigraphy in IE is the lower spatial resolution, which results in a limited diagnostic sensitivity, especially in NVIE with small vegetations (21). Therefore, most of the upcoming emerging bacterial tracers are PET tracers offering a method-inherent upgrade in resolution.
[18F]FDG PET in VGEI offers a high sensitivity of 89%–98% with a comparatively lower specificity of 59%–81% (5,22–25). Parallel acquisition of CT (PET/CT) improves the diagnostic accuracy by reducing false-negative and false-positive findings (5,25,26). The risk of false-positive findings is particularly high in the early postoperative period, 6–8 wk after surgery, but can persist for many years (5,27). In addition, the nonspecific uptake of [18F]FDG in inflammatory, noninfected tissue around the vascular graft is dependent on the graft material used (27). The use of methods of interpretation beyond visual assessment of uptake intensity (sensitivity, 0.90; specificity, 0.59), namely uptake patterns (heterogeneous and focal or multifocal) (sensitivity, 0.94; specificity, 0.81) and quantitative uptake assessment (SUVmax) (sensitivity, 0.95; specificity, 0.77), improves diagnostic accuracy (27). Leukocyte scintigraphy, including SPECT/CT, offers comparable sensitivity, with a pooled sensitivity of 0.99 in a metaanalysis (25) but improved pooled specificity (leukocyte scintigraphy without SPECT, 0.88; with SPECT, 0.82) compared with [18F]FDG PET/CT (pooled sensitivity, 0.95; pooled specificity, 0.80) (25). Leukocyte scintigraphy offers low false-positive rates in the early postoperative phase, but the patient numbers in the studies are relatively low and comparative studies to [18F]FDG PET are missing (5). Detailed information on the grade of evidence for the use of [18F]FDG PET/CT and leukocyte scintigraphy are given in the current European Association of Nuclear Medicine guideline on imaging infections in vascular grafts (5).
[18F]FDG PET and leukocyte scintigraphy face some additional tracer- and method-inherent advantages and disadvantages in imaging all types of cardiovascular infections. [18F]FDG is taken up by a variety of inflammatory cells and not by the pathogen itself, thus reflecting the host response to the infection, or, if false-positive, for example, to a recent surgical intervention. In addition, the host response can be altered, especially in elderly and immunosuppressed patients, who are also associated with an atypical clinical presentation (7).
[18F]FDG PET has a higher spatial resolution than leukocyte scintigraphy, even when SPECT/CT is performed. The accuracy of PET has improved with technologic improvements over time,as studies on IE before 2015 showed sensitivity significantly inferior to that of studies after 2015 (13). Implementation of respiratory and cardiac gating may further improve sensitivity; however, studies regarding cardiovascular infections are still lacking. [18F]FDG PET offers the advantage over the standard procedure, echocardiography, of providing whole-body images. Septic foci in other organs can be easily detected (Fig. 1), especially the portal of entry or septic emboli. Whole-body imaging allows for the detection of significantly more septic complications, as reported for IE by Kestler et al. (28). The systematic use of [18F]FDG PET resulted in a 2-fold reduction in recurrence in that study (28). In CDRIE, the involvement of different parts of the device defined by [18F]FDG PET has implications for clinical management and outcome (29,30). The diagnostic accuracy of imaging in general is further hampered by the time between initial diagnosis and imaging—for example, an 8-d median for [18F]FDG PET in IE in a large prospective dataset, with a large interquartile range of 5–15 d (4). By this time, patients have already been treated by empiric antibiotics and other medications influencing both the target microorganism and [18F]FDG uptake in host immune cells. This might be one of the reasons that evidence on the value of [18F]FDG PET for assessing response to antibiotic treatment is still sparse in cardiovascular infection, currently restricted to case reports (31,32).
Representative PET maximum-intensity projections (top) and PET/CT images (middle and bottom) of different types of infection in 3 patients. On left is [18F]FDG-positive VGEI (circle), in middle is infection of driveline of left ventricular assist device (arrows), and on right is [18F]FDG-positive PVIE (small arrow) and spondylodiskitis (large arrow).
To summarize, novel bacteria-specific tracers with a high target signal and a low background signal might improve diagnostic accuracy, especially in PVIE and CDRIE patients with a high risk of false-positive findings in the first months after surgery or in NVIE patients with false-negative results and inconclusive echocardiography results. In VGEI, novel bacteria-specific imaging agents may complement [18F]FDG PET in the postoperative setting to better differentiate active infection from postoperative inflammation. The potential implications of novel bacteria-specific tracers for response assessment and prediction would be highly speculative at this point, as clinical translation is still lacking for most of the promising candidates presented in the following sections.
TARGET SELECTION FOR IMAGING OF BACTERIA
Directly targeting the pathogen involved in the infection has several advantages over indirect imaging approaches that target the host’s immune response: the first is differentiation of infection from inflammation, the second is that imaging could be specific for a subgroup of pathogens or bacteria, the third is that imaging is less affected by immunosuppressive medication, and the last is that response to therapy may be easier to assess (33).
The development of specific tracers requires a thorough understanding of the biology of the bacteria involved in cardiovascular infections. Compared with a mammalian host cell with a volume of 100 to 1,000 μm3 a pathogenic bacterium ranges from 0.5 to 3 μm3 (2,34). Thus, high specificity of the imaging agent for procaryotic cells and signal amplification strategies are a prerequisite for adequate signals that can be detected by clinical PET scanners. Data on the exact number of bacteria during different stages of infections in different locations are limited. From previous publications on infections in various locations, it can be concluded that the bacterial concentration is at least 108 colony-forming units/mL in active infections (2,35,36). However, the bacterial burden might be decreased and even more difficult to detect in chronic infections or after initiation of treatment.
As principal targeting strategies, bacteria-specific differences in cell metabolism, proteins, and cell wall components can be used to develop bacteria-specific tracers, precluding uptake by mammalian cells (37). In addition, it seems important to choose targets that are maintained and preserved in different bacterial milieus. In the context of clinical imaging, which is often initiated only a few days after the start of antibacterial treatment, targets need to be still addressable (2).
Ordonez et al. identified potential targeting moieties through screening of different molecules as substrates for unique metabolic pathways in representative bacteria. This systematic approach resulted in 3 promising tracers for preclinical evaluation and potential clinical translation, underlining the need for biochemical preclinical research on potential targets (37). In the following sections, we discuss potential targets and corresponding specific tracers previously evaluated (Fig. 2; Table 2).
Currently existing and emerging bacteria-specific imaging agents and their targets. PVD = pyoverdine; YbT = yersiniabactin.
Advantages and Disadvantages of Different Bacteria-Specific Imaging Approaches, with Emphasis on Cardiovascular Infections
RADIOLABELED ANTIBIOTICS
Since antibiotics were designed to target and inactivate bacteria, these substances were of immediate interest for molecular imaging techniques. Therapeutically administered antibiotics have long biologic half-lives for effective treatment and are therefore of a lipophilic character to create high plasma protein binding with an equilibrium between bound and unbound that yields a nearly constant concentration of the free drug over many hours. For radiotracer-based imaging techniques such as SPECT or PET, this is counterproductive as it results in a high background accumulation and low image contrast. Nevertheless, many groups tried to label frequently used antibiotics for diagnostic infection imaging and studied their pharmacokinetic profiles. The fluoroquinolone ciprofloxacin (as [99mTc]ciprofloxacin) has been suggested as a promising candidate for infection imaging. In a large patient cohort of 879 patients, but only 26 patients with IE, this tracer presented a low sensitivity of 62.5% but a specificity of 100% in IE patients (38). In addition, unspecific binding of ciprofloxacin to DNA of both bacteria and host cells limits its use (39,40). Moreover, this tracer did not yet show high diagnostic accuracy for the differentiation of infection from sterile inflammatory processes (41). Thus, the signal might be based on passive leakage due to increased endothelial permeability in infection and inflammation but not on specific binding. Another general disadvantage of radiolabeled antibiotics is the emergence of antibiotic-resistant bacteria (39,40).
In 2017, Sellmyer et al. reinforced the attempt at using radiolabeled antibiotics for imaging with [18F]fluoropropyltrimethoprim, which has been evaluated in animal models discriminating cancer, sterile inflammation, and Escherichia coli infection (42). They expanded the portfolio to a 11C-labeled version ([11C]trimethoprim) and showed its uptake regardless of bacterial resistance against trimethoprim. The first clinical data underlined these findings in 6 human case studies with proven infections but not of the cardiovascular system (43). Besides the general disadvantages of radiolabeled antibiotics mentioned before, this approach faces tracer-specific disadvantages. Preclinical evaluation has revealed nonspecific absorption in the presence of an excess of unlabeled substance and minimal binding to heat-inactivated bacteria, raising concerns about its specificity (42,44). Therefore, we believe that there are more promising candidates for clinical translation in imaging infection.
ANTIMICROBIAL PEPTIDES
Ubiquicidin 29-41 is a synthetic peptide accumulating in the bacterial wall because its positive charge interacts with negatively charged phospholipids. Labeled with 99mTc for SPECT and 68Ga for PET, this tracer showed promising results in animal studies. Clinical translation has been attempted (45,46), but not in patients with cardiovascular infections (47). Concerns arose about the only 80% specificity found for the 99mTc-labeled compound in the first clinical studies (46). No reliable data on the sensitivity and specificity of the 68Ga-labeled compound exist so far (45,48), although the first case reports were published more than 10 y ago (49). This fact is underlined by a very recent systematic review by Marjanovic-Painter et al. with the resumé that larger clinical trials are needed to gain a clear view on applicability in clinical routine (48). Recently, Chen et al. critically discussed the strength and specificity of the binding mechanism, relying only on differences in positive and negative charge (50). Future general use of antimicrobial peptides in clinical routine (e.g., in cardiovascular infections) therefore seems less likely than for other more promising bacterial imaging agents.
ANTIBODIES
In the early 1980s, bacteria-specific antibodies and antibody fragments were the first steps in developing bacteria-specific imaging agents, starting with a 99mTc-labeled antistaphylococcal antibody for imaging endocarditis, which was very promising taking into account the technical status of imaging at that time (51). Four decades later, it has to be summarized that the specificity of imaging is limited by the fact that by targeting a surface antigen, a differentiation between bacteria in active infection and degraded bacteria after host immune reaction cannot be made (2). Moreover, antibody imaging is limited by disadvantageous penetration depths as well as slow blood pool clearance resulting in higher radiation exposure, low contrast, and prolonged imaging protocols hampering clinical translation to cardiovascular infection imaging. Nevertheless, many further attempts were made, for example, the detection of lipoteichoic acid on the surface of S. aureus in animal models of joint infection, but without convincing results (52).
MOLECULAR IMAGING AGENTS BASED ON PROKARYOTIC METABOLISM
For imaging of bacterial infections, high tissue penetration in areas of often-poor vascularization and perfusion and high target-to-background ratios are needed to image relatively low numbers of bacteria. Here, metabolic radiotracers that are small and are steadily accumulating seem to be the most promising candidates for sensitive and specific infection imaging. There are a few approaches targeting exclusive metabolic pathways in bacteria that eukaryotic cells are lacking.
First, carbohydrate transport and use are of great interest when developing an infection imaging agent. There are different possible substrates such as sugar alcohols, disaccharides, and oligosaccharides (37). The first suggested specific imaging agents in this area were based mainly on carbohydrates such as maltodextrins (53), which are metabolized exclusively by bacteria and not by eukaryotic cells. Although there were very promising initial results in preclinical applications for such agents as 18F-labeled maltohexaose (54), 18F-labeled maltose (55), and 18F-labeled maltotriose (56), none of these candidates have been translated into clinical settings so far. Especially, 6″-[18F]fluoromaltotriose seems to be promising as it showed a significant diagnostic and treatment monitoring capability in a murine model of S. aureus IE and the authors claimed its potential to change the clinical management of patients with bacterial endocarditis (57). Reasons for limited clinical translation could be the differences between artificial preclinical disease models such as subcutaneous infections and the complex clinical infection reality. Further, shifts in bacterial metabolism toward the use of alternative energy sources in complex infection scenarios, lower tracer uptake triggered by the local concentration of glucose and other metabolic substrates, or the formation of biofilms are challenging. A detailed view on the relation between degradation by starch-degrading enzymes and the chain length of different maltodextrins yielded an optimal scaffold of 3 glucose units (58). If specific positions on the carbohydrate-reducing and carbohydrate-nonreducing ends are blocked with 19F also, maltotetraose would be enzymatically stable enough (59). Nevertheless, these maltodextrins offer a high bacterial specificity as shown in Figure 3 (58).
Example of maltohexaose-based bacteria-specific imaging in mouse model of footpad infection with S. aureus in left hind paw. In vivo SPECT (left) shows [99mTc]MB1143 uptake (maltohexaose derivative) in infected footpad 3 h after injection, in clear contrast to healthy right foot. Correlative [18F]FDG PET/CT (right) in same mouse shows complementary signal surrounding site of infection, reflecting inflammatory response. %ID = percentage injected dose. (Reprinted with permission of (58).)
Also, sugar alcohols are a promising class as possible bacteria-specific radiotracers since this substance class is digested exclusively by bacteria. There should be many different sugar alcohols suitable for bacteria-specific imaging. The first and, up to now, most promising, [18F]-2-fluoro-d-sorbitol ([18F]FDS), is selectively taken up by gram-negative Enterobacteriaceae but not by gram-positive prokaryotes or mammalian cells (37). At first developed for tumor imaging (60) or functional renal imaging (61), [18F]FDS is actually of great interest for infection imaging. In comparison to the mentioned maltodextrin-based radiotracers, [18F]FDS can easily be synthesized by reduction of [18F]FDG (62). It is taken up via a sorbitol-specific phosphotransferase, which is highly conserved, and thus [18F]FDS can be used in antibiotic-sensitive and multidrug-resistant infection. After successful preclinical evaluation, [18F]FDS has been translated into different clinical settings, showing favorable renal excretion in the first human studies (63). Issues hampering clinical translation of [18F]FDS into cardiovascular infection imaging are discussed in the section on clinical translation.
FOLATE BIOSYNTHESIS
Besides the carbohydrate use, which may not be usable for bacteria-specific imaging in a broad spectrum of bacteria, all bacteria exclusively biosynthesize the essential folic acid themselves. In antibiotic therapy, this pathway is often targeted by trimethoprim inhibiting dihydrofolate reductase, which has been used for imaging as the radiolabeled antibiotic [11C]trimethoprim. Further, it has been shown that p-aminobenzoic acid (PABA) is also incorporated into the bacterial folic acid biosynthesis (37,64). PABA-derived radiotracers, [11C]PABA (65) and [18F]PABA (66), were shown to distinguish well between infection and sterile inflammation, with favorable pharmacokinetics, and can be used in both gram-positive and gram-negative bacterial infections. [11C]PABA PET imaging could accurately detect infections due to pyogenic bacteria in different orthopedically relevant animal models. In first-in-humans studies on healthy volunteers, [11C]PABA was safe and well tolerated and had a favorable pharmacokinetic behavior, with low background activity (65). Clinical translation into different infectious diseases is still lacking but should be of great interest not only in orthopedic scenarios but also in cardiovascular infections. Regarding the uptake mechanism, targeting the folate synthesis is a more general approach including gram-negative and gram-positive bacteria. During clinical translation, it seems of the utmost importance to compare this technique with clinically approved [18F]FDG PET and leukocyte scintigraphy as even more general approaches including the host’s response.
D-AMINO ACIDS
A further difference between prokaryotic and eukaryotic cells is related to bacterial cell wall biosynthesis, since bacteria use d-amino acids, especially d-alanine and d-glutamine. In a murine myositis model, d-[5-11C]glutamine was able to detect gram-negative E. coli and gram-positive S. aureus infections and discriminate those from sterile inflammation (67). d-[3-11C]alanine was evaluated in rodent models of diskitis–osteomyelitis and P. aeruginosa pneumonia (68). In uptake studies with 14C-labeled d-amino acids, d-methionine was identified as the most promising imaging agent candidate (69). The authors claimed that d-methionine may be incorporated into peptidoglycan muropeptides by transpeptidases, and the developed d-[11C]methionine could selectively differentiate both E. coli and S. aureus infections from sterile inflammation in vivo. Its synthesis is similar to the well-established synthesis of l-[11C]methionine, which is clinically established for brain tumor imaging. In a first small clinical study on 5 patients with joint infections, d-[11C]methionine showed low background uptake and fast urinary excretion, and areas of suspected infections presented moderately elevated uptake (70). However, these preliminary results have to be interpreted with caution for multiple reasons: chronic infection with negative histopathologic or microbiologic verification even after surgery, ongoing antibiotic treatment, and metal artifacts around protheses influencing attenuation correction and uptake quantification (70). Thus, these first results may also be explained by at least partly unspecific endothelial leakage due to the higher vascular permeability in an inflammatory and chronic infectious process and not only by specific uptake in the bacterial cell wall.
SIDEROPHORES
Evolutionarily, bacteria developed mechanisms to acquire metals from the surrounding tissue, maintaining metal homeostasis within the infected tissue. One common mechanism is the secretion of siderophores, which deliver the metal they capture to the bacteria via specific receptors (71). Specific siderophores are expressed, forming highly stable complexes mainly for Fe3+ but also for other metals. The structure of such siderophores is often specific for particular bacterial species or categories (72). This specificity makes broad-spectrum infection imaging impossible but, if needed, species-specific imaging of bacterial infection potentially available. Many of these structurally different Fe3+ complexes are significantly more stable than mammalian heme-iron complexes, allowing bacteria to steal iron from their mammalian hosts (73). As Ga3+ has the same charge and a similar radius, it is chemically interchangeable with Fe3+ regarding its complexation characteristics. In this regard, the 68Ga-labeled siderophore pyoverdine was successfully used for imaging of P. aeruginosa (74). Its renal elimination with almost no unspecific background uptake results in a clear signal and very high sensitivity. The same group also labeled desferrioxamine B with 68Ga ([68Ga]Ga-DFO-B), showing high and specific uptake of [68Ga]Ga-DFO-B by P. aeruginosa and S. aureus in vivo but no relevant uptake in E. coli, Candida albicans, and Klebsiella pneumoniae and claimed a great potential for clinical translation (Fig. 4) (75). Targeting the outer-membrane FyuA receptor expressing bacteria such as E. coli and K. pneumoniae, another group was able to detect infection sites in murine models with 64Cu-labeled yersiniabactin (76). These preclinical results showed that radiolabeled siderophores give excellent signal-to-background ratios and offer high selectivity for defined bacterial strains. Thus, a potential role might be as an adjunct to [18F]FDG PET as a niche application to image the presence of distinct bacterial populations, but siderophores will likely not replace [18F]FDG PET as a broad-spectrum imaging agent in imaging infection. In IE and VGEI, S. aureus accounts for a relevant percentage of infections, making the use of [68Ga]Ga-DFO-B interesting for clinical translation in cardiovascular infection. However, polymicrobial infections are frequent in VGEI, potentially limiting the use of imaging agents for bacterial subgroups as an adjunct to, for example, [18F]FDG PET.
Example of siderophore-based bacteria-specific in vivo PET/CT in mouse model of acute murine myositis (A) and in lung-infected Lewis rats (B). (A) [68Ga]Ga-DFO-B uptake in S. aureus–infected area (yellow arrow) vs. heat-killed area (1, white arrow), sterile inflammation (2, white arrow), and E. coli–infected area (3, white arrow) 45 min after injection. (B) Yellow arrow clearly indicating site of P. aeruginosa infection (2), in contrast to noninfected animal (1). (Reprinted from (75).)
CLINICAL TRANSLATION
After profound preclinical evaluation, clinical translation is needed for promising imaging agents to overcome preclinical model–specific disadvantages. In most preclinical studies, subcutaneous infection models are used. These cannot depict well the complex clinical infection reality affecting multiple organs and metabolic up- and downregulation over time. After first-in-humans pilot studies, preferably in a prospective setting (77), followed by biodistribution and radiation dosimetry phase I trials (65,78), the role of the bacteria-specific imaging agent has to be evaluated in different clinical settings in prospective and retrospective analyses. Therefore, there has to be a clinical need and an appropriate bacterial burden (2). Such a scenario would be VGEIs with a high amount of bacteria involved but possible unspecific morphologic imaging and [18F]FDG uptake related to postsurgical inflammation in the first months hampering sensitivity and especially specificity (5).
Prospective studies are needed to image a well-defined subgroup of patients at an early stage of the disease with no or only limited previous antibiotic treatment. These prospective studies and a subsequent metaanalysis were performed for the role of [18F]FDG PET in IE and VGEI (13,26). In IE, the result was recently published international guidelines proving the impact of [18F]FDG PET in the clinical work-up of IE (7,12).
Recent clinical translation of [18F]FDS into a first-in-humans prospective study allowed for specific detection of infection with gram-negative Enterobacteriaceae and successful differentiation from sterile inflammation or malignant lesions. Moreover, [18F]FDS PET monitored the response to antibiotic treatment (63). Twenty-six patients were prospectively enrolled, with 18 patients having enterobacterial infection predominantly in bone, soft tissue, and lung. None of the patients had an infection affecting the cardiovascular system (63). In IE, generally less than 10% of patients have an infection with gram-negative pathogens (10); in VGEI, up to one third of the patients have an infection with gram-negative pathogens, but polymicrobial infections are frequent (9,11). Thus, the role and potential of [18F]FDS PET are limited in infections of the cardiovascular system. However, in VGEI patients with a relevant proportion of responsible gram-negative pathogens, the absence or presence of gram-negative bacteria as shown by [18F]FDS PET might influence antibiotic treatment. Currently, treatment of cardiovascular infections often encompasses empiric antibiotic treatment for a prolonged time of up to several weeks. Tailored antibiotic treatment is of importance as the number of multidrug-resistant bacterial infections increases (33). Other subgroup-specific or even strain-specific tracers, namely siderophores, might play a role as an adjunct to [18F]FDG PET toward a personalized medicine approach influencing antibiotic treatment strategies. In particular, the siderophore tracer [68Ga]Ga-DFO-B targeting gram-positive S. aureus seems to be promising for cardiovascular infection, taking into account the underlying pathogens in IE and VGEI. More general bacteria-specific tracers targeting multiple gram-positive and gram-negative bacteria, such as other sugars or tracers of folate metabolism, will have to compete directly with [18F]FDG to define their role.
Data on clinical translation of novel bacteria-specific tracers are sparse and often limited to first-in-humans application for biodistribution purposes or case series (65). However, first larger prospective studies arise on the horizon, such as a prospective interventional study with 30 patients investigating the role of [68Ga]Ga-DFO-B, a siderophore tracer, in VGEIs (NCT05285072).
FUTURE DIRECTIONS
Potential drawbacks of bacteria-specific imaging, as previous antibiotic treatment hampers sensitivity, can partly be overcome by technologic innovations such as concomitant diagnostic imaging by integrated PET/CT. This was already shown for [18F]FDG PET/CT in VGEI, with similar sensitivity but improved specificity compared with [18F]FDG PET alone (PET only: pooled sensitivity, 0.94; specificity, 0.70; PET/CT: pooled sensitivity, 0.95; specificity, 0.80) (25).
Further enhancement of the signal-to-noise ratio adjacent to the diaphragm may aid in the diagnosis of IE. Respiratory and cardiac motion correction influences image quality and quantitative uptake parameters (79), with a potential impact on diagnostic performance in cancer patients (80). However, data on use in cardiovascular infections are not available.
Recently introduced PET scanners with a large field of view (total-body PET) will revolutionize molecular imaging. By increasing the sensitivity of the PET system, we might more accurately detect small infective foci (e.g., in NVIE). The lower radiation exposure of total-body PET (81) is of special interest in benign disease such as cardiovascular infections, opening the door for multiple-time-point PET to monitor treatment response (2). Multiple-tracer studies, including a bacteria-specific agent in direct comparison to [18F]FDG, become feasible. Dynamic imaging of the whole body allows for kinetic modeling without invasive arterial blood sampling for the differentiation of specific tracer binding from unspecific accumulation. This capability is of special interest in imaging infections (e.g., in the postoperative setting in PVIE, CDRIE, and VGEI) because of unspecific leakage of tracers due to the increased vascular endothelial permeability that occurs in infection and inflammation. Most first-in-humans studies of novel tracer candidates in a small number of patients do not distinguish between the 2 components of the signal—active binding and vascular leakage—but positive results are misinterpreted as active binding (44).
Advances in cell labeling including multiple methods described in other reviews may improve our understanding of the role of different immune cells, especially in inflammation but also in distinction from infection (82).
Radiomics and other advanced artificial intelligence applications are now frequently used in morphologic cancer imaging. There is a growing body of evidence on a holistic approach including the best of both worlds—advanced morphologic and molecular imaging—into artificial intelligence–based analysis (83,84). Imaging cardiovascular infection offers opportunities for improvement, such as in terms of the currently limited visual assessment and sensitivity, for example, in NVIE. Another aim of artificial intelligence in infective cardiovascular disease is to gain predictive and prognostic imaging values (85).
The availability of PET in general and of novel imaging agents must be increased to allow fast imaging after symptom onset and suspicion of cardiovascular infection. Development of new bacteria-specific tracers is cost-intensive, and prospective phase II or III studies need substantial funding. In comparison to studies on cancer, dedicated funding for infection is marginal, resulting in limited support for the development of new bacteria-specific PET tracers (2). In the future, nuclear medicine should focus on prospective clinical evaluation of the most promising preclinical candidates. Targets or strategies that have been discussed for decades, with general concerns about binding specificity (e.g., antibiotics and antimicrobial peptides) or unfavorable kinetics (e.g., antibodies), are less likely to realize a clinical breakthrough in cardiovascular infection imaging. Currently, [18F]FDG PET and leukocyte scintigraphy are still the methods of choice for cardiovascular infection imaging, with high diagnostic accuracies in many clinical scenarios. Therefore, new tracers have to be compared with, or established as an adjunct to, these approaches. Albeit first data on the predictive and prognostic value of [18F]FDG PET exist, future research in this field is warranted to strengthen our knowledge about the currently available imaging methods.
Besides specific imaging methods, the development of more specific and more rapid blood-based detection of microorganisms causing infection of the cardiovascular system would further improve diagnostic accuracy. For rapid improvements in the field of imaging and treatment of infections, a multidisciplinary approach encompassing microbiologists, molecular imaging specialists, radiologists, and clinical infectious disease physicians, including cardiologists and cardiothoracic and vascular surgeons, is needed (33).
CONCLUSION
Bacteria-specific imaging in cardiovascular infections is often considered a highly promising approach with the potential to overcome well-known disadvantages of [18F]FDG PET. Over the years, it has turned out that many bacteria-specific tracers face either target-specific issues, such as low binding strength and specificity, or unspecific accumulation in hyperemic areas of infection or inflammation in first-in-humans applications, comparable to [18F]FDG PET. Prospective clinical translation for the remaining promising preclinically evaluated bacteria-specific tracers targeting prokaryotic metabolism, folate biosynthesis, or siderophores is warranted. Currently, [18F]FDG PET is the imaging method of choice for cardiovascular infection imaging, with accumulating data and inclusion into guidelines. It still remains to be proven whether bacteria-specific imaging agents might complement or replace imaging with [18F]FDG in cardiovascular infections.
DISCLOSURE
This work was supported by DFG-CRC1450-431460824 (projects A03, A04, and C01), Münster, Germany. No other potential conflict of interest relevant to this article was reported.
ACKNOWLEDGMENT
Figure 2 was created with Biorender.com.
Footnotes
Published online Nov. 1, 2023.
- © 2023 by the Society of Nuclear Medicine and Molecular Imaging.
REFERENCES
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.
- 24.
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- Received for publication February 16, 2023.
- Revision received May 31, 2023.