Identifying the rules of engagement enabling in silicoleukocyte rolling activation and adhesionJonathan Tang C Anthony HuntUniversity of California San FranciscoThe UCSFUCB Joint Graduate Group in Bioengineering

Objectives Leukocyte rolling activation and adhesion onendothelial cell surfaces are key events during leukocyterecruitment to tissues during the inflammatory process Theinteractions between leukocytes and endothelial cell surfacesduring rolling activation and adhesion are complex and involveligand-binding events between several combinations of receptor-ligand pairs Spatially restricted signaling events betweenchemokine receptors and local integrins as well as lateral diffusionand clustering of integrins are thought to play important roles inmediating the transition from rolling to adhesion The objective ofthis project was to develop a novel computational model thataimed at gaining a fundamental understanding of the spatiotem-poral events and key rules of engagement that determine when andhow leukocytes are able to adhere to endothelial cell surfacesduring inflammatory conditionsMethods We used the synthetic modeling approach to construct amultilevel multiagent in silico model of leukocyte rollingactivation and adhesion Object-oriented software componentswere designed verified plugged together and then operated inways that represent the mechanisms and processes believedresponsible for leukocyte rolling and adhesion The in silico systemconstructed was a discrete event discrete space and discrete timeanalogue of the entire parallel plate flow chamber systemcommonly used to study leukocyte rolling and adhesion in vitroThis modeling approach allowed us to represent the spatial anddiscrete event phenomena that are thought to occur in vitro andin vivoResults We previously reported details of the model and initialresults of in silico rolling and adhesion on P-selectin and VCAM-1substrate in the presence of GRO-alpha chemokine (PMID17408504) in silico leukocyte movement is a function of (1)number and type of bonds and (2) location of the bond-containingpatches at the interface Our current effort has extended the modeland increased resolution to enable exploration of the role of lateraldiffusion and clustering of LFA-1 integrin in mediating rolling andadhesion to ICAM-1 substrateConclusions These new methods represent important progress ingaining a fundamental understanding of the spatiotemporalevents and key rules of engagement that determine when andhow leukocytes adhere to endothelial cell surfaces duringinflammation The current model and results represent animportant step in developing scientifically useful validatedsimulations of leukocyte recruitment under physiologic andpathophysiologic conditions

doi101016jjcrc200710017

Impact of analysis interval on heart rate and blood pressurevolatility in a murine model of sepsisAnupam Gupta Bryan Foley Josh WeinstockKarine Hageboutros Jad Skaf Joseph E ParrilloSergio Zanotti Steven M HollenbergDepartment of Cardiology and Critical CareCooper University Hospital Camden NJ USA

Objectives Nonlinear analysis of beta-to-beat heart rate (HR) andblood pressure (BP) variability may provide insights into diseasepathophysiology not available from standard linear measures Wehave used HR and BP volatility (SD variability) as a means ofassessing variability that may minimize artifact-induced error andhave used this method to demonstrate decreased HR and BP volatilityin 5-minute intervals in a murine model of septic shock Shorteranalysis intervals may allow for more detailed time course analysesbut aremore time consuming In this study we examined the effects ofthe time segment of analysis on HR and BP volatility measurementsMethods Radiotelemeters for noninvasive measurement of bloodpressure were implanted into the ascending aorta of C57Bl6 mice(8-12 weeks n = 23) After 5 to 7 days of recovery postimplantationbaseline data were collected for 24 hours The mice were then madeseptic by cecal ligation and puncture (CLP) and resuscitated withfluids and antibiotics every 6 hours controls underwent shamligation HR and systolic BP were calculated from BP waveformsand HR and BP SDswere calculated in on each 1- 5- and 10-minutetime segments for the 72-hour course of the experiment For eachanimal the SD cutoff of both HR and BP that represented the lowest5 was determined and the of low SDs (low volatility) in theentire experimental period was defined by this cutoffResults Threshold values of the 5 SD cutoff did not differ for1- 5- and 10-minute analysis intervals for either HR (206 plusmn 125340 plusmn 184 382 plusmn 160 respectively P = NS) or BP (19 plusmn 0430 plusmn 06 39 plusmn 09 respectively P = NS) HR volatility differedsignificantly between control and septic groups for all 3 analysisintervals (1 minute 44 plusmn 43 vs 432 plusmn 270 P = 01 5 minutes47 plusmn 38 vs 455 plusmn 326 P = 02 10 minutes 48 plusmn 72 vs 471 plusmn337 P = 02) BP volatility differed significantly between controland septic groups for the 5- and 10-minute intervals (5 minutes117 plusmn 46 vs 392 plusmn 196 P = 01 10 minutes 139 plusmn 58 vs453 plusmn 235 P = 02) Low BP volatility intervals were morecommon for septic than control animals for the 1-minutemeasurement interval as well (380 plusmn 241 vs 146 plusmn 50) butdid not reach statistical significance (P = 07) Examination of BPwaveforms revealed that increased artifact influenced BP volatilityin the 1-minute analysis intervalConclusions less demanding and more intuitive as a means ofmeasuring variability than spectral analysis and artifact increasesSD and so biases the data to the null hypothesis In our study thisdecreased the sensitivity of the 1-minute analysis interval to detectchanges between experimental and control subjects The use oflonger-time segments or better filtering techniques may benecessary to minimize the influence of measurement artifacts inBP volatility analysis

doi101016jjcrc200710018

Integer heart rate multiscale entropy and variability earlyindependent predictors of mortality in 3145 traumaICU patientsPatrick Norrisab Judith Jenkinsa John MorrisacaDivision of Trauma Burn and Surgical Critical CareVanderbilt UniversitybDepartment of Biomedical Engineering Vanderbilt UniversitycDepartment of Biomedical Informatics Vanderbilt University

Objectives We have shown that integer heart rate multiscaleentropy (MSE) and variability (HRV) separately predict trauma

341Abstracts

patient death We hypothesized that early MSE and HRVcontribute independent information when combined in a mortalitymodel and predictions are valid even for long lengths ofstay (LOS)

Variable OR plusmn SE LOS AUC plusmn SE

MSE (sum) 096 plusmn 01 1 day 081 plusmn 04HRV short () 103 plusmn 01 2-3 days 063 plusmn 05HRV long (bpm) 113 plusmn 02 4-9 days 069 plusmn 02

ge10 days 069 plusmn 05

P b 001

Methods Three thousand one hundred forty-five patients had3 hours of suitable HR data within the first day of ICU stay Missingdata were interpolated and a publicly available MSE algorithm wasapplied (wwwphysionetorg m = 2 r = 015) Short-term HRV wascomputed as in past studies the percentage of 5-minute intervalswith HR SD from 03 to 06 Long-term HRV was computed as theHR SD over 3 hours Multivariate logistic regression models fordeath were built using half of the cases (randomly selected) andevaluated on the other half using area of the receiver operator curve(AUC) over subgroups stratified by LOSResults MSE (sum of sample entropies) and HRV measuredindependently predict death (overall AUC = 069 plusmn 002) Asexpected AUC improves with shorter LOS patients who aredischarged or die earlier seem easier to distinguish Even so themodel remains predictive past day 10 Not shown Includingindependent covariates (age severity) increases AUCs by sim08Conclusions (1) Within hours of ICU arrival MSE and HRVpredict trauma patient outcome (2) Predictive ability isimproved in patients who rapidly die or are discharged butremains significant in those surviving past 10 days (3) MSEspredictive ability is independent of variability suggesting thatinteger HR correlation structure is an important determinant ofpatient risk

doi101016jjcrc200710019

Mathematical modeling of innate immunity From endotoxintolerance to sepsisCatalin Vasilescua Mircea Olteanub Paul FlondorabaFundeni Clinical InstitutebPolitehnica University Bucharest

Objectives The innate immune system is responsible for the firstreaction to infections Despite the advanced research in themolecular biology our understanding of the biologic basis ofsepsis remains incomplete A complex system like the innateimmune system consist of a large number of components thatdisplay marked variability and a high degree of connectivity andinterdependence The emergent properties of such a systemcannot be understood by studying the isolated individual parts Ithas recently been appreciated that patients with severe sepsissuffer from altered innate and adaptive immune responsesleading to an impaired clearance of microorganisms Thisdecrease of the innate immune defense seems to be similar tothe phenomenon of endotoxin tolerance Efforts are underway byseveral groups to create mathematical models of sepsis and to

evaluate their usefulness in the design of clinical trials ofinvestigational new drugsMethods To gain further insights into the physiologic significanceof the immune depression of the patients with abdominal sepsis amathematical model of lipopolysaccharide (LPS) signaling basedon endotoxin tolerance was established This model was applied todifferent scenarios of sepsis and systemic inflammatory responsesyndrome The results of the numerical simulations were comparedwith the data obtained from patients using the ex vivo method ofwhole blood stimulation We tested the hypothesis that the anergy ofthe innate immunity in abdominal sepsis is similar to theexperimental phenomenon of endotoxin tolerance In our approachthe LPS kinetics was the same in all scenarios to compare thedifferent behavior of the immune response expressed by thedynamics of TNF-α concentration The differences among the 4cases consist in various states of the innate immunity system torespond to the same LPS stimulus as expressed by the TNF-αreleasing capacity In the present approach we incorporate 2 keyaspects of the innate immune response to LPS stimulation theLPS kinetics and the down-regulation of the LPS signalingmachinery after a previous LPS challenge The mathematicalmodel considers a nonautonomous first-order differential system of2 equations with 2 unknowns that is the concentrations of TNF-αand of the inhibitorResults This model has demonstrated its utility in simulating TNF-α release under endotoxin stimulation in systemic inflammatoryresponse syndrome and abdominal sepsis The results argue thatendotoxin tolerance (a main assumption that goes into theconstruction of the model) represents a keystone of the responseof innate immunity in systemic inflammatory response syndromeand sepsisConclusions Endotoxin tolerance may be a component of theimmune dysregulation that complicates the clinical evolution ofthe patient with abdominal sepsis However endotoxin toleranceis a dynamic not a static state characterized by continuouslychanging levels of circulating mediators Our results seem toshow a good concordance with the experimental and clinicaldata The proposed model is a good approximation forreal situations

doi101016jjcrc200710020

Measurement units may impact results of pathway analysisMary F McGuirea M Sriram Iyengara David MercerbaUniversity of Texas School of Health Information SciencesHouston USAbTrauma Research Center Department of Surgery University ofTexas Medical School at Houston Houston USA

ObjectivesDuring development of a method to identify biologicsignaling pathways implicated in multiple organ failure weexpected that the qualitative results of the molecular pathwayanalysis would be independent of the numerical units usedMethods We analyzed BioRad immunoassay data for 11 serumcytokines from 48 patients sustaining major torso trauma (exclud-ing severe head injuries) who met standardized criteria Cytokinelevels were measured every 4 hours from the start of resuscitationData in picograms per milliliter for the first 24 hours postinjurywere grouped by time period (4-hour spans starting at 2 hours) andby patient outcome of multiple organ failure (MOF) or not (NMOF)

342 Abstracts