Erythropoietin in the critically ill: do we ask the right questions?

There is a plethora of experimental data on the potential therapeutic benefits of recombinant human erythropoietin (rhEPO) and its synthetic derivatives in critical care medicine, in particular in ischemia/reperfusion injury. Most of the recent clinical trials have not shown clear benefits, and, in some patients, EPO-aggravated morbidity and mortality was even reported. Treatment with rhEPO has been successfully used in patients with anemia resulting from chronic kidney disease, but even a subset of this patient population does not adequately respond to rhEPO therapy. The following viewpoint uses rhEPO as an example to highlight the possible pitfalls in current practice using young healthy animals for the evaluation of therapies to treat patients of variable age and underlying chronic co-morbidity.


Introduction
Th e potential therapeutic benefi ts of recombinant human erythropoietin (rhEPO) and its synthetic derivatives in critical care medicine have recently been reviewed in various journals [1][2][3] and even occasioned impassioned correspondences [4,5]. Th e results of pre-clinical trials suggest that rhEPO could be used not only to ameliorate anemia, but also to limit organ injury/dysfunction associated with stroke, myocardial infarction, trauma, hemorrhage and sepsis. However, despite the promising pre-clinical results, most recent clinical trials have not shown clear benefi ts, and, in some patients, EPO-aggravated morbidity and mortality was even reported. Since the 1990s, rhEPO has been successfully used to treat chronic kidney disease (CKD)-related anemia, but a subset of approximately 10% of these patients does not adequately respond to rhEPO therapy [6]. Th is condition is referred to as 'EPO resistance' , which is characterized by either a need for higher doses of EPO to maintain the recommended hemoglobin (Hb) or even a lack of response to EPO at all [6]. Th e confl icting results of the recent clinical trials that evaluated the cytoprotective eff ects of rhEPO beg the question of the applicability of our preclinical models in the clinical setting. Th e following viewpoint uses rhEPO as an example to highlight the possible pitfalls in current practice using young healthy animals for the evaluation of therapies to treat patients of variable age and underlying chronic co-morbidity.

Erythropoietin and its receptor
While EPO is mainly produced in the peri-tubular cells of the kidney in response to hypoxia, low levels of EPO mRNA have also been reported in the central nervous system, lungs and spleen. EPO is well-known as a regulator of erythrocyte production to optimize tissue oxygenation: A drop in local O 2 tension leads to the stabilization of hypoxia inducible factor, which binds to the hypoxia-responsive elements of the EPO gene activating its transcription. EPO needs a receptor (EPO-R) to perform its function, and this EPO-R is expressed on erythroid cell progenitors and in a variety of tissues and cell types -for example, the brain, retina, heart, kidney, vascular smooth muscle cells, myoblasts and vascular endothelium. Administration of EPO upregu lates EPO-R expression and increases endothelial nitric oxide (NO) production. EPO-R expression was also confi rmed in primary human kidney tubular epithelial cells, in rat cortical and medullary tubules as well as in porcine wound healing fl uid, granulation tissue, and kidney [7][8][9]. However, children with acute kidney injury presented with elevated EPO-R expression in the kidney but decreased EPO plasma levels [10], and diff erential regulation of EPO-R expression in renal tissue biopsies from young, healthy versus older, co-morbid swine was reported [11].
Accruing evidence suggests that EPO exerts tissueprotective properties via a diff erent heteroreceptor

Abstract
There is a plethora of experimental data on the potential therapeutic benefi ts of recombinant human erythropoietin (rhEPO) and its synthetic derivatives in critical care medicine, in particular in ischemia/ reperfusion injury. Most of the recent clinical trials have not shown clear benefi ts, and, in some patients, EPOaggravated morbidity and mortality was even reported. Treatment with rhEPO has been successfully used in patients with anemia resulting from chronic kidney disease, but even a subset of this patient population does not adequately respond to rhEPO therapy. The following viewpoint uses rhEPO as an example to highlight the possible pitfalls in current practice using young healthy animals for the evaluation of therapies to treat patients of variable age and underlying chronic co-morbidity.
EPO-R isoform, which has been proposed to comprise a classic EPO-R homodimer and the cytokine β-common receptor (βcR). Gorio and colleagues [12] demonstrated both the association of the βcR subunit and the EPO-R as well as the need for the heteroreceptor combination for the recovery of motor function after spinal cord compression injury. Saqib and colleagues [9] showed in a porcine model of wound healing that EPO was associated with an increase of granulation tissue, and demonstrated higher expression and the co-localization of EPO-R and βcR in the cellular constituents of the granulation tissue. It is noteworthy that the βcR is involved in EPO-mediated endothelial nitric oxide synthase (eNOS) activation in endothelial cells [13]: both EPO-and eNOS-derived NO inhibit neo-intima formation and improve re-endo thelialization in a dose-dependent manner [14]. Furthermore, synthetic EPO derivatives like carbamylated EPO (cEPO) provided additional insight into the properties of the EPO hetero-receptor complex: cEPO does not bind to the hematopoietic EPO-R and thus does not increase the hematocrit, but exerted cytoprotective eff ects in cerebral infarction, spinal cord trauma, and kidney ischemia/ reperfusion (I/R) injury [15]. Recently, cEPO was even reported to more eff ectively reduce kidney infl ammation in brain-dead rats than rhEPO [16], and a newly developed cEPO-FC fusion protein was at least as protective as rhEPO in a porcine aortic balloon occlusion-induced spinal cord I/R injury [17]. At present, four clinical trials have evaluated the safety and pharmacokinetics of cEPO for acute ischemic stroke (ClinicalTrials.gov identifi ers NCT00870844 and NCT00756249), in advanced kidney cancer (NCT00035243) as well as in patients with the Friedreich´s ataxia (NCT01016366): so far results are not yet available.

Erythropoietin clinical trials
Corwin and coleagues' report on the CRIT Study [43] examined the incidence of anemia and red blood cell transfusions in critically ill patients and determined that trauma patients were more likely to be transfused than non-trauma patients. Four separate randomized, placebocontrolled studies using rhEPO in this context were conducted, which enrolled 160, 1,302, and 1,460 anemic (total Hb concentration of <12 g dL -1 ) critically ill patients [44][45][46] and 86 'long-term acute care patients' [47]. Th e fi rst two trials demonstrated a reduction of transfusion requirements, and the second even had an increased survival rate in the treatment arm. Due to a lack of data of specifi c trauma events that could aff ect the outcome, however, a defi nitive assessment was impossible. Interestingly, in the third trial no transfusion reduction was observed with treatment despite the increase in Hb content. Furthermore, there was a clinically signifi cant increase in thrombovascular events in rhEPO-treated patients in comparison to vehicle [44,45]. Finally, the most recent long-term trauma outcome study evaluating the role of rhEPO in anemic (Hb <12 g dL -1 ) trauma subjects found no diff erences in physical function or safety between the treatment and control arms [46]. Ehrenreich and colleagues [48] showed promising neuro protective eff ects of rhEPO in a pilot study of ischemic stroke. Th e subsequent large double-blind, placebo-controlled, randomized muticenter rhEPO stroke trial not only failed to show any neuroprotective benefi t, but, contrary to all expectations, patients treated both with rhEPO and tissue plaminogen activator presented with increased intracerebral hemorrhage and mortality [49]. Yip and colleagues [50] also tried to assess the benefi ts of rhEPO after acute ischemic stroke: they reported an increase in endothelial progenitor cells and decrease of 90-day major adverse neurological events. Th e commentary by Minnerup and colleagues [51] highlighted the fact that the two trials focused on diff erent primary endpoints: a reduced incidence of recur rent strokes at day 90 does not necessarily imply improved neurological function.
Th e two trials on EPO eff ects on spinal cord injury, Evaluation of Tolerability and Effi cacy of EPO Th erapy in Spinal Shock (NCT00220675) and EPO Spinal Cord Compression Randomized Trial (NCT00220675) were both terminated prematurely. An additional trial looking at the benefi ts of rhEPO without prophylactic anticoagulation in elective spinal surgery noted an increase in deep vein thrombotic events. Th e study concludes with the recommendation to add anti-thrombotic prophylaxis to rhEPO in the surgical setting [52].
Th e Reduction of Infarct Expansion and Ventricular Remodeling with Erythropoietin after Large Myocardial Infarction (REVEAL) trial enrolled 222 patients and showed unchanged infarct size after treatment compared to vehicle. Interestingly, in the treatment arm, older patients (aged >70 years) even presented with a doubling in infarct size in the fi rst week [53].
In the setting of acute kidney injury (AKI) a study of 71 patients undergoing elective coronary artery bypass graft surgery had a reno-protective eff ect [54], whereas the larger (n = 162) Early Intervention in Acute Renal Failure (EARLYARF) trial, evaluating rhEPO therapy in a heterogeneous group of ICU patients, found no such eff ects [55]. Another clinical trial (Recombinant Human Erythropoietin use in Intensive Care Unit Patients: Does it prevent acute renal failure; NCT00676234) recruited 80 patients and was completed in 2009, but no data are available so far. Finally, a very recent follow-up report from the aforementioned trial [55] on the incidence of end-stage renal disease and mortality showed that rhEPO reduced all-cause mortality and development of endstage renal disease in patients that had previously suff ered from AKI [56]. Th is subset of patients with AKI comprised 21 patients, 14 in the placebo group and 7 in the rhEPO group. Interestingly, patients in the placebo group were older (67 to 84 years; 10 of the 14 patients were >70 years) than those in the rhEPO group (58 to 75 years; 3 of the 7 were >70 years). It may be too early to make defi nitive conclusions from these data, but the Table 1   REVEAL study suggests putative harm by rhEPO in patients aged over 70 years [53]. Whether or not age (and the presence or not of underlying CKD) may provide useful information defi ning who may be best served by rhEPO therapy warrants further investigation.

Chronic kidney disease and erythropoietin resistance
In an eff ort to understand why results from recent clinical trials to treat AKI are mixed, let us take a look at renal disease and CKD-related co-morbidity and EPO resistance. Renal disease is associated with a graded increase in both infl ammatory and oxidative markers: i) patients with CKD presented with increased blood lipid hydroperoxide, oxidized low density lipoproteins, F2-isoprostanes, TNF-α, IL-6, and IFN-γ when compared with patients with normal kidney function [6,57]; ii) in subgroup studies from clinical trials, patients with CKD responded diff erently to pharmaceutical interventions compared to patients with normal kidney function [58]; iii) atherosclerosis, which is characterized by an increase in low density lipoproteins, a decrease in high density lipoproteins, oxidative stress, endothelial dysfunction and infl ammation, is prevalent in CKD, increases with age, and is the main risk factor for cardiovascular disease [57]. Finally, atherosclerosis is also associated with reduced NO bioavailability [59], and the constitutive production of NO has been shown to be attenuated in patients with CKD [60]. Th e mechanisms underlying EPO resistance are poorly understood and most likely multi-factorial, since endogenous EPO levels tend to be higher in these patients than in control subjects [61]. Age and the manifold aspects of ageing add to this complexity: in a geriatric cohort higher EPO blood levels were directly related to mortality [6]. Nevertheless, there is general consensus that infl ammation and oxidative stress are key players [6,62,63]: the pro-infl ammatory cytokines IL-6, IFN-γ, and TNF-α may antagonize the actions of EPO by inhibiting erythroid progenitor cells, activating suppressor of cytokine signaling, down-regulating EPO-R expression and generating reactive oxygen species that lead to lipid peroxidation of red cell membranes [6,62,63]. Moreover, EPO activates eNOS, which is considered to be critical for its tissue protective eff ects: genetic eNOS deletion is associated with a loss of response of endothelial progenitor cells to EPO stimulation ex vivo [64], and in vivo EPO not only failed to exert any vasoprotective eff ects but even worsened remodeling after vascular injury [59]. In rats with heminephrectomyinduced polycythemia, EPO aggravated arterial hypertension and only partially attenuated the fall of the glomerular fi ltration rate caused by non-selective NO synthase inhibition with L-NAME (N G -nitro-L-argininemethyl ester) [65].

Animal models
Animal models that use young and healthy animals are essential for the understanding of basic pathophysiological mechanisms. Any investigator will try to reduce inter-individual variation as much as possible and choose animals of the same sex, age and strain in order to control for physiology and establish reproducible and defi ned conditions. Such models are valuable inasmuch as they provide unique insights into the pathophysiology of specifi c experimental scenarios and even identify novel therapeutic targets. However, one of the problems with research conducted in naive animals is that a dramatic benefi t is often observed that cannot be reproduced in the clinical study: a systemic review of pre-clinical and clinical trials concluded that the discordance was due, at least in part, to the failure in the pre-clinical trial to properly mimic the clinical disease [66]. A single factor such as age may have major eff ects: antibiotic therapy in cecal ligation puncture-induced murine sepsis halved mortality in young animals, while this intervention had no benefi t in older mice [67]. In contrast to the epidemiology in patients, who usually present with variable co-morbidity, all data on EPO-related organ protection against I/R injury originate from models investigating young and healthy animals. Th is limitation thus assumes importance in light of failed clinical studies in comparison to pre-clinical trials. We found a similar lack of protec tion against I/R injury of rhEPO in adult swine with ubiquitous atherosclerosis resulting from familial hyper cholesteremia [11] and an atherogenic diet (socalled familial hypercholesteremia Bretoncelles Meishian (FMB) swine) when compared to otherwise young and healthy animals [17,39]. Th ese FBM swine present with hypercholesteremia and increased markers of oxidative stress, while creatinine clearance, blood levels of NO metabolites, and renal tissue expression of EPO-R are reduced -that is, this strain shows a biomarker pattern comparable with that found in patients with hypercholesteremia-induced atherosclerosis [11,68]. As agematched wild-type and young (6 months) FBM swine without the atherogenic diet showed the same EPO-R expression as young and healthy animals, the reduced EPO-R expression may not only provide a plausible explanation for ineff ectiveness of EPO in this model, but also potentially hint at one of the underlying causes of 'EPO resistance' .

Conclusion
Th e promise of pre-clinical data on organ-protective eff ects of rhEPO has not been matched by successful clinical trials. Th e results from the animal models using young, healthy animals provide us with very important pathophysiological mechanistic information. Th e mechanisms may apply, but often other factors, including gender, age, and, in particular, co-morbidity, confound the therapeutic strategy. Th e distinct contrast in the experimental results in kidney I/R injury between the young, healthy swine and the FBM swine might help to underline the importance of pre-existing co-morbid conditions for the design of pre-clinical experimental models. Th ese results may not only off er a potential expla nation for the diff ering results of receptor expression in human samples, which may be reconciled when the etiology of disease of the donors are better understood, but also suggest that animal models that more closely mimic the human disease conditions may provide better guidance for future therapeutic strategies. Finally, it is tempting to speculate whether pre-existing impairment of kidney function and decreased renal tissue EPO-R expression may explain the controversial eff ects of rhEPO in clinical trials.