Commercial albumin solution enhances endotoxin‐induced vasoplegia and inflammation

The Gram‐negative bacterium Escherichia coli, commonly involved in severe sepsis and septic shock, shed endotoxin that upon detection by the host triggers an inflammatory cascade. Efficiency of albumin solutions to restore hypovolemia during sepsis has been debated. To aid identification of subgroups of sepsis patients that may respond positively or negatively to treatment with albumin we investigated if preparations of albumin for medical use could affect endotoxin‐induced inflammatory response.

example, tumor necrosis factor-α (TNF-α) and interleukins (IL), can lead to endothelial dysfunction, characterized by increased capillary permeability and vasodilatation, 8 tissue hypoxia and organ failure. 9 Intravenous (IV) fluids and vasopressors are among the most important and frequently used evidence based interventions in sepsis.
Amounts, infusion rates, and types of fluids are still a topic for discussion. Most studies find no difference between albumin solutions and crystalloids concerning mortality, 10-12 among these a Cochrane review from 2011 summarizing 38 randomized controlled studies. 13 The Surviving Sepsis Campaign 2016 Guidelines, however, recommend albumin in addition to crystalloids, for patients who require very large amounts of crystalloids. 14,15 Positive effects of albumin solutions not only attributed to expansion of plasma volume have been proposed. There has been evidence that albumin can act as a scavenger of nitric oxide, thereby counteracting peripheral vasodilatation. 16 Albumin has also been suggested to have anti-oxidative as well as anti-inflammatory effects. 17 Adverse effects of albumin are less studied. Perdomo-Morales and colleagues demonstrated contamination of (1,3)-β-glucan derived from filters during the production process of human albumin solutions (HAS). 18,19 The (1,3)-β-glucan contributes to an increase in IL-6 release by monocytes, 18 and other studies have suggested a synergetic effect with endotoxins and (1,3)-β-glucan leading to more intense pyrogenic reactions in human models. 19,20 Plasma proteins can indeed affect host response to bacterial products. M1 protein, a PAMP released by Streptococcus pyogenes, increases IL-6 production and decreases the contractile response to noradrenaline in human blood vessels. We have previously demonstrated that the action of M1 protein is enhanced by fibrinogen. 10 Likewise, the plasma protein lipopolysaccharide-binding protein (LBP) enhances the inflammatory response to LPS by presenting it to CD14 and TLR4. 21 In order to enable identification of subgroups of sepsis patients that may respond positively or negatively to treatment with albumin solutions, in the present study we investigated if preparations of albumin for medical use could affect LPS-induced inflammatory response of the vascular wall and monocytes. Since immunoglobulin G and fibrinogen are main components of plasma and therefore potential contaminants in commercial albumin solutions, for comparison we also explored the effect of these on the response of human tissue to LPS.

| Human arteries
The project was approved by the local research ethics committee (LU 18-93 and 2015/801). After written informed consent, pieces of omentummajus were obtained from six patients undergoing abdominal surgery for pancreatic, ovarian, or uterine cancer. Omental arteries were dissected free from fat and connective tissues and cut into eight 2-4 mm long segments.

| Incubations of arteries
Arterial segments were incubated at 37°C in Dulbecco's modi-

| Measurement of smooth muscle contraction in human arteries
In the first set of experiments, the following combinations of substances were added to the incubation medium: (a) control without additives, (b) control with albumin (4 mg mL − 22 Potassium chloride (83 mmol L −1 ) was then added and a resulting smooth muscle contraction confirmed viability. After wash-out, noradrenaline (NA, 10 −10 -10 −4 mol L −1 ) was added cumulatively in 0.5 10 log units. The resulting contraction was registered and concentration-response curves were constructed.

Editorial Comment
The effect of albumin on inflammation continues to be a subject of debate. This experimental study adds to the discussion, demonstrating here that albumin in an in vitro setting enhances inflammation induced by bacterial endotoxins. Direct translation to the clinical situation from these findings is not possible and clinical study design to test this issue remains challenging. Still, these findings add support to the idea that albumin administration potentially could have some negative effects in sepsis patients.
After measurement each artery segment was weighed and contraction was expressed in mN mg −1 .

| Measurement of cytokine and nitrite/nitrate production by human arteries
In another set of experiments, similar incubations as above were used except for that LPS at 50 EU mL −1 was chosen since this gave a sufficient submaximal cytokine production in pilot experi-

| Isolation of monocytes
Venous blood from healthy volunteers was collected into heparin coated glass vials. Peripheral blood mononuclear cells (PBMC) were separated with density gradient centrifugation using Lymphoprep™ After 2 hours of incubation at 37°C in an atmosphere containing 5% CO 2 , the medium was changed to remove non-adherent cells. 24 Adherent cells consist mainly of mononcytes. 24 Cells were then equilibrated for another 20 hours.

| Incubations of monocytes
After equilibration, wells were rinsed twice using sterile PBS and

DY210-05) according to the manufacturer's instructions. Standard
curves were generated diluting TNF-α to known concentrations using the different media employed, that is, DMEM alone as well as DMEM containing 4 mg mL −1 albumin solution A1, A2, B, C, or D, respectively.
In a separate set of experiments, mononcytes were incubated with LPS as above with or without albumin solution A1 at 2, 4, or 8 mg mL −1 . Following incubation, TNF-α levels were measured in all incubations. In the incubations without or with albumin at 4 mg mL −1 , IL-1β, IL-6, IL-8, and IL-10 were measured with CBA as described.

| Endotoxin levels in albumin solutions
Albumin solutions were diluted × 5 in endotoxin-free PBS and analyzed using a high sensitivity assay based on Limulus amoebocyte lysate as previously described. 25 A standard curve was con-

| Measurement of (1-3)-β-glucan in albumin solutions
Levels of (1-3)-β-glucan in the five different albumin solutions were measured using enzyme-linked immunosorbent assay according to the manufacturer's instructions. The detection level was 0.8 pg mL −1 .
After washing, membranes were incubated with a secondary rabbit anti-goat IgG-HRP Conjugate antibody, diluted in the blocking solu-  analysis of variance on ranks followed post hoc testing against a control group using Dunnett's method and for multiple pairwise comparisons using Student-Newman-Keuls test. A value of P < .05 was considered to indicate statistical significance.

| Analysis of mass spectrometry data
The raw DDA data were analyzed with Proteome Discoverer™ 2. Master, Protein Unique Peptides is greater than or equal to 3. respectively, Figure 1).

| Smooth muscle force
Median AUC after incubation in the presence of LPS with albumin tended to be lower than after incubation with LPS alone but the difference did not reach statistical significance (P = .061). Incubation with albumin alone did not affect the concentration-response curve for noradrenaline (not shown).

| Cytokines and nitrite/nitrate released by blood vessels
After incubation with LPS alone there was a statistically non-significant tendency toward an increase in median release of IL-6 and IL-8 from human arteries compared to control (Figure 2). When LPS was combined with albumin, however, statistically significantly greater amounts of IL-6 and IL-8 were released compared to control as well as to with LPS alone (Figure 2). Release of IL-1β, IL-10, and TNF-α from the blood vessels was around three orders of magnitude lower than release of IL-6 and IL-8 ( Figure 3). Incubation with LPS alone did not affect cytokine release to any larger extent compared to control.
Adding albumin to the incubation increased median release of IL-1β, IL-10, and TNF-α compared to incubation with LPS alone (Figure 3).
Immunoglobulin G and fibrinogen did not affect LPS-induced cytokine release. Incubation with albumin, immunoglobulin G, or fibrinogen alone did not affect release of cytokines from the omental arteries (data not shown).
No detectable levels of nitrite/nitrate were observed in any of the incubations. Exposure to any of the albumin solutions alone did not affect TNF-α release compared to control (n = 8, not shown). When albumin solution A1, A2, B, or C was present during incubation with LPS, median TNF-α release was, however, statistically significantly greater than in the presence of LPS alone (P < .05, Figure 4). Albumin solution D did not affect median TNF-α release compared to incubation with LPS alone.

| Effect of different albumin solutions on cytokine release by monocytes
Median TNF-α release was increased by incubation with Albumin A1 at 2, 4, and 8 mg mL −1 and LPS compared to incubation with LPS alone (P < .05, not shown). No concentration-dependent effect of albumin was detected.
Release of IL-6 and IL-10 was significantly higher during incubation with albumin and LPS compared to LPS alone (P < .05, Figure 5).

Release of IL-1β and IL-8 during incubation with LPS differed from
control neither in the presence nor in the absence of albumin (data not shown).

| Contents of the Albumin solutions
Electrophoretic separation of all albumin solutions diluted 100 times showed a prominent band with a molecular weight consistent to albumin. In all solutions a fainter band could be seen with a molecular weight approximately 130 kDa ( Figure 6A). Diluting the albumin 1000 times narrowed the albumin band but no fainter bands were revealed next to it ( Figure 6B).

F I G U R E 1 Contraction induced by noradrenaline in human
omental arteries incubated in the presence or absence of lipopolysaccharide (LPS, 100 EU mL −1 ) with or without albumin (4 mg mL −1 ). The concentration-response curve to noradrenaline was shifted to the right after incubation with LPS (regardless of the presence of albumin) reflected by a decrease in median pEC 50 -value compared to control (P < .05). After incubation with LPS, median maximum response to noradrenaline was decreased compared to control (P < .05), an effect further enhanced by albumin (P < .05). In-gel digestion and mass spectrometry of the 130 kDa band showed that albumin was 100-fold more abundant than any other protein, which strongly suggests that this band consists of dimeric albumin. 26 The two albumin solutions from manufacturer A were found to contain (1-3)-β-glucan at 9 and 17 pg mL −1 , respectively, while levels in all the other solutions were below detection level.
The level of endoxins was below detection level (0.15 EU mL −1 ) in all albumins solutions except in solution A1 (close to detection level) and B (0.24 EU mL −1 ). originating from the donors, may contribute to priming the tissue to exaggerated sensitivity to LPS without having any detectable effect of its own.

| Western blot
One possible mechanism to the increased inflammatory effect observed in the present study, could be that albumin helps to create a more stable state of LPS, via binding of the lipid A portion, thereby to a greater extent extracting and transferring endotoxin to specific cellular targets such as TLR 4.Thisinteraction has previously been described, but demonstrated to include LBP and CD14. 34  We have previously demonstrated that M1 protein from S pyogenes incubated with human arteries in combination with fibrinogen increases IL-6 and IL-8 secretion and weakens smooth muscle contraction. 37 Albumin solution did not affect the response to M1 pro- The ultimate question is whether the increase in host immune response by albumin solutions indicated in the present study might worsen outcome for patients with Gram-negative sepsis. If so, it would be important to identify patients with endotoxaemia as a subgroup which might experience negative effects from treatment with albumin solutions. Further studies ought to have priority since sepsis patients are exposed to albumin treatment on a daily basis in ICUs worldwide.

| CON CLUS ION
We have shown that HAS triggers an aggravated LPS-induced proinflammatory cytokine activation in human omental arteries and monocytes as well as vasoplegia. A possible explanation could be that albumin forms complexes with LPS activating TLR 4 in an LBP independent manner.

ACK N OWLED G EM ENT
We thank the Local MS Support at Medical Faculty, Lund University for mass spectrometry analysis andGisela Hovold for providing excellent technical assistance with the CBA analysis. We also acknowledge Ravi Bhongir, research engineer at Infection Medicine, for brilliant work regarding additional incubations of monocytes and analysis of the results.

CO N FLI C T O F I NTE R E S T S
The authors declare that they have no competing interests.

AUTH O R S' CO NTR I B UTI O N
Viveka Björck was involved in laboratory work, statistical analysis, and drafting of manuscript. Mikael Bodelsson was involved in statistical analysis and drafting of manuscript. Linnea Andersson and Lisa I.
Påhlman were involved in laboratory work and scientific discussions.