Global Metabolic Phenotyping in an Experimental Laparotomy Model of Surgical Trauma James M. Kinross,† Nawar Alkhamesi,† Richard H. Barton,‡ David B. Silk,† Ivan K. S. Yap,‡ Ara W. Darzi,† Elaine Holmes,‡ and Jeremy K. Nicholson‡,* Section of Biosurgery and Surgical Technology, St. Mary’s Hospital, Praed Street, London W2 1NY, United Kingdom, and Department of Biomolecular Medicine, Department of Surgery and Cancer, Faculty of Medicine, Imperial College London, Alexander Fleming Building, South Kensington, London SW7 2AZ, United Kingdom Received April 9, 2010
Surgical trauma initiates a complex series of metabolic host responses designed to maintain homeostasis and ensure survival. 1H NMR spectroscopy was applied to intraoperative urine and plasma samples as part of a strategy to analyze the metabolic response of Wistar rats to a laparotomy model. Spectral data were analyzed by multivariate statistical analysis. Principal component analysis (PCA) confirmed that surgical injury is responsible for the majority of the metabolic variability demonstrated between animals (R2 Urine ) 81.2% R2 plasma ) 80%). Further statistical analysis by orthogonal projection to latent structure discriminant analysis (OPLS-DA) allowed the identification of novel urinary metabolic markers of surgical trauma. Urinary levels of taurine, glucose, urea, creatine, allantoin, and trimethylamine-N-oxide (TMAO) were significantly increased after surgery whereas citrate and 2-oxoglutarate (2-OG) negatively correlated with the intraoperative state as did plasma levels of betaine and tyrosine. Plasma levels of lipoproteins such as VLDL and LDL also rose with the duration of surgery. Moreover, the microbial cometabolites 3-hydroxyphenylpropionate, phenylacetylglycine, and hippurate correlated with the surgical insult, indicating that the gut microbiota are highly sensitive to the global homeostatic state of the host. Metabonomic profiling provides a global overview of surgical trauma that has the potential to provide novel biomarkers for personalized surgical optimization and outcome prediction. Keywords: surgery • systems metabolism • 1H NMR spectroscopy • systems integration • microbiome
Introduction Approximately 46 million surgical and interventional procedures are performed in the U.S.A. each year, but despite significant technological advances, complications from surgery still cause a significant economic and health burden.1 Nearly 30% of all severe sepsis patients admitted to critical care units have undergone a surgical procedure, with an estimated cost of $23 billion per annum.2,3 Reduction of this surgical morbidity is the goal of preoperative optimization strategies, which ensure an individual’s physiome and metabonome have the capacity to recover after surgically induced trauma, thereby reducing their risk of complications such as sepsis. Routine biochemical assays when used alone4,5 or as part of a clinical risk score6 provide a simplistic overview of an individual’s metabolic state, yet they are commonly used by clinicians for prognostication of surgical risk.7 More importantly, current therapeutic strategies used for minimizing the physiological insult of surgery or * To whom correspondence should be addressed. Professor Jeremy Nicholson, Head of Department, Surgery and Cancer, Section of Biomolecular Medicine, Imperial College London, The Sir Alexander Fleming Building, South Kensington, London SW7 2AZ, U.K. Tel.: +44 (0)20 7594 3195. Fax: +44 0207 595 3220. E-mail:
[email protected]. † Section of Biosurgery and Surgical Technology. ‡ Section of Biomolecular Medicine. 10.1021/pr1003278
2011 American Chemical Society
for the treatment of its complications are based on a linear mechanistic understanding of mammalian metabolism. Interindividual variation in the metabolic response to a surgical insult is known to be considerable, although it is generally accepted that patients progress through a metabolic “ebb and flow” response to injury.8,9 Mobilization of energy metabolites precedes the ebb phase, which refers to a state of physiological shock at the time of injury, typified by a reduced basal metabolic rate, hypothermia and a rise in endogenous stress hormone release. Flow refers to a hyperdynamic catabolic state that begins during recovery and is described by hyperthermia, insulin and cortisol release and raised plasma glucose, lactate, protein and free fatty acids (FFAs), developing what, is in effect, a metabolic state similar to type II diabetes.10 Recent developments in analytical biochemistry are challenging long held beliefs about mammalian metabolism, heralding a new era of personalized surgical healthcare. This impacts upon the development of pharmaceuticals and medical practices that are targeted to individuals based on their specific genetic code.11 If a personalized strategy for managing patients exposed to physiological insults such as surgery is ever to be realized, quantitative and qualitative analysis of environmental risk is essential, and its complex relationship with the human genome must be understood. Therefore, a new systems-level Journal of Proteome Research 2011, 10, 277–287 277 Published on Web 09/20/2010
research articles understanding of surgical trauma is necessary. Metabolic profiling strategies such as high throughput spectral analysis by nuclear magnetic resonance (NMR) spectroscopy or mass spectrometry (MS) permit a more global metabolic overview of complex organisms to be achieved, not only making the simultaneous interpretation of biological systems possible, but also the quantification of environmental influences on the host genome that impact surgical survival.12,13 Metabonomic analysis of biological samples provides a novel systems perspective from which surgical trauma may be studied14-16 and has previously been applied to characterize both acute and chronic stress,17 and to model sepsis18 and individual organ regeneration after surgically induced injury.19 This is because the simultaneous profiling of multiple metabolites provides a more sensitive and specific diagnosis than more conventional single clinical assays currently available. Most importantly, these techniques are able to indirectly measure complex transgenomic cometabolic interactions that are often modulated by surgical interventions20,21 and which are vital for human health. The notion of microbial-mammalian metabolic cooperation is defined through the concept of the human metabonome.22 It is not understood how intestinal microbiota modulation impacts on host health and the metabonome after surgical trauma or injury.23 This is because it is not currently possible to noninvasively measure the influence of the intestinal microbiota on postoperative outcome, as analysis is still predominantly based on traditional culture based analysis of stool samples. Here we have used a laparotomy rat model of surgical trauma to determine the feasibility of studying global changes in host metabolism after a generic but significant surgical insult and to determine if novel biomarkers can be elucidated to help define and predict surgical metabolic stress.
Methods Ethical approval for all animal work was attained in accordance with the Animals (Scientific Procedures) Act 1986. All surgical work was performed in the CBS facility at St. Mary’s Hospital, Imperial College London. Male Wistar rats (n ) 6) were purchased from Harlan aged 8-10 weeks old. The rats weighed between 200 and 250 g and underwent an acclimatization period of one week prior to surgery. The mean weight was 277 g (( 6.4 SEM). Animals were randomly housed in cages of four and two animals with distinct markings for animal identification. Animals were kept at a temperature of 22 °C with a relative humidity of 60-70% on a 12 h light/dark cycle. All animals had free access to standard chow and water preoperatively. Surgical Protocol. All surgery was performed during the morning to minimize diurnal variation between animals. Batch surgery of two animals was performed at a time using randomly chosen cages. All animals received a preoperative 5 mL subcutaneous bolus of normal saline and were anaesthetized using isoflurane. The core temperature was maintained between 35 and 37.5 °C in all animals and the animal’s pulse and blood oxygen saturations were monitored using a pulse oximeter. A midline laparotomy was performed and the intestines were delivered onto the abdominal wall and wrapped in a swab soaked in warmed normal saline and the Superior Mesenteric Artery (SMA) was identified to mimic a surgical exploration of the abdominal cavity. The intestines were then returned to the abdomen which was closed using interrupted 5.0 undyed Vicryl 278
Journal of Proteome Research • Vol. 10, No. 1, 2011
Kinross et al. sutures and then covered with a polyurethane dressing to prevent further fluid loss. Sampling Protocol. Femoral cannulation was performed after induction of anesthesia and 0.4 mL of blood, from which plasma was obtained, was collected preoperatively and at 60 min of surgery time. At 150 min after the induction of anesthesia a cardiac puncture was performed and the animal was exsanguinated. Blood samples were collected in lithium heparin tubes and then stored on ice. After 2 h, they were centrifuged at 1200 g for 15 min after which 200 µL of plasma samples were aliquoted into 1.5 mL eppendorf tubes and stored at -80 °C. Preoperative urine was collected at 08:00 h each day before surgery by placing the animal in a metabolism cage for two hours. Intra-operative urine samples were taken at the same time points as plasma collection by direct bladder massage. After surgery a necroscopy was performed to ensure that no physical damage to the intestine had been caused. Urine Sample Preparation. Urine samples were prepared using 100 mL of urine to which was added 400 mL of 200 mM Phosphate buffer at pH 7.4. The buffer was prepared into a 80:20 H2O:D2O mixture containing 1 mM TSP (trimethylsilylpropionic acid) for the chemical shift reference and 3 mM of sodium azide as a bacteriostatic agent. This was transferred into a 5 mm outer diameter NMR tube. All 1H NMR spectra were acquired using a Bruker DRX600 spectrometer (Rheinstetten, Germany) with a 5 mm TXI probe operating at 600.13 MHz. The field frequency was locked on D2O solvent. Primary acquisitions were made using a standard 1-D pulse program [Recycle delay (RD)-90°-t1-90°-tm-90°-acquire Free Induction decay (FID)]. The water peak was suppressed by irradiation during RD of 2 s and mixing time (tm) of 100 ms. t1 was fixed to 3 µs. The 90° pulse length was adjusted to approximately 10 µs. 128 scans were recorded into 32 k data points with a spectral width of 20 ppm. An exponential function was applied to the FID prior to the Fourier transformation, which resulted in a line broadening of 0.3 Hz. Plasma Preparation. Plasma samples were centrifuged at 10 000 g for 5 min to remove any solid matter. Sodium chloride solution (400 µL) (0.9% w/v, made up in 80% H2O/20% D2O) was added to 100 µL of the plasma sample in 3 mm Wilmad NMR tubes. D2O (deuterium oxide) was added as a fieldfrequency lock for the NMR spectrometer. 1H NMR spectra of the plasma samples were acquired on a Bruker DRX 600 MHz spectrometer (600.13 MHz 1H-observation frequency) at a probe temperature of 300 K employing two 1D NMR experiments. These were a standard one-dimensional pulse sequence (as described for the urine) and a Carr-Purcell-Meiboom-Gill (CPMG) [RD-90°-(τ-180°-τ)n -acquire FID] (here, τ ) 2nt, where n ) the number of spin echoes and t ) CPMG delay time). A spin-spin relaxation delay of 64 ms was used for all samples and water suppression irradiation was applied during the relaxation delay (2 s). Typically in the standard one-dimensional and CPMG experiments, the spectral width was 20 ppm and 256 transients were collected into 32 k data points. CPMG experiments filter broad resonances from proteins and lipids, permitting latent biomarkers of smaller molecular weight to be visualized. This is particularly useful in a surgical model where tissue trauma and autolysis may lead to bias created by lipid or protein resonances. Data Processing and Multivariate Analysis. All urine and plasma 1H NMR spectra were phased and baseline corrected within XWINNMR (version 3.1, Bruker Biospin, Rheinstetten, Germany), and the chemical shifts were referenced either to
research articles
Metabolic Phenotyping of Surgical Trauma the TSP peak at δ0.0 (urine) or to the R-glucose anomeric resonance at δ5.23 (plasma). These data were collected into Excel (Microsoft, Excel 97, SR-2) data tables, where each row comprised the integral descriptors for an NMR spectrum. Plasma spectra were analyzed using non-normalized data, whereas urine samples were normalized to unit area after removal of the spectral region containing the suppressed water resonance so as to reduce the effects of concentration differences. The normalization procedure is a mathematical correction applied to experimental data in an attempt to adjust for uncontrollable systemic technical variation (bias). This is necessary in urine during surgery because it takes into account the variations in metabolite excretion caused by altered renal function. Scaling refers to the standardization applied to spectra before chemometric analysis may take place. This is necessary in NMR spectroscopy because the signal intensity varies across spectra as not all of the metabolites in a biofluid have the same signal intensities. Therefore scaling is applied to minimize the effect of the variation in intensity and to ensure that larger peaks do not overshadow the contribution of smaller peaks, and bias model interpretation. Urine data were mean-centered by subtraction of the average value of each variable from the data and analyzed without further scaling. Unit-variance scaling (UV) utilizes a formula whereby each variable is corrected by dividing it by the calculated standard deviation, hence because animals that suffer extremes of physiology are prone to excretion of metabolites of a larger molecular weight (such as lipoproteins). UV scaling was applied to the Plasma data sets. Recursive segment-wise peak alignment (RSPA) was also used to account for variations in peak position across samples in the spectra for both urine and plasma.24 This method refines the segmentation of reference and test spectra in a top-down fashion, subdividing the initial larger segments into smaller ones, as required, to improve the local peak alignment. The spectra were imported using a MATLAB (version 7, The Mathworks, Inc.; Natwick, MA) script developed in-house. The region containing the water resonance was removed from each spectrum to eliminate baseline effects of imperfect water saturation. The resulting data matrices were analyzed using SIMCA-P (version 11.5, UMETRICS AB, Box 7960, SE90719 Umeå, Sweden). Principal Component Analysis (PCA) scores plot were constructed for both plasma and urine data sets for samples taken at all time points to visualize any inherent clustering of the samples based on surgical injury. Biomarkers of surgical injury were subsequently identified from the values of the Principal Component loadings (which indicated the importance of each variable to the distribution of the samples in the scores plots) and the corresponding NMR spectral regions. The R2 value provides a quantitive measure of the goodness of fit, indicating the degree of variation within the data set which can be explained by the model. The Q2 value represents the percentage of the variation in the data set (X with PCA and Y with the PLS-DA) predicted by the model according to cross validation, that is, the Q2 value indicates how well the model predicts new data. A large Q2 (>0.5) indicates good predictivity. Despite recent data suggesting that orthogonal projection to latent structure discriminant analysis (OPLS-DA) on the spectra does not remove the interindividual variation25 it minimizes the degree of extreme variation found during periods of extreme physiological extremis caused by surgery. Therefore, Orthogonal Projection to Latent Structure Discriminant Analysis
26
OPLS-DA was performed in a MATLAB environment using in-house protocols to optimally model class differences and to systematically identify metabolites contributing to the differences between pre and post operative states. The OPLS-DA models were constructed using NMR spectral data as the X-matrix and class information as the Y-matrix. One orthogonal component was used to remove variation unrelated to class. A post processing step was used with transformation of the data and the reconstruction of a spectral representation of loadings plot. This was constructed with a weighted scale based on the data scaled to unit variance and it is highlighted by a color code projected onto the spectrum to indicate the correlation of the metabolites discriminating between the preoperative and post operative state. Red indicates a high correlation and dark blue denotes no correlation with sample class. The direction and magnitude of the signals relate to the covariation of the metabolites with the classes in the model. A coefficient of 0.71, corresponding to 5% significance level (i.e., the significance of the Pearson product-moment correlation coefficient for each metabolite to the description of the intra operative state (r) )
