Capillary leak accompanying systemic inflammatory response conditions is a significant clinical problem. In the present study, we describe and verify a method for studying capillary leak that is based on the injection of proteins that differ significantly in size and have spectrally distinguishable fluorophores. Control (n=11) and post-CLP (caecal ligation and puncture; n=14) Sprague–Dawley rats were injected with tracer amounts of albumin and PEG–Alb [albumin covalently linked to methoxy-poly(ethylene glycol)] labelled with fluorescein and Texas Red. Blood samples were withdrawn between 5 min and 144 h, and the fluorescence of the labelled proteins was determined. The relative retention of the PEG–Alb and albumin was assessed via measurement of the TER (transcapillary escape rate; in %/h) and the t50% estimate, defined as the time when the actual concentration reached 50% of its baseline. The concentration–time trends for both albumin and PEG–Alb tracers exhibited two-compartmental behaviour and were analysed using bi-exponential modelling. Retention times were significantly greater for PEG–Alb in both control and CLP rats. TERPEG-Alb was significantly lower than TERalbumin for both control (8.1±5.6 compared with 14.8±7.1 %/h respectively; P<0.01) and CLP (14.8±6.6 compared with 22.5±7.3 %/h respectively; P<0.001) rats. The t50%[PEG–Alb] was substantially greater than the corresponding t50%[albumin] for both control (29.8±9.8 compared with 7.2±2.0 h respectively; P<0.001) and CLP (12.9±5.6 compared with 5.1±1.6 h respectively; P<0.001) rats. The result was similar irrespective of the fluorophore–protein combination, validating the multifluorophore technique. In conclusion, the double-fluorophore approach described in the present study may provide the future basis for a method to quantify capillary leak in disease.
Attaching PEG [poly(ethylene glycol)] to proteins (PEGylation) increases their intravascular retention time [t½ (half-life)] by decreasing physiological turnover (e.g. protecting against proteolysis) and their antigenicity . For example, PEGylation increases the t½ of glucagon 16 times [2,3] and the t½ of α-interferon more than 12 times. The consequences of PEGylation can be as significant as observed with α-interferon, where PEGylation dramatically increased its therapeutic effectiveness in treating hepatitis C .
We have reported previously  that PEGylated albumin [PEG–Alb, i.e. albumin covalently linked to poly(ethylene glycol)], which is 16 times larger than albumin, extravasates less in capillary leak conditions associated with CLP (caecal ligation and puncture) and lipopolysaccharide models of severe sepsis . Indirect physiological evidence (haemodynamics and colloid osmotic pressure), fluorescent lung section imaging (qualitative) and in vitro biophysical data (size-exclusion chromatography and surface-enhanced laser-desorption ionization–time of flight MS) are consistent with increased vascular retention of PEG–Alb . However, to date, the in vivo retention and clearance of PEG–Alb compared with albumin has not been reported.
Capillary leak varies directly with the plasma volume and the total intravascular albumin . To eliminate some of the uncertainties related to changes in plasma volume, two proteins with different molecular masses have been used in the present study. We employed this double-fluorophore technique to compare retention of albumin and PEG–Alb under control and capillary leak conditions.
In principle, preferential vascular retention could be demonstrated using radiolabelled albumin and PEG–Alb (e.g. 125I and 131I) . This method, however, exposes subjects to radioisotopes and the associated contamination hazards, employs isotopes with relatively short t½ and has technical problems in measuring the two isotopes in the same sample. The present study describes and validates an experimental method that allows (i) simultaneous assessment of intravascular retention of albumin and PEG–Alb using two spectrally distinct fluorophores to label albumin and PEG–Alb, and (ii) simultaneous determination of their concentrations over time.
MATERIALS AND METHODS
Preparation of dye-conjugated albumin and PEG–Alb
Preparation of PEG–Alb and their dye conjugates of albumin and PEG–Alb has been described previously . To compare the degree to which PEG–Alb was retained compared with albumin, we employed two distinct fluorophores so that the clearance of the proteins could be evaluated in the same animal, minimizing the confounding effects of animal-to-animal variation. Albumin and PEG–Alb were labelled with maleimide derivatives of FL (fluorescein) or TR (Texas Red) (Molecular Probes) via Cys34 of albumin. The excitation and emission spectra of TR (excitation, 590 nm; emission, 617 nm) and FL (excitation, 492 nm; emission, 515 nm) are sufficiently separated that mixtures of the two dyes can be examined quantitatively in the same sample. As necessary, measurements of intensity were corrected for the inner-filter effect [8,9]. Interference from other fluorescent material in serum was negligible at the dilutions (1:100–1:200) employed.
Animals and the experimental protocol
Measurements were conducted in 14 Sprague–Dawley rats with sepsis and 11 control Sprague–Dawley rats (body weight, 300–400 g; Charles River Laboratories). Sepsis was induced via CLP surgery and was compared with no surgery in control rats. Animals were housed in the University of Toledo College of Medicine animal laboratory. Food and water were allowed ad libitum. The Institutional Animal Care Committee approved the experimental protocol.
CLP sepsis model
Rats were fasted 10–12 h prior to pre-experiment surgery with water available ad libitum. Sepsis was induced using a CLP model [10,11]. Briefly, rats were anaesthetized with sodium pentobarbital (50 mg/kg of body weight, intraperitoneal injection), followed by pentobarbital (12.5 mg/kg of body weight, intraperitoneal injection) as required. A laparotomy was performed through a midline abdominal incision. The caecum was ligated just below the ileocaecal valve with 3-0 silk ligatures such that intestinal continuity was maintained. The caecum was perforated with a 16-gauge needle in two locations and gently compressed to extrude faeces. The bowel was returned to the abdomen, and the incision was closed with proline sutures for the muscle and 3-0 silk sutures for skin. A total of 3 ml of sterile 0.9% saline/100 g of body weight was administered subcutaneously for resuscitation. Following surgery, rats were provided with moistened chow and had free access to water. One rat died following CLP surgery and before data collection was started.
Blood sampling procedures in control and CLP rats were identical. Rats were anaesthetized 18 h following CLP and were injected with tracer amounts of a mixture of PEG–Alb–FL+albumin–TR or albumin–FL+PEG–Alb–TR (100 μl of the mixture/100 g of body weight) through an internal jugular vein canula.
The initial blood sample (100–150 μl) was taken 5 min after injection to allow for sufficient mixing, and the sample taken at this time was determined to be t=0. Thereafter samples were taken at 15 min, and 0.5, 1, 3, 5 and 8 h through a jugular line. The jugular line was removed and the rats were returned to their cages. Longer-term blood samples were taken through a tail vein over 5–6 days. Serum samples for fluorescence were diluted 100–200-fold in 10 mmol/l potassium phosphate buffer (pH 7.5) and 150 mmol/l NaCl. Blank spectra on buffer and on diluted serum from animals that did not receive the labelled albumins were used to correct for endogenous fluorescence.
Bi-exponential modelling of fluorophore concentration–time data
Concentration–time data were determined from blood serum fluorescence collected at known time intervals after bolus injection of the fluorophores labelled with PEG–Alb and albumin. The respective concentration–time [C(t)] data for each chromophore were then analysed separately using a two-compartment or bi-exponential model Cm(t) as follows:
where A and B are the respective amplitudes of the fluorescence attributable to (i) the rapid redistribution (capillary leak) phase defined by parameters A and α; and (ii) the slower elimination (degradation/catabolism) compartment defined by parameters B and β.
The measured concentration–time data were normalized to the total fluorescence measured in the first serum sample (assigned in the model as t=0); the amplitude A and B in the model were constrained during fitting of the individual rat data such that A+B=1. The rate constants α and β were not constrained, and three model parameters (B, α and β) that best fitted the concentration–time data were estimated separately for the labelled albumin and PEG–Alb in individual rats (non-linear regression tool box; SigmaPlot 2002).
Using the parameter estimates α and β for each chromophore, the corresponding t½ was estimated for the two compartments as t½,α=0.693/α and t½,β=0.693/β respectively. TER (transcapillary escape rate; in %/h) was calculated based on the rapid redistribution/capillary leak phase kinetics as follows:
Albumin and PEG–Alb volumes of distribution (Vd) were estimated using the relationship :
where the tracer concentration, as determined by fluorescence, is Ci and the injected volume (or dose) is Vi. Cβ(0) was determined as the elimination phase normalized amplitude B multiplied by the initial injected chromophore concentration [Cβ(0)=B×Ci].
Vascular retention of the chromophores was quantified using the corresponding t50% estimate defined as the time when the concentration reaches 50% of its baseline (t=0) value. This estimate is not phase-specific, and was derived via a computer algorithm applied to the two-compartment model fit. All modelling parameters and derived estimates were compared using a paired Student's t test (for albumin compared with PEG–Alb values) or unpaired Student's t test (for CLP compared with control). Statistical significance is reported at the P<0.05 level.
Multifluorophore results and method validation
Fluorescence measurements were performed in 25 rats (11 controls and 14 CLP rats). Four of the 14 CLP rats with sepsis died early at 54, 67, 67 and 68 h after injection of the tracers. The time-dependent decrease in the albumin and PEG–Alb tracer concentrations closely followed a bi-exponential behaviour, as shown by the model fitted to the pooled control and CLP results in Figure 1. Figure 2 shows the albumin and PEG–Alb concentration–time decay curves for representative individual control (n=6; upper panels) and CLP (n=6; lower panels) rats. These results showed that (i) the decay curves in semi-logarithmic plots were non-linear for albumin and PEG–Alb attached, indicating multicompartmental behaviour of their intravascular elimination, and (ii) in control and CLP rats, albumin decay was faster than the PEG–Alb decay.
Fluorescence of albumin and PEG–Alb, indexed to the concentration at injection time (
t=0), as a function of time in control (upper panel) and CLP (lower panel) rats
Representative albumin and PEG–Alb concentrations over time in six control (upper panels) and six CLP (lower panels) rats
To validate the double-fluorophore approach, the fluorophores were interchanged in two subcohorts of the 14 CLP rats: (i) nine rats were injected with PEG–Alb–FL and albumin–TR (Figure 3, upper panel), and (ii) five rats were injected with PEG–Alb–TR and albumin–FL (Figure 3, lower panel). The relative serum concentrations of the labelled albumin and PEG–Alb decreased with time after tracer injection in a similar fashion, irrespective of the fluorophore–protein combination. These results confirm that the differences observed in clearance reflect the properties of PEG–Alb and albumin and not the properties of the fluorophore employed.
Comparison of albumin and PEG–Alb concentrations over time in 14 CLP rats with reversal of the protein–fluorophore combinations
Comparison of TERalbumin and TERPEG-Alb in control and CLP rats
The concentration–time decay curves were analysed using the bi-exponential (two-compartment) model shown in eqn (1). Example of the model fitted to the simultaneously collected albumin and PEG–Alb decay results in six control rats and six CLP rats are shown in Figure 2. The model closely described the time-dependent decrease in tracer concentrations in all rats. The derived model parameters [A, B, τ1 (1/α) and τ2 (1/β)] for albumin and PEG–Alb in control and CLP rats are summarized in the insets in Figure 5. TER data derived on the basis of the fast-compartment kinetics (A and τ1) were consistent with the 1 h determination, averaging 8.2 compared with 17.3%/h (PEG–Alb compared with albumin) in control rats and 12.3 compared with 21.5%/h (PEG–Alb compared with albumin) in CLP rats.
Comparison of the bi-exponential-model-derived parameters for PEG–Alb and albumin in control and CLP rats
Individual t½,α and t½,β for the rats were calculated based on the model-derived τ1 (1/α) and τ2 (1/β). These, along with the parameter B (the slow-compartment amplitude), were compared for the control and CLP rats and are shown in Figure 5. Note, that in ten out of 11 control and 11 out of 14 CLP rats, the PEG–Alb B estimate was greater than the corresponding albumin B estimate. In the remaining four rats, the B coefficient values were essentially identical. This result is consistent with a larger volume of distribution of albumin compared with PEG–Alb [B is inversely proportional to Vd (see the Materials and methods section)].
The two-compartment model estimation was also used to quantify the intravascular retention of the two tracers in terms of the phase-independent t50%. The t50% [PEG–Alb] was consistently and substantially greater than the corresponding t50%[albumin] in the controls (Figure 5, lower-right-hand panel).
Developing a reliable method to quantify capillary leak in the setting of SIRS (systemic inflammatory response syndrome) conditions is a desirable yet, thus far, elusive goal. If available, such a measure might provide a clinically useful quantification of SIRS severity, which is an important precursor of MODS (multiorgan dysfunction syndrome) [13,14]. Much of the efforts towards achieving this goal have focused on analyses of serial measurements of intravascular protein concentrations. However, the multiple factors contributing to protein concentrations, including varying forms of protein transcapillary transport, lymphatic return, protein synthesis, protein catabolism/degradation and variations in circulating plasma volumes, limit the ability to do so.
In the present study, we approached this question via serial simultaneous measurements of unmodified albumin and a larger albumin, with an approx. 16 times larger hydrodynamic radius, modified via the attachment of multiple PEG groups (PEG–Alb). The two injected tracer proteins were tagged with fluorescently distinct fluorophores and their respective serial concentration–decay curves were constructed and analysed in both healthy animals (control) and animals with SIRS (CLP sepsis). Our main findings are as follows: (i) PEG–Alb concentrations were greater than the corresponding albumin concentrations measured in both CLP and control rats; and (ii) for either protein, the decay in tracer concentrations as a function of time were greater in CLP compared with control conditions.
The initial phase of the tracer disappearance from the circulation represents redistribution of the protein, which includes both the physiological and pathological capillary leak processes. This phase is primarily dependent on diffusion, since there is initially a large concentration gradient . Our present results show that the capillary leak of albumin (defined as the fraction of albumin that leaks to the extravascular space per unit time) was increased approx. 2-fold compared with PEG–Alb in both control and CLP rats. These capillary leak rates of albumin compare well with other sepsis studies in rats .
Although proteolysis and immune recognition are important routes for protein degradation or elimination , under normal physiological conditions a fixed fraction of albumin is catabolized daily, and this form of catabolism is known to be concentration-dependent . However, the modelling results suggest less degradation of PEG–Alb compared with albumin, despite the fact that PEG–Alb concentrations were higher at all times. This is probably a consequence of PEGylation by (i) protein shielding , (ii) suppressing protein immunogenecity and antigenecity , and (iii) reducing protein phagocytosis and preventing opsonization by reticular activator system and liver cells .
Using one labelled protein to measure capillary leak severity is plagued with difficulty, especially by the ongoing change in vascular volume. Using two or more labelled proteins of different sizes allows a more accurate quantification and assessment of capillary leak during the progression of SIRS conditions. Walker et al.  used two iodine-labelled proteins (131I-albumin and 125I-globulin) injected simultaneously into humans. However, quantifying the radioactivity was difficult and clinically impractical, because detecting the radioactivity of 125I required sufficient time (weeks) for the 131I to decay. Present techniques used for measuring capillary leak consist of using radioactive and colorimetric tracers . These techniques require a correction with radiolabelled red cells as an intravascular marker, since a substantial fraction of tissue activity is intravascular, in addition to the inconvenience of housing radioactively contaminated animals. Other researchers have applied a mass balance technique to calculate plasma volume and capillary leak based on repeated blood sampling of haemoglobin and plasma proteins [22,23].
Importantly, conjugation of TR or Evans Blue to albumin was shown to provide comparable concentration–decay results with radiolabelling of albumin with 131I . Advantages of our present technique over existing techniques are as follows: (i) it does not require isotopes; (ii) the chromophore is covalently linked to the protein, so the leak of the dye from the vascular space is not possible; (iii) availability of non-toxic chromophores, such as FL (excitation, 492 nm) or Indocyanine Green (excitation, 748 nm), allow injection of higher tracer concentrations, which will improve the accuracy of measurements by increasing the signal to background ratio ; (iv) it does not alter the protein physical or osmotic properties; and (v) with this approach, each animal acts as its own control, avoiding the limitations of a single-tagged protein.
The full experimental concentration–decay data set (including both phases) showed a consistently better retention of PEG–Alb compared with albumin; however, the timing and frequency of the early samples in some animals might not have been sufficient in the early phase to allow an accurate characterization of the fast or redistribution compartment. This concern was, however, mitigated by the similar results of the TER calculated in all instances (albumin and PEG–Alb in control and CLP rats) based on blood sample measurements between 0 and 60 min, as well as using the slope of the redistribution/capillary leak phase. Another limitation of our present study was that separate model analyses applied to the albumin and PEG–Alb tracer concentration–decay data implicitly assume that the intravascular plasma volume was unchanged over the entire time period. Usually, these variations can be corrected using concomitant haematocrit determinations, but this is not possible in a small animal model. We suggest further that the ratio of the simultaneously measured PEG–Alb (larger protein) and albumin (smaller protein) fluorophore concentrations as introduced in the present study will be independent of plasma volume variations. Therefore modelling of this ratio may provide a basis for quantifying capillary leak severity in a clinically relevant setting (such as varying vascular volume due to fluid administration, diuresis and/or transcapillary leak). Finally, we do not have results on renal excretion of albumin and PEG–Alb, such an excretion will increase the elimination (or concentration decay) and may confound the interpretation of the model parameters.
In summary, we have described a novel double-fluorophore tracer method based on two fluorescent proteins of distinct molecular masses (albumin and PEG–Alb) that are simultaneously injected in tracer amounts, and are subsequently tracked over time. In the present study, both proteins are assessed in the same subject and, hence, are studied under the same pathophysiological conditions (including intravascular volume). This technique may be applied to other proteins combinations (two or more), and developing it may allow the early detection and quantification of capillary leak, which precedes the development of overt organ injury by several days. In the present study, we provide evidence supporting the conclusion that the improved plasma volume expansion achieved with PEG–Alb compared with albumin in sepsis and endotoxaemia (i.e. capillary leak conditions) is a consequence of the increased intravascular retention of PEG–Alb .
The study was sponsored, in part, by a start-up fund from the Department of Medicine, University of Toledo College of Medicine. R. A. A., J. I. S. and J. D. D., together with the University of Toledo College of Medicine, have filed a patent for PEG–Alb (US patent number 7,037,895B2).