An extracorporeal blood-cleansing device for sepsis therapy

Technical Reports An extracorporeal blood-cleansing device for sepsis therapy npg © 2014 Nature America, Inc. All rights reserved. Joo H Kang1–3,7, Michael Super1,7, Chong Wing Yung1,2, Ryan M Cooper1,2,4, Karel Domansky1, Amanda R Graveline1, Tadanori Mammoto2, Julia B Berthet1, Heather Tobin2, Mark J Cartwright1, Alexander L Watters1, Martin Rottman1,6, Anna Waterhouse1, Akiko Mammoto2, Nazita Gamini1, Melissa J Rodas1, Anxhela Kole1, Amanda Jiang2, Thomas M Valentin1, Alexander Diaz1, Kazue Takahashi5 & Donald E Ingber1–3 Here we describe a blood-cleansing device for sepsis therapy inspired by the spleen, which can continuously remove pathogens and toxins from blood without first identifying the infectious agent. Blood flowing from an infected individual is mixed with magnetic nanobeads coated with an engineered human opsonin—mannose-binding lectin (MBL)—that captures a broad range of pathogens and toxins without activating complement factors or coagulation. Magnets pull the opsonin-bound pathogens and toxins from the blood; the cleansed blood is then returned back to the individual. The biospleen efficiently removes multiple Gram-negative and Gram-positive bacteria, fungi and endotoxins from whole human blood flowing through a single biospleen unit at up to 1.25 liters per h in vitro. In rats infected with Staphylococcus aureus or Escherichia coli, the biospleen cleared >90% of bacteria from blood, reduced pathogen and immune cell infiltration in multiple organs and decreased inflammatory cytokine levels. In a model of endotoxemic shock, the biospleen increased survival rates after a 5-h treatment. The presence of microbial pathogens in the bloodstream triggers systemic inflammation and can lead to sepsis, which often overcomes the most powerful antibiotic therapies and causes multiorgan systems failure, septic shock and death1–4. Sepsis afflicts 18 million people worldwide every year5, with a 30–50% mortality rate even in state-ofthe-art hospital intensive care units3–8, and its incidence is increasing because of the emergence of antibiotic-resistant microorganisms6. Sepsis treatment usually involves the use of empiric, broad-spectrum antibiotic therapy because it takes days to identify the source of the infection and blood cultures are often negative. But these broadspectrum agents are not as effective as therapeutics targeted against specific microbes, and they can produce severe side effects7; as a result, mortality rates increase as much as 9% for every hour before the correct antibiotic therapy is administered8. This situation is even more devastating in patients with antibiotic-resistant pathogens because of the lack of effective drugs, and in immunosuppressed patients and neonates with fungal infections because of the high toxicity of antifungal agents6,9. Other methods used to treat sepsis (for example, administration of fluids, anti-thrombosis therapy, hemofiltration of inflammatory mediators and extracorporeal organ support10) do not address the root of the problem, which is the persistence of both live and dead pathogens that release toxins into the bloodstream2. With the 2011 withdrawal of the Eli Lilly drug Xigris (drotrecogin alpha), the recombinant form of activated protein C, there are now no drugs specifically approved to treat sepsis11. We initiated our project based on the observations that blood pathogen load is known to be a major contributor to both disease severity and mortality in patients with sepsis12, and that many patients respond to appropriately targeted antibiotic therapies that work exclusively by lowering the number of live pathogens. Thus, we set out to develop an extracorporeal blood-cleansing therapy that can rapidly remove microorganisms and endotoxins from blood without the need to first identify the source of the infection and without altering blood contents. Although combined micromagnetic-microfluidic techniques have been used for many applications13–15, including pathogen isolation16,17, this approach has not been used to cleanse whole blood. Here we describe an improved extracorporeal microfluidic device that incorporates a flow channel design inspired by the microarchitecture of the spleen, which can be used to cleanse pathogens from the flowing blood of patients with sepsis when combined with broad-spectrum magnetic opsonins. These capture agents are composed of magnetic nanobeads coated with a genetically engineered version of human MBL that binds to a wide variety of pathogens and is easily manufactured but lacks key functional domains that could complicate therapy. 1Wyss Institute for Biologically Inspired Engineering, Boston, Massachusetts, USA. 2Vascular Biology Program, Boston Children’s Hospital and Harvard Medical School, Boston, Massachusetts, USA. 3Harvard School of Engineering and Applied Sciences, Cambridge, Massachusetts, USA. 4Harvard-MIT Health Sciences and Technology Graduate Program, Cambridge, Massachusetts, USA. 5Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, USA. 6Current address: Université de Versailles St-Quentin en Yvelines, Versailles, France and Service de Microbiologie, Hôpitaux Universitaires Île de France Ouest, AP-HP, Garches, France. 7These authors contributed equally to this work. Correspondence should be addressed to D.E.I. (don.ingber@wyss.harvard.edu). Received 30 December 2013; accepted 25 April 2014; published online 14 September 2014; doi:10.1038/nm.3640 nature medicine VOLUME 20 | NUMBER 10 | OCTOBER 2014 1211 npg © 2014 Nature America, Inc. All rights reserved. Technical Reports RESULTS Engineering a human opsonin for blood cleansing To develop a broad-spectrum opsonin that can be used without complication to cleanse whole blood flowing through an extracorporeal fluidic circuit, we genetically engineered the large (650 kDa) native MBL protein by deleting its collagen-helix domain that contains the molecule’s natural complement-fixation and blood coagulation– promoting activities as well as its ability to bind to receptors on macrophages18,19 (Fig. 1a). The remaining carbohydrate-recognition domain and neck domains of the human MBL were fused to human IgG1 Fc, enabling high expression and secretion as well as efficient, rapid and low-cost purification of the smaller (90 kDa), recombinant protein20. The Fc-containing MBL (FcMBL) was biotinylated at its N terminus by addition of the tripeptide sequence alanine-lysinethreonine. Site-specific biotinylation of the terminal tripeptide sequence enabled us to uniformly orient FcMBL at high density on the surface of streptavidin-coated superparamagnetic nanobeads (128-nm diameter) to create multivalent magnetic opsonins (Fig. 1a). Notably, the engineered FcMBL retained the same potent pathogenand endotoxin-binding capabilities as reported for wild-type MBL21–23 when we tested the magnetic opsonins in standard in vitro binding assays against Candida albicans, Saccharomyces cerevisiae, E. coli and S. aureus (Fig. 1b and Supplementary Fig. 1a), a broad range of other Gram-negative and Gram-positive bacteria and fungi (Supplementary Table 1 and Supplementary Fig. 1a), and lipopolysaccharide (LPS) endotoxin from Gram-negative bacteria (Supplementary Fig. 1b). The FcMBL also bound to multiple antibiotic-resistant organisms, including methicillin-resistant S. aureus, Staphylococcus epidermidis, Klebsiella pneumoniae and E. coli (Supplementary Table 1)24–26. The FcMBL also displayed greatly decreased coagulation-promoting, complement-activation and DNA-binding activities compared with MBL (Supplementary Fig. 1c,d and Online Methods). Bioinspired fluidic device design To efficiently remove magnetically opsonized pathogens and endotoxins from septic blood flowing at liters per hour without loss or dilution of blood, we developed a fluidic device design inspired by the microarchitecture of the living spleen (Fig. 1c and Supplementary Fig. 2a–c). The biospleen unit consists of two adjacent, rectangular fluidic channels. One acts like a vascular channel (2 mm × 0.6 mm × 27 mm; width × height × length) that contains flowing blood (up to 1.25 liter h−1), and the other one (2 mm × 0.34 mm × 27 mm) contains saline under intermittent or slow flow (0–0.1 liter h−1; 5 min on and 60 min off), much like in the spleen’s venous sinusoids. In the living spleen, opsonized pathogens are cleared from the blood by resident macrophages as the blood is filtered through slits between adjacent endothelial cells when flow slows down and trickles from the marginal zone and red-pulp cords into the venous sinuses27. Similarly, the bio­ spleen contains a series of open rectangular slits (2 mm × 0.34 mm × 0.34 mm, spaced every 0.43 mm) that provide direct access between the blood channel and the saline-filled sinusoid channel (Fig. 1c); we based the channel and slit dimensions on our previous device design15. To increase isolation efficiency at higher flow rates, the single inflow channel into the device branches 4 times before entering a Figure 1 Magnetic opsonin and biospleen device. (a) Design scheme MBL CRD for genetic engineering of native mannose-binding lectin (MBL) to produce the generic opsonin FcMBL and coat it on magnetic nanobeads MBL Magnetic neck (128 nm) to produce magnetic opsonins, as shown at the right. FcMBL nanobead Collagen IgG1 was produced by engineering the protein to retain MBL’s carbohydratehelix Fc recognition domain (CRD) while removing its collagen-helix regions. FcMBL (90 kDa) MBL (650 kDa) These regions contain domains that support binding of MBL-associated serine proteases that mediate cleavage of complement and fibrinogen— Magnetic bead responsible for MBL’s complement-fixation and coagulation-promoting E. coli activities36, respectively—as well as domains that are responsible for MBL’s binding to complement and collectin receptors on the surfaces of macrophages18,19. These regions were then replaced with the Fc region of IgG1 to ensure stability and facilitate protein purification, and a S. aureus terminal biotinylation site (not shown) was used to consistently orient Discard out the FcMBL at high density on the surface of the streptavidin-coated Saline in Venous system Magnets nanobeads with the CRD facing outward. (b) Pseudocolored scanning electron micrographs showing multiple opsonin-coated magnetic beads Venous (128 nm) bound to the bacteria S. aureus (orange/brown; left) and Arterial Sinusoid Sinusoid red-pulp slits E. coli (blue; right). Scale bars, 1 µm; arrows indicate pathogen with bound cord Arterial beads. (c) Schematics of a venous sinus in the red pulp of the spleen (left) and a longitudinal view of the biospleen (right), with a photograph of an Magnetic Endothelial opsonin in engineered device (top right). The blood-cleansing biospleen mimics the cells Septic Cleansed Stress fiber architecture and function of the human spleen by incorporating a highblood in blood out flow vascular arterial channel perfused with contaminated whole blood Peristaltic pump and interconnected by open slits to a parallel low- or intermittent-flow Magnetic venous sinusoid channel perfused with isotonic sterile saline; this mimics opsonins Anesthesia the arterial red-pulp cord and venous system separated by sinusoid slits between neighboring endothelial cells. Magnetic opsonins are added Jugular to the flowing septic (contaminated) blood and passed through an Static mixer catheters in-line spiral mixer and incubation loop that promote nanobead-pathogen binding (shown in d) before entering the arterial channel of the biospleen. Stationary magnets positioned directly above the sinusoid channel pull Incubation loop the magnetic opsonins and bound pathogens through the open slits, into the saline-filled channel and into a discard collection vial (Supplementary Magnets Magnetic separator unit Video 1). A micromachined aluminum magnetic separator unit of the biospleen is shown at the right above the diagram (scale bar, 50 mm); the inset shows a higher-magnification view of the slit regions of the same device (scale bar, 20 mm). (d) Photograph of the experimental animal setup for extracorporeal blood cleansing, with components labeled. b c ∆B ∆B ∆B d 1212 VOLUME 20 | NUMBER 10 | OCTOBER 2014 nature medicine Technical Reports npg © 2014 Nature America, Inc. All rights reserved. Removal of pathogens and endotoxin from blood in vitro When we tested the efficiency of the biospleen at removing pathogens prebound to magnetic opsonins and spiked into saline or blood in vitro, we confirmed removal efficiencies of 99% and 90% for S. aureus in saline and blood, respectively, at a flow rate of 535 ml h−1 (Supplementary Fig. 3d), as predicted by our theoretical model (Supplementary Fig. 3c). Moreover, we still could clear ~85% and ~60% of these pathogens when the flow was raised to 1 liter h −1 (Supplementary Fig. 3d). However, many patients with sepsis are infected with multiple microbes simultaneously (for example, because of an intestinal injury). To model this condition, the cecal contents of meat-fed rats (Online Methods) were spiked into banked whole human blood (3.5 × 104 anaerobe CFU and 1.5 × 104 aerobe CFU in 10 ml) in an anaerobic environment, which mimics intra-abdominal sepsis. We determined pathogen-removal efficiencies for clearance of this complex mixture of anaerobic and aerobic bacteria using the continuous-flow biospleen device at a flow rate of 10 ml h−1 with the magnetic opsonins (0.5 mg ml−1) injected at 7.1 µl min−1. These studies revealed that >98% and >80% of anaerobic and aerobic bacteria were removed from blood, respectively, after only a single pass through the blood-cleansing device; it also successfully removed other pathogens, including C. albicans, S. aureus and E. coli (all at ~104 CFU ml−1), from blood with efficiencies consistently above 90% (Fig. 2a). Using fresh, whole human blood that we anticoagulated with heparin (16 units ml−1) and experimentally contaminated with S. aureus (104 CFU ml−1), we removed about 60% of the bacterial load with each pass through the device at a flow rate of 10 ml h−1. Notably, more than 90% of these Gram-positive pathogens were removed from the fresh, flowing blood after five sequential passes through nature medicine VOLUME 20 | NUMBER 10 | OCTOBER 2014 Isolation efficiency (%) 50 50 co li ro ce bic c Ae al r ce obic ca l 0 Experimental Kinetic analysis 0 2 4 Time (h) ae Untreated Biospleen + FcMBL beads 4 2 0 0 100 An re u au lb ic a S. C .a c E. s 0 ns Isolation efficiency (%) b 100 d LPS levels (OD450) a total of 16 magnetic separator channels, which then merge into a single outlet channel (Fig. 1c). To carry out blood cleansing, we continuously injected the nanomagnetic opsonins into flowing whole blood and thoroughly mixed them with pathogens by flowing them through a Kenics inline mixer28 followed by an incubation loop composed of helically coiled tubing to enhance transverse convective flux29 for 5–10 min before they entered the magnetic separator (Fig. 1d, Supplementary Fig. 2d and Online Methods). This design was based on a theoretical analysis that predicted >99% bead capture at a maximum flow rate of ~560 ml h−1 (Supplementary Fig. 3c and Online Methods), assuming the magnetic flux density gradient is >1.0 T2 m−1 with a blood channel height of ~600 µm. This analysis assumed that a single spherical pathogen (~1-µm diameter) will be bound by about 40 magnetic opsonins (128-nm diameter) on average based on our electron micros­ copy imaging results (Fig. 1b) and our past findings16. a E. coli levels (log CFU ml–1) Figure 2 Magnetic capture efficiency of the biospleen device in vitro. (a) Pathogen-isolation efficiency for various types of pathogens, including C. albicans fungi, Gram-positive S. aureus bacteria, Gram-negative E. coli bacteria and mixed populations of anaerobic or aerobic cecal microbial flora, when spiked into human blood and flowed through the bloodcleansing device at 10 ml h−1 while FcMBL-coated magnetic beads (0.5 mg ml−1) were added continuously into the device (7.1 µl min−1) (n = 3). (b) Pathogen-removal efficiencies calculated using a kinetic analysis model (dashed line) and results obtained experimentally (solid line) with the biospleen device using whole blood spiked with S. aureus (104 CFU ml−1) in vitro (n = 3). (c) Depletion of E. coli in blood by continuous cleansing in vitro using the biospleen device at a flow rate of 10 ml h−1 (n = 3; P < 0.03). (d) Depletion of LPS when LPS endotoxin spiked into fresh human whole blood (10 µg ml−1) was flowed through the biospleen device at 10 ml h−1 (n = 3). Error bars, s.e.m. 2 Time (h) 4 Biospleen + no FcMBL beads Biospleen + FcMBL beads Background 1.0 0.5 0 0 2 Time (h) 4 the ­biospleen device over 5 h, which correlates closely with a theoretical model based on Monod kinetics (Fig. 2b and Online Methods). Moreover, we also were able to significantly (P < 0.03) deplete Gramnegative pathogen levels when we continuously infused E. coli into human blood at a dose and rate (5 × 108 CFU ml−1 h−1) previously used in an animal model of sepsis30 and then flowed it through the biospleen device with magnetic opsonins for 5 h in vitro (Fig. 2c). Finally, as bacterial toxin levels in blood are a crucial contributing factor in sepsis31, we also assessed how efficiently the blood-cleansing device can remove LPS endotoxin from flowing blood. When we spiked heparinized, fresh, whole human blood (10 ml) with LPS (10 µg ml−1) and continuously flowed it through the blood-cleansing device with magnetic opsonins at 10 ml h−1 for 5 h in vitro, endotoxin levels decreased by a factor of more than 4, almost to background levels (Fig. 2d). In contrast, endotoxin levels remained consistently high when we used the biospleen device with uncoated magnetic beads (Fig. 2d), hence demonstrating the critical requirement for FcMBL in this blood-cleansing therapy. Although we efficiently removed pathogens or endotoxin bound to multiple magnetic opsonins, we removed only ~80% of unbound, single nanomagnetic beads because they have extremely small magnetic moments due to their small size (128 nm). To more effectively remove all magnetic opsonins (and prevent potential complications of the beads entering a patient’s circulation in the future), we administered larger (1-µm diameter), uncoated superparamagnetic beads along with the 128-nm, FcMBL-coated beads; these larger beads act as local magnetic field gradient concentrators that magnetize and attract the smaller beads when exposed to an external magnetic field (i.e., only when passing through the magnetic separator) (Online Methods). Experimental studies confirmed that addition of these larger beads increased magnetic capture of the smaller magnetic opsonins from ~80% to 99.6 ± 0.6% when the beads were flowed through the biospleen at 10 ml h −1 (Supplementary Fig. 4c). Extracorporeal blood cleansing in living animals To test the blood-cleansing capability of the biospleen in vivo, we linked the device (containing the static mixer, incubation loop, magnetic separator unit and a syringe pump injecting the magnetic opsonin beads) to the jugular veins of living, anesthetized rats using 1213 b Untreated Biospleen alone Biospleen + FcMBL beads E. coli in blood (log CFU ml–1) S. aureus in blood (log CFU ml–1) a 2 1 0 10 c Biospleen + uncoated beads Biospleen + FcMBL beads 3 2 1 0 0 12 Time (h) – – + – 14 – – + + 2 Time (h) 0 + – + + – – 4 + – + + + CD45 S. aureus S. aureus Biospleen e -6 IL G M α 0 -4 Kidney IL Spleen -C Lung -γ 0 50 -1 1 100 N 2 Untreated Biospleen + FcMBL beads IL Cytokine levels (pg ml–1) Untreated Biospleen + FcMBL beads Kidney SF d Spleen IF Lung S. aureus intensity (a.u.) Figure 3 In vivo blood cleansing using the biospleen blood-cleansing device in a rat bacteremia model. (a) Depletion of pathogen levels in the blood of rats injected intraperitoneally with S. aureus (5 × 108 CFU) by biospleen treatment 10 h later with (n = 6) or without (n = 3) opsonin-coated magnetic nanobeads as compared to untreated controls (n = 5). (b) Depletion of pathogen levels in the blood of rats continuously infused with E. coli (5 × 108 CFU h−1) by immediate biospleen treatment using FcMBL-coated nanobeads (solid line) or the same beads without the coating (dashed line) (biospleen with FcMBL beads, n = 3; biospleen with uncoated beads, n = 3). (c) Immunofluorescence micrographs of lung, spleen and kidney removed from normal rats (left) or rats injected intraperitoneally with S. aureus that were either untreated (middle) or treated (right) for 5 h with the biospleen device, showing reduced levels of S. aureus in each organ. Top images were stained with antibodies that recognize S. aureus (green); bottom images identify CD45+ immune cells (green); nuclei were stained with DAPI (blue) in both sets of images. Scale bar, 50 µm. (d) Quantitative analysis of the images obtained from six random fields (untreated, n = 3; biospleen with FcMBL beads, n = 3) shows that biospleen treatment significantly lowered S. aureus levels (a.u., arbitrary units) in each organ (P < 0.05). (e) Blood levels of five types of proinflammatory cytokines—granulocyte-macrophage colony–stimulating factor (GM-CSF), interferon-γ (IFN-γ), interleukin-1α (IL-1α), IL-4 and IL-6—in the circulation of rats infected with S. aureus with (black bars) or without (white bars) subsequent treatment with the biospleen device (biospleen with FcMBL beads, n = 3; untreated, n = 3). Error bars, s.e.m. medical-grade tubing and catheters (Fig. 1d and Supplementary Fig. 2d). We first tested the biocompatibility of the system (Online Methods) and found that flowing the animal’s blood continuously through the biospleen for 5 h had no detectable effect on the animal’s physiology (temperature and breathing rates), nor did it induce blood coagulation as measured by thrombin–antithrombin complex formation (Supplementary Fig. 4e). Much as we observed in our in vitro studies, the device removed ~90% of live S. aureus (Fig. 3a) and E. coli (Fig. 3b) pathogens from the septic rat’s blood within 1 h using FcMBL-coated nanomagnetic opsonins combined with the larger magnetic concentrator beads (Online Methods). Use of the biospleen device with magnetic FcMBLcoated beads also maintained significantly (P < 0.01 for S. aureus; a LPS intensity (a.u.) Survival (%) b Biospleen Untreated 100 50 0 0 1 c 2 3 Time (h) + – 4 + + 0.6 0.4 0 Lung + – + + Spleen – – Kidney + – + + CD45+ Lung Spleen 80 60 40 20 0 0 1214 1 2 3 Time (h) Kidney e Temperature (°C) d 4 5 P < 0.004 for E. coli) lower levels of pathogens in the septic rat’s circulation for the entire 5-h treatment period than those in infected animals that did not experience blood cleansing or rats treated with the biospleen either in the absence of added magnetic opsonin beads (S. aureus) or using beads that lacked the FcMBL coating (E. coli). Control biospleen studies clearly showed that the blood must be treated with magnetic FcMBL opsonins to produce effective blood cleansing as passage of blood through the magnetic separator without beads or with uncoated beads had no effect on pathogen clearance (Fig. 3a,b). Notably, histological analysis revealed that reduction of blood pathogen levels using the biospleen device with the FcMBL magnetic opsonins also resulted in a major decrease in both pathogen load and the level of inflammatory cell (CD45+) infiltrate and interstitial edema in the lung, spleen and kidney (Fig. 3c,d). This likely explains why the treated animals did not exhibit detectable symptoms of clinical distress (such as labored breathing and lower body temperature) that we observed in untreated animals. Treating the bacteremic rats with the biospleen using magnetic 0.2 5 – – Untreated Biospleen 0.8 LPS LPS – Biospleen – Respiratory rate (min–1) npg © 2014 Nature America, Inc. All rights reserved. Technical Reports 38 36 34 32 0 0 1 2 3 Time (h) 4 5 Figure 4 In vivo blood cleansing using the biospleen device in a rat acute endotoxic shock model. (a) Kaplan-Meier plot showing increased survival in rats treated with the biospleen after intravenous injection with a lethal dose of LPS endotoxin (13.6 × 106 endotoxin units per ml) (biospleen with FcMBL beads (solid line), n = 9; untreated (dashed line), n = 7). (b) Quantitative analysis of immunostained histological sections of lung, spleen and kidney (untreated, n = 3; biospleen, n = 3) (c) obtained from six random fields (n = 3). (c) Immunofluorescence micrographs of lung, spleen and kidney harvested from normal rats (left) or rats injected intravenously with LPS endotoxin only (middle) or with subsequent treatment for 5 h with the biospleen device (right) (scale bar, 50 µm). Top images were stained with antibodies that recognize LPS endotoxin (green); bottom images identify CD45 + immune cells (green); nuclei were stained with DAPI (blue) in both sets of images. (d,e) Respiratory rates (d) and body temperature (e) (untreated, n = 7; biospleen, n = 9) measured in control rats with endotoxemia (untreated, dashed line) produced by intravenous LPS injection versus similar endotoxemic animals treated with the biospleen device (solid line). Error bars, s.e.m. VOLUME 20 | NUMBER 10 | OCTOBER 2014 nature medicine npg © 2014 Nature America, Inc. All rights reserved. Technical Reports opsonins for 5 h also significantly (P < 0.05) lowered levels of multiple proinflammatory cytokines, including granulocyte-macrophage colony–stimulating factor, interferon-γ, interleukin-1α, interleukin-4 and interleukin-6 (Fig. 3e), all of which play an important role in the sepsis cascade32. This extended blood-cleansing treatment significantly (P < 0.05) decreased levels of cell-free hemoglobin in plasma as well (Supplementary Fig. 4d), which is important because this is a well-known complication of sepsis caused by the release of toxic hemoglobin from large numbers of damaged erythrocytes33. The 5-h biospleen treatment by itself only reduced the hematocrit from 40% to 30% in control animals, which was likely due to the fluidresuscitation therapy required in this animal protocol. Hence, the reduced free hemoglobin levels we measured with the blood-cleansing therapy are likely due to the lower pathogen levels, which resulted in reduced hemolysis. Finally, when we injected rats with a lethal dose of LPS endotoxin to model acute endotoxemic shock and treated them with the bio­ spleen device, we found that we could significantly (P < 0.02) improve animal survival (Fig. 4a). Treatment with the blood-cleansing device also resulted in a significant decrease in the amount of LPS endotoxin and leukocytes (P < 0.05) that infiltrated the lungs and other organs of these animals (Fig. 4b,c). Because of the potential toxicities of long-term inhalational anesthesia, all rats had to be killed at 5 h to minimize pain and distress. However, these studies revealed that although 86% of untreated rats died after intravenous injection of LPS, 89% of animals treated with the blood-cleansing device survived for the entire 5-h experiment. Also, it should be noted that endotoxemic rats treated with the biospleen device exhibited significant (P < 0.01) improvement in various physiological responses, including respiratory rate (Fig. 4d), temperature (Fig. 4e) and restoration of normal leukocyte levels (which drop rapidly to ~1.2 × 109 l−1 after LPS injection and return to 3.0 × 109 l−1 within 5 h in biospleentreated animals). DISCUSSION Despite extensive efforts at eradicating sepsis over the past century, treating patients with sepsis remains a daunting task. Our biologically inspired, extracorporeal blood-cleansing system addresses the root cause of sepsis by removing pathogens and endotoxins simultaneously. The biospleen also allows researchers and clinicians to overcome two major practical barriers that have not been previously addressed in the field. First, use of the broad-spectrum FcMBL opsonin makes it possible to rapidly treat systemic blood infections and prevent sepsis progression without having to first identify the pathogen. This is a crucial advantage because many patients with sepsis have negative blood cultures, and thus it is not possible to select an optimal antibiotic therapy. Second, the whole blood volume of a patient can be quickly processed, and multiple rounds of blood cleansing can be carried out, without producing detectable blood coagulation or significantly altering blood composition. Sepsis progresses rapidly, eventually resulting in extremely high mortality rates due to progressive development of multiorgan failure and septic shock34. Many patients with severe sepsis are already in intensive care units on extracorporeal circuits; thus, upon observing disease progression (for example, decreased kidney or lung function), a clinician could add a biospleen module like the one we describe into the existing extracorporeal circuit. Initial human clinical trials likely would be carried out with patients in this setting. In contrast, in patients with earlier-stage disease, physicians would likely require confirmation of septicemia before they could nature medicine VOLUME 20 | NUMBER 10 | OCTOBER 2014 initiate this type of extracorporeal therapy. Blood culture is not sufficient because most patients with fulminant sepsis are bloodculture negative; however, the results shown here suggest that it might in future be possible to use FcMBL magnetic opsonins to detect the presence of live and dead pathogens, as well as released toxins, within blood samples. It is also notable that sepsis caused by antibiotic-resistant bacteria is one of the most critical public health issues we face today, even in well-resourced countries35. As FcMBL binds to multiple clinical isolates of antibiotic-resistant organisms (for example, methicillin-resistant S. aureus and K. pneumonia) (Supplementary Table 1), the biospleen also could offer an effective therapeutic strategy for these patients who are not helped by existing drug therapies. Our dialysis-like blood-cleansing system allows the entire blood volume to be processed multiple times through the device during a single treatment. Hence, even if only a portion of pathogens are removed in a single pass, the pathogen load in the bloodstream can be significantly lowered by circulating the blood several times over a 24-h period. Although this therapy will not remove pathogens present within organs or abscesses, our results suggest that it will significantly reduce the spread of infectious agents to distal sites and lower levels of circulating endotoxin, toxic free hemoglobin and inflammatory cytokines. All of these factors are key contributors to disease progression, and thus, this treatment could critically extend the time available to identify the nidus of infection and initiate optimal antibiotic therapy. In fact, wide-spectrum or targeted antibiotic therapy can be coadministered with this blood-cleansing therapy because the magnetic FcMBL opsonins bind to both dead and live pathogens. The ability of the biospleen device to efficiently capture pathogens also provides a potential way to rapidly collect large numbers of living infectious agents; therefore, this device could greatly accelerate pathogen identification and antibiotic susceptibility determination. Finally, this blood-cleansing device is a platform technology in that it could be used to remove proteins (such as cytokines or autoantibodies) as well as other types of cells (such as circulating cancer cells, stem cells, fetal cells in maternal circulation, etc.) from the whole blood volume of patients simply by coating the magnetic beads with appropriate cell- or protein-specific ligands. Methods Methods and any associated references are available in the online version of the paper. Accession codes. KJ710775 Note: Any Supplementary Information and Source Data files are available in the online version of the paper. Acknowledgments This work was supported by Defense Advanced Research Projects Agency grant N66001-11-1-4180 and contract HR0011-13-C-0025, Department of Defense/ Center for Integration of Medicine and Innovative Technology and the Wyss Institute for Biologically Inspired Engineering at Harvard University. We thank M. Montoya-Zavala and D. Breslau for micromachining of the blood-cleansing microdevice and technical support; P. Snell and J. Tomolonis for microbiology assistance; A. Schulte for assistance in biospleen experiments; J. Weaver for help with scanning electron microscopy; R. Betensky for statistical analysis assistance; and A. Onderdonk, M. Puder, A. Nedder and their teams for assistance in developing the rat cecal contents sepsis model. We thank J. Fiering and his team for helpful discussions during the early phase of this project. J.H.K. is a recipient of a Wyss Technology Development Fellowship from the Wyss Institute and a professional development postdoctoral award from Harvard University. Scanning electron microscopy images were obtained at the Center for Nanoscale Systems at Harvard University, supported by the National Science Foundation under award no. ECS-0335765. 1215 Technical Reports AUTHOR CONTRIBUTIONS J.H.K. designed and performed blood-cleansing experiments with assistance from R.M.C., J.B.B., N.G., A.R.G., A.D. and A.W., and analyzed the data and prepared the manuscript. K.D. contributed to the design and integration of the biospleen device. J.H.K., C.W.Y., A.R.G. and H.T. established the rat sepsis models, and J.H.K. and A.R.G. conducted animal studies with help from T.M., A.M. and A.J.; C.W.Y. designed the prototype biospleen device and obtained preliminary data. M.S. and A.L.W. designed, engineered and produced FcMBL with assistance from M.J.R., J.B.B. and A.K.; M.S., M.J.C. and M.R. performed blood analysis for quantitating LPS levels with help from N.G. and helped establish an endotoxemia model in rats. T.M.V. fabricated devices, performed scanning electron microscopy and assisted with conducting studies. K.T. performed experiments to characterize FcMBL versus native MBL. D.E.I. led efforts to design the device and the opsonin, assisted in data analysis and helped write the manuscript. COMPETING FINANCIAL INTERESTS The authors declare no competing financial interests. npg © 2014 Nature America, Inc. All rights reserved. Reprints and permissions information is available online at http://www.nature.com/ reprints/index.html. 1. Bone, R.C. The pathogenesis of sepsis. Ann. Intern. Med. 115, 457–469 (1991). 2. Cohen, J. The immunopathogenesis of sepsis. Nature 420, 885–891 (2002). 3. Hotchkiss, R.S. & Karl, I.E. The pathophysiology and treatment of sepsis. N. Engl. J. Med. 348, 138–150 (2003). 4. Kumar, A. Optimizing antimicrobial therapy in sepsis and septic shock. Crit. Care Clin. 25, 733–751 (2009). 5. Anonymous. Focus on sepsis. Nat. Med. 18, 997 (2012). 6. O’Brien, T.F. Global surveillance of antibiotic resistance. N. Engl. J. Med. 326, 339–340 (1992). 7. Lineaweaver, W. et al. Topical antimicrobial toxicity. Arch. Surg. 120, 267–270 (1985). 8. Garnacho-Montero, J. et al. Timing of adequate antibiotic therapy is a greater determinant of outcome than are TNF and IL-10 polymorphisms in patients with sepsis. Crit. Care 10, R111 (2006). 9. Stoll, B.J. et al. Late-onset sepsis in very low birth weight neonates: the experience of the NICHD Neonatal Research Network. Pediatrics 110, 285–291 (2002). 10. Neff, L.P. et al. Extracorporeal organ support following trauma: the dawn of a new era in combat casualty critical care. J. Trauma Acute Care Surg. 75, S120–S129 (2013). 11. Angus, D.C. Drotrecogin alfa (activated). a sad final fizzle to a roller-coaster party. Crit. Care 16, 107 (2012). 12. Chuang, Y.-C., Chang, S.-C. & Wang, W.-K. High and increasing Oxa-51 DNA load predict mortality in Acinetobacter baumannii bacteremia: implication for pathogenesis and evaluation of therapy. PLoS ONE 5, e14133 (2010). 13. Gijs, M.A.M. Magnetic bead handling on-chip: new opportunities for analytical applications. Microfluid. Nanofluidics 1, 22–40 (2004). 14. Kang, J.H., Choi, S., Lee, W. & Park, J.-K. Isomagnetophoresis to discriminate subtle difference in magnetic susceptibility. J. Am. Chem. Soc. 130, 396–397 (2008). 15. Kang, J.H. et al. A combined micromagnetic-microfluidic device for rapid capture and culture of rare circulating tumor cells. Lab Chip 12, 2175 (2012). 1216 16. Xia, N. et al. Combined microfluidic-micromagnetic separation of living cells in continuous flow. Biomed. Microdevices 8, 299–308 (2006). 17. Yung, C.W., Fiering, J., Mueller, A.J. & Ingber, D.E. Micromagnetic-microfluidic blood cleansing device. Lab Chip 9, 1171 (2009). 18. Dommett, R.M., Klein, N. & Turner, M.W. Mannose-binding lectin in innate immunity: past, present and future. Tissue Antigens 68, 193–209 (2006). 19. Sheriff, S., Chang, C.Y. & Ezekowitz, R.A.B. Human mannose-binding protein carbohydrate recognition domain trimerizes through a triple alpha-helical coiled-coil. Nat. Struct. Biol. 1, 789–794 (1994). 20. Lo, K.M. et al. High level expression and secretion of Fc-X fusion proteins in mammalian cells. Protein Eng. 11, 495–500 (1998). 21. Neth, O. et al. Mannose-binding lectin binds to a range of clinically relevant microorganisms and promotes complement deposition. Infect. Immun. 68, 688–693 (2000). 22. Townsend, R., Read, R.C., Turner, M.W., Klein, N.J. & Jack, D.L. Differential recognition of obligate anaerobic bacteria by human mannose-binding lectin. Clin. Exp. Immunol. 124, 223–228 (2001). 23. Takahashi, K., Ip, W.E., Michelow, I.C. & Ezekowitz, R.A.B. The mannose-binding lectin: a prototypic pattern recognition molecule. Curr. Opin. Immunol. 18, 16–23 (2006). 24. Orsini, J. et al. Microbiological profile of organisms causing bloodstream infection in critically ill patients. J. Clin. Med. Res. 4, 371–377 (2012). 25. Castanheira, M., Farrell, S.E., Krause, K.M., Jones, R.N. & Sader, H.S. Contemporary diversity of β-lactamases among Enterobacteriaceae in the nine U.S. census regions and ceftazidime-avibactam activity tested against isolates producing the most prevalent β-lactamase groups. Antimicrob. Agents Chemother. 58, 833–838 (2014). 26. Morrow, B.J. et al. Activities of carbapenem and comparator agents against contemporary US Pseudomonas aeruginosa isolates from the CAPITAL surveillance program. Diagn. Microbiol. Infect. Dis. 75, 412–416 (2013). 27. Mebius, R.E. & Kraal, G. Structure and function of the spleen. Nat. Rev. Immunol. 5, 606–616 (2005). 28. Meijer, H.E.H., Singh, M.K. & Anderson, P.D. On the performance of static mixers: a quantitative comparison. Prog. Polym. Sci. 37, 1333–1349 (2012). 29. Jayakumar, J.S., Mahajani, S.M., Mandal, J.C., Iyer, K.N. & Vijayan, P.K. CFD analysis of single-phase flows inside helically coiled tubes. Comput. Chem. Eng. 34, 430–446 (2010). 30. Pass, L.J., Schloerb, P.R., Pearce, F.J. & Drucker, W.R. Cardiopulmonary response of the rat to Gram-negative bacteremia. Am. J. Physiol. 246, H344–H350 (1984). 31. Wiesel, P. et al. Endotoxin-induced mortality is related to increased oxidative stress and end-organ dysfunction, not refractory hypotension, in heme oxygenase-1– deficient mice. Circulation 102, 3015–3022 (2000). 32. Netea, M.G. Proinflammatory cytokines and sepsis syndrome: not enough, or too much of a good thing? Trends Immunol. 24, 254–258 (2003). 33. Janz, D.R. et al. Association between cell-free hemoglobin, acetaminophen, and mortality in patients with sepsis: an observational study. Crit. Care Med. 41, 784–790 (2013). 34. Rello, J. Severity of pneumococcal pneumonia associated with genomic bacterial load. Chest 136, 832 (2009). 35. Levy, S.B. & Marshall, B. Antibacterial resistance worldwide: causes, challenges and responses. Nat. Med. 10, S122–S129 (2004). 36. Takahashi, K. et al. Mannose-binding lectin and its associated proteases (MASPs) mediate coagulation and its deficiency is a risk factor in developing complications from infection, including disseminated intravascular coagulation. Immunobiology 216, 96–102 (2011). VOLUME 20 | NUMBER 10 | OCTOBER 2014 nature medicine ONLINE METHODS npg © 2014 Nature America, Inc. All rights reserved. Biomimicry of human blood opsonins. To develop a viable clinical therapy using this blood-cleansing approach, we studied how the living spleen naturally works in combination with the innate immune system to remove pathogens from blood with high efficiency (for example, virtually all fungal pathogens are captured by the spleen within one minute after intravenous injection in rabbits; Supplementary Fig. 5). In the human body, pathogen capture is mediated by natural blood opsonins, such as mannose-binding lectin (MBL). MBL is a soluble, circulating, calcium-dependent (C-type) lectin that binds to terminal mannose and fucose residues on branched oligosaccharides that are expressed on the surface of over 90 different pathogen species (including Gram-negative and Gram-positive bacteria, fungi, protozoa and viruses) as well as endotoxin but not on mammalian cells37–39. The MBL-opsonized pathogens are then bound and phagocytosed by macrophages in the spleen as flow slows and the blood trickles from the arterial circulation through the marginal zones and red-pulp cords and into venous sinuses. Fabrication of magnetic opsonins. Native human MBL was genetically engineered to create the Fc-fusion opsonin protein FcMBL, which consists of the tripeptide alanine-lysine-threonine (AKT) followed by the hinge, CH2 and CH3 domains of human IgG1 Fc (residues E216–G446 with two point mutants, C220S and N297D), followed by a single alanine linker fused to the neck and carbohydrate-recognition domain (CRD) region of human MBL2 (residues 81–228 of the mature protein). cDNA for this construct (GenBank accession code: KJ710775) was cloned downstream of the Ig κ–light chain leader sequence in an HTLV-driven expression plasmid for transient transfection in HEK293F Freestyle cells (Invitrogen) and into an HTLV expression plasmid with DHFR under the control of an IRES for stable transfection into CHODG44 cells (Invitrogen), following the manufacturer’s suggested protocols. FcMBL was purified from conditioned medium using protein A Sepharose (GE Healthcare Life Sciences) on an AKTA-Avant system (GE Healthcare Life Sciences). Using expression of the FcMBL from CHO cell lines and a single-step, Fc-based purification technique, we were able to produce large amounts (0.112 g liter−1) of single band–quality FcMBL that were efficiently immobilized on the magnetic nanobeads. Purity was confirmed to >95% by SDS-PAGE electrophoresis, and functionality was confirmed by the ability to bind to mannan-coated plates. By fusing the tripeptide AKT into the N terminus for site-specific biotinylation, we are able to orient coupling of the FcMBL to streptavidin-coated superparamagnetic beads (Ademtech AS) so that the N-terminal Fc hinge attached to the bead surface and the CRD faced outward. The N-terminal amine of FcMBL was modified to an aldehyde using pyridoxal5-phosphate monohydrate treatment as previously described 40,41; the AKT tripeptide at the N-terminus has been identified as an optimal sequence for this reaction42. Briefly, FcMBL protein at 4 mg ml−1 was incubated overnight at room temperature in 7.5 mM PLP (Alfa Aesar) in 50 mM sodium phosphate buffer, pH 6.4. The PLP-treated protein was dialyzed 3 times against 50 mM sodium phosphate, pH 6.4, and then incubated overnight in the presence of 25 µg of aminooxy biotin (Biotum) per 1 mg of FcMBL. Biotinylated FcMBL was coupled to 128-nm streptavidin-coated magnetic beads (Ademtech AS) at a ratio of 25 µg of protein mg−1 of beads for 30 min at room temperature. To create the magnetic opsonin beads, streptavidin-coated superparamagnetic nanobeads (128 nm; Ademtech Inc.) were incubated with biotinylated FcMBL (25 µg protein mg−1 beads) for 30 min at room temperature, and remaining unbound streptavidin sites on the beads were blocked with polyethylene glycol (PEG)-biotin (1 kDa PEG; Nanocs Inc.). The beads (5 mg beads ml−1) were resuspended in 1% BSA in PBS and stored at 4 °C until use. The PEG was used to minimize nonspecific binding of serum proteins and blood components to the magnetic beads as it has been shown to effectively prevent nonspecific binding of molecules in vitro43 and in vivo44,45, in addition to already being used for this purpose in multiple FDA-approved blood-compatible clinical products. In addition, we confirmed that the FcMBL magnetic beads do not deplete blood cells nonspecifically (Supplementary Fig. 4b) when we treated human blood with the biospleen devices using the same conditions as in the rat study, and then carried out complete blood counts using a commercial CBC machine. In addition, we confirmed that FcMBL failed to induce significant coagulation or complement activation in blood versus native doi:10.1038/nm.3640 recombinant human MBL (rhMBL) (Supplementary Fig. 1c,d, using previ­ ously described methods46–48 to quantitate the thrombin-like activity and complement-binding activity of MBL, respectively (rhMBL manufactured by Enzon Pharmaceuticals Inc. was provided by K. Takahashi (Massachusetts General Hospital)49). Moreover, when FcMBL-coated beads were incubated with 50 µg of DNA isolated from salmon sperm, they did not exhibit any detectable binding. This is in stark contrast to wild-type MBL, which has been shown to bind to DNA50. Generic ligands for pathogen binding have been developed in the past. For example, the synthetic organic molecule zinc(II)-dipicolylamine binds to enriched anionic phospholipids of bacterial cell walls by forming coordination complexes51,52. However, their binding capacity is limited: they do not bind to as wide a range of pathogens, toxins and viruses, and their binding strength is much weaker than that exhibited by MBL for its microbial cell-surface ligands53. The major advantage of the FcMBL opsonin is that the clinical development path should be much more straightforward as it is an engineered version of a human blood protein. FcMBL also can be produced in large quantities at low cost using protein A purification, and the Fc moiety has been previously used in multiple US Food and Drug Administration–approved protein therapeutics54. Biospleen device design and fabrication. The microdevice developed in our past studies had insurmountable limitations that prohibited its clinical application for sepsis therapy55,56. First, as the identity of the infectious agent is not known when a patient first presents with symptoms of systemic infection, it is not possible to select appropriate microbe-specific antibodies, and the cost of using numerous antibodies directed against virtually all potential pathogens would be prohibitive. Second, endotoxins that are released by many microbes when killed by endogenous immune responses or antibiotic therapies are a major trigger for the inflammatory cascade that leads to sepsis57, yet these toxins were not cleared with the initial method. Third, the microfluidic channel design used in the original device could not support the high flow rates (liters per hour) necessary for clinical use without causing blood loss or dilution. The dimensions of the channels and slits were initially based on our previous device design58, which enabled high (>99%) magnetic bead capture efficiencies but only at very low flow rates (<0.0012 liter h−1). To further optimize system function, magnetic flux density gradients across the blood channel were estimated using computational simulations59 (Supplementary Fig. 3b), and the distance the magnetic bead–bound pathogens would travel perpendicular to the blood channel while flowing through the device was calculated as a function of the magnetic flux density gradient and flow rate (Supplementary Fig. 3c). We then redesigned the channels to obtain a removal efficiency of >99% using these results. The channel width (2 mm) and overall device dimensions were determined by the size of the permanent magnets (50 mm in length) and capacity of the micro-milling machine. We calculated the magnetic force acting on a pathogen bound with magnetic opsonin beads and how fast they are pulled by the magnetic flux density gradient of the permanent magnets (Nd2Fe14B, BY042SH; magnetization, M = 106 A m−1) (KJ Magnetics, Jamison, PA, USA). The magnetic drag velocity of a pathogen bound with nanomagnetic opsonin beads was determined by the equation59 Vmag (x ) = 2R2 ∆ c B(x) 9 m0h (1) where Vmag(x) is the magnetic drag velocity, R is a radius of a magnetic bead, ∆χ is difference in magnetic susceptibility between magnetic beads and solution, µ0 is the vacuum permeability, η is the dynamic viscosity and B(x) is the magnetic flux density gradient. Then, Dt can be calculated by equations (2) and (3) Dt = Vmag (x ) × tres tres = whL Q (2) (3) where tres is the residence time of the magnetic beads in the channel while they are flowing through the device; w, h and L are width, height and length of the channel; and Q is the flow rate (ml h−1). The saline channel of the ­magnetic nature medicine © 2014 Nature America, Inc. All rights reserved. npg separator unit of the blood-cleansing microdevice is 340 µm in height, whereas the blood channel is 600-µm high; thus, the magnetic force acting on the nanomagnetic opsonin and magnetically labeled pathogens is much enhanced, resulting in further increased pathogen removal efficiency. Both the blood and saline channels are 2 mm wide. To enhance flow throughput, the single inflow channel into the device was designed so that it undergoes 4 branches before entering a total of 16 magnetic separator channels, which then join in a symmetrical quadruple branching pattern to merge into a single outlet channel (Fig. 1c (top) and Supplementary Fig. 2a). Sheets (1.58 mm) of aluminum (Fig. 1c), medical-grade polysulfone (Supplementary Fig. 2b) or PMMA (Supplementary Fig. 2c) were micromachined using a Microlution 5100-S (Microlution Inc.) machine, and the upper and lower surfaces of the device containing the fluidic channels were covered with thin layers of transparent PMMA adhesive tape (76-µm thickness, McMaster-Carr); similar isolation efficiencies were obtained with all devices from three types of materials. As shown in Supplementary Figure 2a, two upper and two lower connector blocks were then aligned with the ends of the taped device using dowel pins, and assembled into a single integrated device that was held together by socket head cap screws (McMaster-Carr, 92290A117). Four O-rings also were located between through-holes on the connector blocks and the device to prevent leakage, and luer lock connectors were inserted into each port to create a bio­ spleen device that can connect to fluid sources (Supplementary Fig. 2a,c) and capture magnetic opsonins flowing through its slits when stationary magnetics are placed on its top surface (Supplementary Fig. 2a). Similar results were obtained with the aluminum (Fig. 1c) and polymeric devices (Supplementary Fig. 2b,c) in our in vitro studies, and we used the aluminum biospleen device for treating septic rats. The assembled blood-cleansing magnetic separator unit and attached tubing and connectors were sterilized using ethylene oxide before use in animal studies. One of the key challenges of the blood-cleansing microdevice is that we need to continuously mix blood with nanomagnetic opsonins and then incub­ ate them without damaging blood in laminar flow. We incorporated a Kenics static inline mixer60 into the blood-cleansing microdevice system (Fig. 1d, Supplementary Figs. 2d and 3a) to ensure that the continuously injected nanomagnetic opsonins were thoroughly mixed with blood so that pathogens in blood could bind to the FcMBL nanomagnetic beads. After passing through the inline mixer, pathogens and nanomagnetic opsonins need to be mixed in the flowing blood before it reaches the magnetic separator unit of the bloodcleansing microdevice. To accomplish this, we inserted an incubation loop composed of helically coiled tubing (internal diameter, 1.58 mm; 100 cm long) between the mixer and the separator unit (Fig. 1d and Supplementary Fig. 2d) to enhance transverse convective flux along the tubing61. This flux is determined by the Dean number (De = Re (r Rc−1)0.5, De = 0.17) in a curved duct, where Re is the Reynolds number, r is an inner radius of the tubing (1.59 mm) and Rc is a coil radius (6.3 mm). We confirmed that incorporation of the mixer with this loop design increased binding and removal efficiencies for S. aureus bacteria from ~40% to 99% when they were spiked (104 CFU ml−1) into banked whole blood or buffer flowing at 10 ml h−1 (Supplementary Fig. 3a). Fluidic resistance measurements. It is critical to control the fluidic resistance of the extracorporeal system because changes in resistance may affect blood pressure of patients and cause blood coagulation due to fluidic resistance–driven shear stress62. We measured the fluidic resistance across the blood-cleansing microdevice system, including tubing and an inline Kenics mixer (FMX8214; Omega Engineering Inc.), at a high flow rate up to 1.6 liter h−1 of saline in a single device and determined this to be 15.5 mm Hg, which is compatible with the pressure drops of other extracorporeal devices63 (Supplementary Fig. 3e). Our final design also ensured that shear stresses within the blood channel were low (<3 dyne cm−2; Supplementary Fig. 4a) and well within the physiological range (<15 dyne cm−2)64. We also demonstrated high throughput by circulating 200 ml of human blood (CPDA-1 anticoagulated blood-bank blood) through a single blood-cleansing microdevice system at a flow rate of 1.25 liter h−1 for over 25 h without any evidence of blood clotting or coagulation. After 25 h of circulation through the device, the blood was passed through a 40-µm cell strainer (Cat. No.: 352340, BD Falcon, BD Biosciences), and no blood clots were seen. Units of human blood nature medicine were obtained from Brigham and Women’s Specimen Bank (protocol number M20403-101). Analysis of system function also confirmed that blood loss and dilution were minimal using this design because the flow through the saline channel was static during the cleansing step, and its total volume (<1 ml) was significantly smaller than the volume of blood that flowed continuously through the blood channel (Supplementary Fig. 3f). Blood-cleansing device operation. Because we planned to carry out proof-ofconcept studies in rat sepsis models, we carried out pathogen-clearance studies in vitro and in vivo using 10 mL of blood flowing at 10 ml h−1 based on the known volume and flow rate of blood in a rat65,66. In the in vitro studies, bacteria (104 CFU ml−1) were spiked into 10 ml of human whole blood and flowed through the device. Magnetic opsonins (0.5 mg ml−1 in saline) were introduced continuously (7.1 µl min−1) into the blood path through a three-way connector (Supplementary Fig. 2d), and then the blood was passed through the spiral mixer and coiled tubing to ensure complete pathogen binding before entering the magnetic separator of biospleen device (Fig. 1d). After passing through the device, the cleansed blood that exits the outlet then returns back to the inlet, producing a closed, extracorporeal blood-circulation loop. Quantitation of pathogen depletion from blood was determined by collecting small (100 µl) blood samples over time and carrying out quantitative blood culture assays to determine pathogen colony-forming units. Controls included running the contaminated blood sample through the device without the magnets or with magnets but without adding the magnetic opsonin beads, and neither produced significant pathogen removal. The capture efficiency was lower in blood than in buffer likely because of the high numbers of blood cells (~109 red blood cells ml−1, ~108 platelets ml−1 and 106 white blood cells ml−1) and higher fluid viscosity that decrease the efficiency of magnetic capture. But in studies with various samples of human blood that had differences in hematocrit over the normal range (40–55%), we did not observe significant differences in capture efficiency. This isolation efficiency can be further improved by increasing magnetic flux density gradients65,67 and modifying the channel geometry to utilize the stronger magnetic forces. Moreover, as flow through this device scales linearly, the biospleen can be adapted for use at rates as high as 10 or more liters per hour, if required for large animal or human studies, simply by connecting multiple (>8) separator units in parallel. Taken together, these studies show that the biospleen device can efficiently cleanse flowing whole blood of various types of pathogens, as well as endotoxins, without significantly altering blood composition or inducing coagulation in vitro. Notably, we were able to obtain near-complete removal of pathogens and endotoxins that were magnetically opsonized because multiple nanobeads bind to each pathogen or endotoxin-containing pathogen fragment. However, we only removed about 80% of unbound, single nanomagnetic beads because they have extremely small magnetic moments due to their small size (128 nm). In future therapeutic applications, it would be advantageous to minimize release of unbound magnetic nanobeads into the individual’s circulation; therefore, we added larger (1-µm diameter), uncoated superparamagnetic beads to the blood (50 µl in 10 ml of blood) along with the 128-nm, FcMBL-coated beads. These larger beads act as local magnetic field gradient concentrators that magnetize and attract the smaller, 128-nm FcMBL beads flowing in the liquid when exposed to an external magnetic field (i.e., only when passing through the magnetic separator). This results in the formation of larger bead aggregates, which increases their effective magnetic radius and, thus, their ability to be magnetically cleared from the blood channel in the biospleen device68. To quantitate removal efficiency of unbound, single nanomagnetic beads (128 nm), we labeled the beads with fluorescein isothiocyanate (FITC) and flowed them through the device in saline with larger (1 µm, Dynabeads MyOne; Invitrogen), uncoated magnetic flux concentrator beads at a flow rate of 10 ml h−1 and then measured the fluorescent signal of the sample collected from the outlet (Supplementary Fig. 4c). Experimental studies confirmed that addition of these larger beads increased magnetic capture of the smaller magnetic opsonins from ~80% to 99.6 ± 0.6% when flowed through the biospleen at 10 ml h−1 (Supplementary Fig. 4c). If additional bead capture is required in the future, a component capable of continuously removing magnetic nanobeads67 could be incorporated downstream in the circuit before blood is returned to the individual. doi:10.1038/nm.3640 npg © 2014 Nature America, Inc. All rights reserved. Microbial cell culture. To model sepsis induced by intestinal injury, we fed Wistar rats a diet of raw ground beef and collected their cecal contents, which contain a complex mixture of aerobic and anaerobic bacteria, including E. coli and B. fragilis. Aerobic bacterial cultures (i.e., Staphylococcus aureus, Bacillus subtilis, Pseudomonas aeruginosa, Klebsiella pneumoniae and Escherichia coli) and aerobic microbes isolated from rat cecal contents as described 69 were grown in LB medium at 37 °C, shaking at 200 rpm. Anaerobic bacteria isolated from rat cecal contents69 were cultured on LB agar plates at 37 °C in an anaerobic chamber (Coy Laboratories). Fungal cultures (i.e., Candida albicans, Pichia pastoris, Candida parapsilosis and Saccharomyces cerevisiae) were grown in YPD medium at 30 °C, shaking at 200 rpm. Cell concentrations were measured with OD600 spectrometry (M5 SpectraMax, Molecular Devices); cell concentrations below 104 cells ml−1 (OD600 ~10−4) were quantified by culture on blood agar plates. Stock aliquots containing a defined number of cells per tube were obtained by FACS (BD FACS Aria, BD Biosciences) and used in various depletion and separation assays to enhance experimental consistency. Cell-binding and capture assays. Initial testing of the pathogen-­capturing efficiency of the magnetic opsonins was carried out by spiking 104 CFU S. aureus into 10 ml of heparinized, whole human blood and then adding 50 µl of the FcMBL beads (5 mg ml−1). Human blood was obtained from healthy donors with informed consent in accordance with the Harvard University Faculty of Medicine Committee on Human Studies and the Defense Advanced Research Projects Agency (protocol number M20403-101). After mixing in an inverting mixer (Invitrogen) for 20 min, the blood samples were flowed through the biospleen magnetic separator unit at 10 ml h−1, and samples of the collected blood (100 µl) were plated on blood agar plates for quantitating pathogen-capture efficiency. In subsequent studies, aliquots of C. albicans, S. cerevisiae, S. aureus, E. coli and cecal contents with a known pathogen concentrations stored at −80 °C were thawed and diluted to 104 CFU ml−1 and used to test magnetic opsonin binding and capture. Characterization of bead binding to different pathogens was carried out using 500-µl samples of each pathogen strain diluted in 0.1 M Tris buffered saline, pH 7.4, with 0.05% Tween 20 (TBST) (Boston BioProducts) buffer with 5mM CaCl2 or diluted in human blood with 5 mM CaCl2 and incubated for 20 min with FcMBL-coated nanobeads (5 mg ml−1) in a tube with agitation on an inverting mixer (Invitrogen); a positive control for each strain was taken before the beads were added. The solution containing the beads and pathogens was then flowed through the biospleen device at 10 ml h−1 to determine the efficiency of removal of magnetic opsonin-bound pathogens. This flow rate and the volume (3.5 ml) used to prime the system with saline were determined based on the previous reports using rats of similar weight to those we eventually used for our in vivo studies65,66. Aliquots of the fluid remaining in each tube after magnetic removal of the bound pathogens were plated out onto blood agar plates to determine the number of microbes that eluded binding and capture. Fungi were plated out onto dextrose agar plates and cultured at 30 °C; bacteria were cultured on LB agar (or blood agar) plates at 37 °C. Colonies were counted 24 h later to determine the pathogen concentration in the positive control and depleted samples. Binding efficiencies were determined by measuring the ratio of bound to unbound pathogens in each experiment. Endotoxin-binding assay. Lipopolysaccharide (LPS) endotoxin extracted from E. coli was spiked into TBST buffer with 5 mM CaCl2 at concentrations of 20, 2, 0.2 and 0 µg ml−1, and the capture capacity of the magnetic beads coated with FcMBL opsonins was assessed by a sandwich immunoassay using the FcMBL magnetic beads and the horseradish peroxidase (HRP)-labeled FcMBL protein diluted in 0.1 M TBST. The colorimetric reaction was detected with the Pierce 1-Step TMB Substrate (Thermo Scientific) according to the manufacturer’s protocol, and results were read at 450 nm. For measuring LPS levels in blood, we collected blood samples (100 µl) from the biospleen circuit and performed the sandwich immunoassay as described above. The background signal of heparinized, fresh human whole blood is around 0.15 arbitrary units at 450 nm (Fig. 2d). Rat sepsis models. All animal protocols were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of Boston Children’s doi:10.1038/nm.3640 Hospital and Harvard Medical School, and the Animal Care and Use Review Office of the US Army Medical Research and Material Command Office and Department of Defense. Rats (Wistar 12-week-old male, 350 g) used in this study were obtained with double catheters placed in their jugular veins from Charles River Laboratories. Before initiating experiments with infected animals, we tested the biocompatibility of the biospleen device by connecting it to jugular vein catheters of a living rat we randomly selected and circulated the animal’s blood through the extracorporeal circuit and the biospleen device for 5 h. Healthy catheterized rats (n = 3) were anesthetized with isoflurane using a nose cone and two 23-gauge needles were inserted into the catheters and then connected to male luer-barb fitting connectors (EW-45504-00, Cole-Parmer) at each end of the biospleen tubing loop (Tygon 1/16” ID, PVC) (Fig. 1d and Supplementary Fig. 2d). We continuously injected saline (90 ml kg−1 day−1) containing heparin (50 unit kg−1 h−1) as required by IACUC protocols to prevent dehydration of the rats and coagulation of blood while circulating blood through the biospleen circuit. Heparin administration is also standard medical practice in humans receiving extracorporeal hemodialysis or extracorporeal membrane oxygenation. To study the utility of the blood-cleansing device in an animal with a systemic bacterial infection, we used the previously described sepsis model using Wistar rats70 we randomly selected in which 1 ml of PBS containing S. aureus (ATCC No.: 12598, 5 × 108 CFU ml−1) was injected into the peritoneal cavity. In this model, pathogen levels in blood increased 3–4 h after injection and peaked around 10 h (ref. 70). At 10 h, we anesthetized the rats, connected their jugular catheters to the tubing of the biospleen device and initiated peristaltic pumping at a flow rate of 10 ml h−1; the syringe pump was used to inject saline containing heparin (50 unit kg−1 h−1), the magnetic opsonins (0.5 mg ml−1) and larger magnetic flux concentrator beads (0.5 mg ml−1) at a flow rate of 7.1 µl min−1. Every hour, we collected small blood samples (300 µl) through a 3-way valve in the circuit, which were used to measure cytokines (Bio-Plex Pro kit, Bio-Rad) and plasma hemoglobin (using a colorimetric assay) and to carry out blood-culture analysis in a blinded fashion (n = 3 for biospleen treatment; n = 3 for the untreated group). The control rats randomly selected were either untreated or treated with biospleen without adding magnetic opsonins while receiving saline and heparin as described above. In a separate study, we used the E. coli Gram-negative bacteremia model that was reported previously71,72 by injecting a bolus of saline containing 5 × 108 CFU ml−1 of E. coli followed by continuously infusing 5 × 108 CFU ml−1 of E. coli (ATCC No. 8739) for 5 h while treating the rats with the biospleen device (n = 3). In a control experiment (n = 3), we treated the rats with the same experimental biospleen setup except that the magnetic beads were not coated with FcMBL (i.e., they were only coated with PEG). In the bacteremia animal models, the experimental and control animals received the same continuously infused volume of saline and heparin for continuous fluid resuscitation. The only difference was that the experimental animals received FcMBL beads, whereas the controls either received no beads or beads without FcMBL attached. Thus, the fluid due to bead administration (~250 µl of the bead stock solution for 5 h) was minimal compared to the large volume of saline (90 ml kg−1 day−1 into rats of 350-g weight; 6.6 ml total volume in 5 h) that we have to infuse into the animals over time during the extracorporeal treatment to keep them hydrated. To determine the ability of the biospleen to treat animals with endotoxemia, we injected LPS endotoxin (13.5 × 106 endotoxin units ml−1) extracted from E. coli (0111:B4)73 into the jugular catheters of the rats (n = 7 for the untreated group; n = 9 for biospleen treatment) and then immediately connected the rats to the blood-cleansing system. We continuously infused saline with heparin at the same flow rate in the untreated animals as we did for the biospleen treatment group to keep them hydrated, so the only difference was that blood was circulated through the biospleen device in the experimental group. Blood sample collection and analysis were similar to that described above. Rats also were observed continuously to monitor symptoms of morbidity, as measured by body temperature, respiratory rates, mucous membrane color, perfusion quality and labored breathing. After 5 h of blood cleansing, the rats were humanely euthanized according to the protocol and the major organs were harvested, fixed in 10% formalin, cryosectioned and processed for immunohistochemistry and HE staining in a blinded fashion as previously nature medicine © 2014 Nature America, Inc. All rights reserved. described74. The levels of S. aureus present in the organs were quantified using computerized microfluorimetry to measure fluorescence intensity (green) in six random fields out of samples from three rats from each group. These results strongly support our hypothesis that removing pathogens and endotoxins from septic blood should be able to significantly attenuate sepsis progression that leads to organ dysfunction and mortality and, thus, potentially save the lives of septic patients. It is important to note, however, that we were not able to test the long-term efficacy of the biospleen device because our rat studies were limited to 5-h duration by our animal protocol (after which surviving animals had to be euthanized) due to the use of isoflurane anesthesia, which can produce significant respiratory suppression in rats. In future large-animal and human studies, this blood-cleansing device would likely be used continuously in combination with broad-spectrum antibiotics until clinical response is observed. Hematology. To determine whether treatment with the biospleen significantly changes blood chemistry, blood samples (300 µl) were collected from rats after being connected to the extracorporeal circuit containing the blood-cleansing device for 5 h. The blood samples were centrifuged (×500 g) for 15 min and the supernatant-containing, platelet-depleted plasma was collected. Thrombin– anti-thrombin complexes (TAT) were measured using a sandwich ELISA kit according to the manufacturer’s instructions (Wuhan EIAAB Science). Red blood cell lysis also was analyzed to determine release of free hemoglobin in plasma. Plasma hemoglobin was measured by a hemoglobin colorimetric assay (Cayman Chemical, Item #700540). Plasma was diluted 1:10 in hemoglobin detector and incubated at room temperature for 15 min. Absorbance at 575 nm was subsequently measured and concentrations were calculated by reference to an eight-point hemoglobin standard curve (Cayman Chemical, Item# 700453). Plasma cytokine levels were measured using the Bio-Plex Pro Rat Cytokine 24-plex assay (Bio-Rad) according to the kit directions. Briefly, plasma samples were diluted 1:4 before mixing with coupled beads for cytokine measurement. All plate washes were performed with a Bio-Plex Pro magnetic wash station (Bio-Rad). Samples were measured using the Bio-Plex 3D Suspension array system (Bio-Rad) and the accompanying Bio-Plex Manager software was used for all data acquisition and analysis. Kinetic model for estimating pathogen clearance. The systemic pathogen clearance efficiency from septic blood can be calculated by the simplified mathematical model of Monod kinetics56 npg dC p dt Fj  = C p  m −  V   (4) where Cp is the concentration of pathogen in blood, t is time, µ is the growth rate of the pathogen (for example, S. aureus: 0.13 h−1)75, F is the flow rate (ml h−1), ϕ is the pathogen-removal efficiency in a single-round pass and V is the blood volume. The pathogen-removal efficiency was plotted in Figure 2b. Scanning electron microscopy. All scanning electron microscopy images were collected using a Zeiss FESEM Supra at the Center for Nanoscale Systems at Harvard University. Samples were fixed in 2.5% glutaraldehyde in 0.1 M sodium cacodylate and dehydrated through an ethanol gradient followed by hexamethyldisilazane drying. Specimens were captured on a porous membrane for visualization. Statistical analysis. All measured values are reported as an average of at least triplicate samples ± s.e.m., as indicated by error bars in all graphs. Significant differences in survival rates in the endotoxemia rat model were determined using the Mantel-Cox test and the Gehan-Breslow-Wilcoxon test (P < 0.05). Other significant differences were determined by the ANOVA t-test, as defined by P values <0.05. 37. Neth, O. et al. Mannose-binding lectin binds to a range of clinically relevant microorganisms and promotes complement deposition. Infect. Immun. 68, 688–693 (2000). nature medicine 38. Townsend, R., Read, R.C., Turner, M.W., Klein, N.J. & Jack, D.L. Differential recognition of obligate anaerobic bacteria by human mannose-binding lectin. Clin. Exp. Immunol. 124, 223–228 (2001). 39. Takahashi, K., Ip, W.E., Michelow, I.C. & Ezekowitz, R.A.B. The mannose-binding lectin: a prototypic pattern recognition molecule. Curr. Opin. Immunol. 18, 16–23 (2006). 40. Gilmore, J.M., Scheck, R.A., Esser-Kahn, A.P., Joshi, N.S. & Francis, M.B. N-terminal protein modification through a biomimetic transamination reaction. Angew. Chem. Int. Ed. Engl. 45, 5307–5311 (2006). 41. Scheck, R.A. & Francis, M.B. Regioselective labeling of antibodies through N-terminal transamination. ACS Chem. Biol. 2, 247–251 (2007). 42. Witus, L.S. et al. Identification of highly reactive sequences for PLP-mediated bioconjugation using a combinatorial peptide library. J. Am. Chem. Soc. 132, 16812–16817 (2010). 43. Shim, M., Shi Kam, N.W., Chen, R.J., Li, Y. & Dai, H. Functionalization of carbon nanotubes for biocompatibility and biomolecular recognition. Nano Lett. 2, 285–288 (2002). 44. Åkerman, M.E., Chan, W.C.W., Laakkonen, P., Bhatia, S.N. & Ruoslahti, E. Nanocrystal targeting in vivo. Proc. Natl. Acad. Sci. USA 99, 12617–12621 (2002). 45. Korin, N. et al. Shear-activated nanotherapeutics for drug targeting to obstructed blood vessels. Science 337, 738–742 (2012). 46. Takahashi, K. et al. Mannose-binding lectin and its associated proteases (MASPs) mediate coagulation and its deficiency is a risk factor in developing complications from infection, including disseminated intravascular coagulation. Immunobiology 216, 96–102 (2011). 47. Petersen, S.V., Thiel, S., Jensen, L., Steffensen, R. & Jensenius, J.C. An assay for the mannan-binding lectin pathway of complement activation. J. Immunol. Methods 257, 107–116 (2001). 48. Vorup-Jensen, T. et al. Recombinant expression of human mannan-binding lectin. Int. Immunopharmacol. 1, 677–687 (2001). 49. Michelow, I.C. et al. A novel l-ficolin/mannose-binding lectin chimeric molecule with enhanced activity against Ebola virus. J. Biol. Chem. 285, 24729–24739 (2010). 50. Palaniyar, N. et al. Nucleic acid is a novel ligand for innate, immune pattern recognition collectins surfactant proteins A and D and mannose-binding lectin. J. Biol. Chem. 279, 32728–32736 (2004). 51. Lakshmi, C., Hanshaw, R.G. & Smith, B.D. Fluorophore-linked zinc(ii)dipicolylamine coordination complexes as sensors for phosphatidylserine-containing membranes. Tetrahedron 60, 11307–11315 (2004). 52. Lee, J.-J. et al. Synthetic ligand-coated magnetic nanoparticles for microfluidic bacterial separation from blood. Nano Lett. 14, 1–5 (2014). 53. Ji, X., Gewurz, H. & Spear, G.T. Mannose binding lectin (MBL) and HIV. Mol. Immunol. 42, 145–152 (2005). 54. Beck, A. & Reichert, J.M. Therapeutic Fc-fusion proteins and peptides as successful alternatives to antibodies. MAbs 3, 415–416 (2011). 55. Xia, N. et al. Combined microfluidic-micromagnetic separation of living cells in continuous flow. Biomed. Microdevices 8, 299–308 (2006). 56. Yung, C.W., Fiering, J., Mueller, A.J. & Ingber, D.E. Micromagnetic–microfluidic blood cleansing device. Lab Chip 9, 1171 (2009). 57. Wiesel, P. et al. Endotoxin-induced mortality is related to increased oxidative stress and end-organ dysfunction, not refractory hypotension, in heme oxygenase-1– deficient mice. Circulation 102, 3015–3022 (2000). 58. Kang, J.H. et al. A combined micromagnetic-microfluidic device for rapid capture and culture of rare circulating tumor cells. Lab Chip 12, 2175 (2012). 59. Kang, J.H., Choi, S., Lee, W. & Park, J.-K. Isomagnetophoresis to discriminate subtle difference in magnetic susceptibility. J. Am. Chem. Soc. 130, 396–397 (2008). 60. Ling, F. & Zhang, X. A numerical study on mixing in the Kenics static mixer. Chem. Eng. Commun. 136, 119–141 (1995). 61. Jayakumar, J.S., Mahajani, S.M., Mandal, J.C., Iyer, K.N. & Vijayan, P.K. CFD analysis of single-phase flows inside helically coiled tubes. Comput. Chem. Eng. 34, 430–446 (2010). 62. Ofsthun, N.J., Jensen, J.C. & Kray, M. Effect of high hematocrit and high blood flow rates on transmembrane pressure and ultrafiltration rate in hemodialysis. Blood Purif. 9, 169–176 (1991). 63. Krisper, P. & Stauber, R.E. Technology insight: artificial extracorporeal liver support—how does Prometheus compare with MARS? Nat. Clin. Pract. Nephrol. 3, 267–276 (2007). 64. Huh, D. et al. Reconstituting organ-level lung functions on a chip. Science 328, 1662–1668 (2010). 65. Ordodi, V.L. et al. Artificial device for extracorporeal blood oxygenation in rats. Artif. Organs 32, 66–70 (2008). 66. Lee, H.B. & Blaufox, M.D. Blood volume in the rat. J. Nucl. Med. 26, 72–76 (1985). 67. Kang, J.H. & Park, J.-K. Magnetophoretic continuous purification of single-walled carbon nanotubes from catalytic impurities in a microfluidic device. Small 3, 1784–1791 (2007). 68. Cooper, R.M. A Generic Pathogen Capture Technology for Sepsis Diagnosis (MIT Press, 2013). 69. Onderdonk, A.B., Weinstein, W.M., Sullivan, N.M., Bartlett, J.G. & Gorbach, S.L. Experimental intra-abdominal abscesses in rats: quantitative bacteriology of infected animals. Infect. Immun. 10, 1256–1259 (1974). doi:10.1038/nm.3640 73. Remick, D.G., Newcomb, D.E., Bolgos, G.L. & Call, D.R. Comparison of the mortality and inflammatory response of two models of sepsis: lipopolysaccharide vs. cecal ligation and puncture. Shock 13, 110–116 (2000). 74. Mammoto, A. et al. Control of lung vascular permeability and endotoxin-induced pulmonary oedema by changes in extracellular matrix mechanics. Nat. Commun. 4, 1759 (2013). 75. Lindqvist, R. Estimation of Staphylococcus aureus growth parameters from turbidity data: characterization of strain variation and comparison of methods. Appl. Environ. Microbiol. 72, 4862–4870 (2006). npg © 2014 Nature America, Inc. All rights reserved. 70. Liang, J. et al. Enhanced clearance of a multiple antibiotic resistant Staphylococcus aureus in rats treated with PGG-glucan is associated with increased leukocyte counts and increased neutrophil oxidative burst activity. Int. J. Immunopharmacol. 20, 595–614 (1998). 71. Parker, S.J. & Watkins, P.E. Experimental models of Gram-negative sepsis. Br. J. Surg. 88, 22–30 (2001). 72. Pass, L.J., Schloerb, P.R., Pearce, F.J. & Drucker, W.R. Cardiopulmonary response of the rat to Gram-negative bacteremia. Am. J. Physiol. 246, H344–H350 (1984). doi:10.1038/nm.3640 nature medicine Copyright of Nature Medicine is the property of Nature Publishing Group and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.