IAI FigSearch
Home Help [Feedback] [For Subscribers] [Archive] [Search] [Contents]
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrowReprints and Permissions
Right arrow Copyright Information
Right arrow Books from ASM Press
Right arrow MicrobeWorld
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Duncan, D. D.
Right arrow Articles by Eldridge, J. H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Duncan, D. D.
Right arrow Articles by Eldridge, J. H.

 Previous Article  |  Next Article 

Infection and Immunity, July 2004, p. 4282-4285, Vol. 72, No. 7
0019-9567/04/$08.00+0     DOI: 10.1128/IAI.72.7.4282-4285.2004
Copyright © 2004, American Society for Microbiology. All Rights Reserved.

Multivariate Analysis of Cytokine Responses Identifies Distinctive Sensitivities to Lipopolysaccharide in Humans

David D. Duncan,* Lorrie Tiberio, and John H. Eldridge

Wyeth Research, Pearl River, New York 10965

Received 4 September 2003/ Returned for modification 29 January 2004/ Accepted 1 April 2004


   ABSTRACT
Top
 Abstract
Text
 References
 
We describe methods to identify high and low responders in a whole-blood assay of lipopolysaccharide-stimulated cytokine responses. Two multivariate measures of the cytokine responses both captured high and low responses for each of the four individual cytokines that were assayed.


   TEXT
Top
 Abstract
 Text
References
 
Sepsis is an important health concern (18) that has defied a simple and comprehensive description. Recognition of a familial component in sepsis (24) has prompted the search for predisposing host factors, and because cytokines mediate both protective responses and morbidity in infections (21, 26), they have been examined in this regard. Allelic phenotypes have been observed in vitro for various cytokines, including interleukin-6 (IL-6) (12), IL-8 (13), IL-10 (7, 9, 10), and tumor necrosis factor alpha (TNF-{alpha}) (11, 12), but the evidence for a substantial genetic influence on gram-negative sepsis is not conclusive (16). A previous study described only marginal influences of selected host factors and cytokine alleles on lipopolysaccharide (LPS)-stimulated production of IL-6, IL-8, IL-10, and TNF-{alpha} (25). However, we observed that these responses were somewhat correlated (r = 0.438 to 0.599) and suggestive of interactions or concerted control of the four cytokine responses. Here we describe a simple approach to quantify and characterize the overall cytokine response as reflected by the production of these four cytokines and discuss how this technique may complement efforts to understand and treat sepsis.

The whole-blood culture and cytokine assays have been described previously (25). The time periods for phlebotomy, culture, and assay were strictly followed (phlebotomy, 8 to 9 a.m.; 20 h of culture with or without 0.01 µg of Escherichia coli LPS/ml). Means (and standard deviations) of the log transforms of LPS-stimulated cytokine levels were as follows: IL-6, 4.289 (0.261); IL-8, 3.900 (0.289); IL-10, 2.910 (0.180); and TNF-{alpha}, 2.562 (0.316). Z scores of these transforms were used because of the differing magnitudes and variances of the cytokine responses; without standardization the analyses would be dominated by numerical differences rather than comparative differences in response levels.

Two multivariate measures of the responses were generated, using a recently reviewed approach (15). Log-transformed, standardized cytokine levels were plotted as a vector in four dimensions (IL-6, IL-8, IL-10, and TNF-{alpha}), where average responses would be represented by Z = 0 on each of the axes. As illustrated in Fig. 1A, we determined the cosine of the angle between a subject's response vector and the vector of each of the other subjects in turn. The resulting matrix was analyzed, using hierarchical clustering of the Euclidean distances among the cosine values by using complete linkage of the emergent clusters (14). The data were most simply described with high- and low-response clusters (n = 68 and 39 subjects, respectively). The subjects were grouped as high or low responders, and as shown in Fig. 2, comparisons for each of the four responses were significant (P < 0.0001, Student's t test). The coefficients of determination (20) were 0.18, 0.26, 0.50, and 0.47 for IL-6, IL-8, IL-10, and TNF-{alpha}, respectively (as determined by analysis of variance). These proportions of explained variance were notably higher than those for the univariate analyses (25), suggesting that the multivariate analysis provided a more comprehensive model of the responses. Determination of the second multivariate measure, the response index, is illustrated in Fig. 1B. Response indices were determined for all subjects in the study and grouped by high- or low-response category, and these were also found to differ significantly (Fig. 3) (P < 0.05, rank sums test). The response index was determined weekly for nine individuals and remained constant for each subject over a 2-month period, indicating that a single determination of the index adequately reflected the response independently of time (data not shown).



View larger version (10K):
[in this window]
[in a new window]
  		FIG. 1. The two multivariate analyses used in this study. In this illustration, two of the four cytokine responses are plotted on the x and y axes for three subjects, a, b, and c. (A) Categorical analysis. The angles between the response vectors of subjects a and b and a and c are illustrated. The cosine of the angles between all paired combinations of the subjects' response vectors was determined, and the resulting matrix was analyzed by clustering. (B) Quantitative analysis. A low-response reference vector was plotted as Z = –5 in all four dimensions; two dimensions are shown. The angle between the response vector of subject a and the reference vector is indicated. The response index is the cosine of this angle. For subject a, this angle is approximately 180o and the response index is approximately –1. The approximate angle and response index for b would be 90o and 0, respectively, and for c they would be 0o and 1, respectively.

 


View larger version (12K):
[in this window]
[in a new window]
  		FIG. 2. Comparisons of cytokine responses of high- and low-responding groups. The angles among the four-parameter response vectors were clustered on the basis of their similarity. Subjects were grouped into one of the final two clusters, and their cytokine responses were compared. Each comparison was significant (P < 0.0001, as determined by Student's t test).

 


View larger version (21K):
[in this window]
[in a new window]
  		FIG. 3. The response index. The cosine of the angle between each subject's response vector and a low-response reference vector was determined. Subjects were grouped by high- or low-response cluster, and the frequency distribution of the response indices of these two groups was plotted.

 
It has been observed that the incidence and severity of sepsis is greater in men than women (18, 23, 28). Gender-dependent differences in cytokine responses have also been observed (3, 12, 17, 22). In a subset of the subjects analyzed in this study, we found that men made significantly more IL-6, IL-8, and TNF-{alpha} but not IL-10 (25). Comparisons for all 107 subjects in this study gave analogous results (Fig. 4). Of note, the distributions for men and women were similar, except that the IL-6, IL-8, and TNF-{alpha} responses for men were right-shifted. Next, the multivariate response measures of men and women were compared. The proportion of men and women within the high- and low-response groups was nearly significant (P = 0.058, chi-square test), and the response index was significantly higher in men than women (P < 0.01, rank sums test). Interestingly, the index distributions were quite distinctive for men and women (Fig. 5). Men had indices in the intermediate-to-high ranges, and women had a biphasic distribution of high and low indices.



View larger version (17K):
[in this window]
[in a new window]
  		FIG. 4. Comparison of the individual cytokine responses of men (solid line) versus those of women (broken line). Significant comparisons are indicated (determined by Student's t test).

 


View larger version (16K):
[in this window]
[in a new window]
  		FIG. 5. Comparison of the overall cytokine response of men (solid line) versus that of women (broken line).

 
The multivariate response measures described in this report have two useful features. First, they captured significantly different responses for each of the four cytokines (Fig. 2) while summarizing them in a single measure. Second, they provided a more comprehensive model of the cytokine responses, since they accounted for a greater proportion of the response variation than the univariate analyses. A comparison of Fig. 4 and 5 provides a striking visual contrast as well: the univariate analyses indicated that men tended to have higher cytokine responses than women, but the multivariate analysis showed that the difference could be attributed very specifically to differing frequencies of high and low responders among men and women.

Consequently, we anticipate several applications for this technique. Conceivably, the correlations among the cytokines point to a dimension in the responses greater than that of the individual cytokines. By identifying subjects differing at this level, it becomes possible to characterize the basis of the relationships among the individual responses. For example, interindividual differences in the TLR4 signaling pathway are likely to influence many aspects of the response to LPS (1, 19), and in fact, greater NF{kappa}B mobilization has been seen in nonsurviving septic patients than in survivors (2, 4). Alternatively, interactions among the cytokines may intensify their correlation, as, for example, TNF-{alpha}-enhanced production of IL-8 and IL-10 (8, 27). In any case, stratification of subjects by their overall response would facilitate the evaluation of such explanations. There may be clinical applications for the multivariate measures as well. There have been many attempts to correlate prognosis in sepsis with circulating levels of various pro- and antiinflammatory cytokines, with various degrees of success. Although the complexity of cytokine responses in sepsis is appreciated, to our knowledge there have been few attempts to capture multiple cytokine responses in a summarizing metric (5, 6). Since it has been suggested that production of pro- and antiinflammatory cytokines together can lead to complications in sepsis (21), representation of the concerted cytokine response may help clarify the role cytokines play in the disease. In particular, stratification by the multivariate response measures could facilitate experimental evaluations of targeted therapies where it is possible to determine the subjects' response level prior to endotoxin exposure. However, it remains to be seen if our observations apply to the ex vivo responses, or cytokine levels in plasma, of sepsis patients. We have very preliminary evidence that responder status may have some relevance to cytokine levels in plasma: a comparison of IL-8 levels in the unstimulated control cultures of high and low responders was significant (P < 0.05, Student's t test; no other cytokines were consistently detected in the control cultures).

Lastly, note that discriminant functions (14) were generated by using the clustered data as a training sample. With these functions, subjects can be classified as high or low responders solely on the basis of results from the whole-blood assays, without the need for vector analysis and clustering. There was 95.3% agreement in classification by clustering and the discriminant functions, even without adjustments for unequal frequencies of high and low responders or for cost of misclassification, both of which can increase the agreement between the two methods (14). Thus, it would be possible to generate a reference set of data in a primary lab, cluster the responses, and generate the functions. The discriminant functions could then be used for classification by any labs whose methods are consistent with those of the reference lab. However, this approach requires careful consideration of the reference population: it should include subjects whose responses span the clinical spectrum with regard to endotoxemia to ensure that the four-parameter-response space encompasses the responses of interest, e.g., responses from subjects that may be susceptible or resistant to sepsis because of extreme levels of cytokine production.


   FOOTNOTES
 
* Corresponding author. Mailing address: 2005 Countrywood Ct., Encinitas, CA 92024. Phone: (760) 944-0310. E-mail: d_duncan{at}att.net. Back

Editor: F. C. Fang


   REFERENCES
Top
 Abstract
 Text
 References
 
1. Arbour, N. C., E. Lorenz, B. C. Schutte, J. Zabner, J. N. Kline, M. Jones, K. Frees, J. L. Watt, and D. A. Schwartz. 2000. TLR4 mutations are associated with endotoxin hyporesponsiveness in humans. Nat. Genet. 25:187-191.[CrossRef][Medline]
2. Arnalich, F., E. Garcia-Palomero, J. López, M. Jiménez, R. Madero, J. Renart, J. J. Vázquez, and C. Montiel. 2000. Predictive value of nuclear factor {kappa}B activity and plasma cytokine levels in patients with sepsis. Infect. Immun. 68:1942-1945.[Abstract/Free Full Text]
3. Asai, K., N. Hiki, Y. Mimura, T. Ogawa, K. Unou, and M. Kaminishi. 2001. Gender differences in cytokine secretion by human peripheral blood mononuclear cells: role of estrogen in modulating LPS-induced cytokine secretion in an ex vivo septic model. Shock 16:340-343.[Medline]
4. Böhrer, H., F. Qiu, T. Zimmermann, Y. Zhang, T. Jllmer, D. Männel, B. W. Böttiger, D. M. Stern, R. Waldherr, H.-D. Saeger, R. Ziegler, A. Bierhaus, E. Martin, and P. P. Nawroth. 1997. Role of NF{kappa}B in the mortality of sepsis. J. Clin. Investig. 100:972-985.[Medline]
5. Calvano, S. E., S. M. Coyle, K. S. Barbosa, P. S. Barie, and S. F. Lowry. 1998. Multivariate analysis of 9 disease-associated variables for outcome prediction in patients with sepsis. Arch. Surg. 133:1347-1350.[Abstract/Free Full Text]
6. Casey, L. C., R. A. Balk, and R. C. Bone. 1993. Plasma cytokine and endotoxin levels correlate with survival in patients with the sepsis syndrome. Ann. Intern. Med. 119:771-778.[Abstract/Free Full Text]
7. Crawley, E., R. Kay, J. Sillibourne, P. Patel, I. Hutchinson, and P. Woo. 1999. Polymorphic haplotypes of the interleukin-10 5' flanking region determine variable interleukin-10 transcription and are associated with particular phenotypes of juvenile rheumatoid arthritis. Arthritis Rheum. 42:1101-1108.[CrossRef][Medline]
8. DeForge, L. E., J. S. Kenney, M. L. Jones, J. S. Warren, and D. G. Remick. 1992. Biphasic production of IL-8 in lipopolysaccharide (LPS)-stimulated human whole blood. Separation of LPS- and cytokine-stimulated components using anti-tumor necrosis factor and anti-IL-1 antibodies. J. Immunol. 148:2133-2141.[Abstract]
9. Eskdale, J., G. Gallagher, C. L. Verweij, V. Keijsers, R. G. J. Westendorp, and T. W. J. Huizinga. 1998. Interleukin 10 secretion in relation to human IL-10 locus haplotypes. Proc. Natl. Acad. Sci. USA 95:9465-9470.[Abstract/Free Full Text]
10. Gibson, A. W., J. C. Edberg, J. Wu, R. G. J. Westendorp, T. W. J. Huizinga, and R. P. Kimberly. 2001. Novel single nucleotide polymorphisms in the distal IL-10 promoter affect IL-10 production and enhance the risk of systemic lupus erythematosus. J. Immunol. 166:3915-3922.[Abstract/Free Full Text]
11. Hajeer, A. H., and I. V. Hutchinson. 2001. Influence of TNF{alpha} gene polymorphisms on TNF{alpha} production and disease. Hum. Immunol. 62:1191-1199.[CrossRef][Medline]
12. Heesen, M., B. Bloemeke, N. Heussen, and D. Kunz. 2002. Can the interleukin-6 response to endotoxin be predicted? Studies of the influence of a promoter polymorphism of the interleukin-6 gene, gender, the density of the endotoxin receptor CD14, and inflammatory cytokines. Crit. Care Med. 30:664-669.[CrossRef][Medline]
13. Hull, J., A. Thomson, and D. Kwiatkowski. 2000. Association of respiratory syncytial virus bronchiolitis with the interleukin 8 gene region in UK families. Thorax 55:1023-1027.[Abstract/Free Full Text]
14. Johnson, R. A., and D. W. Wichern. 2002. Applied multivariate statistical analysis, 5th ed. Prentice Hall, Upper Saddle River, N.Y.
15. Kuruvilla, F. G., P. J. Park, and S. L. Schreiber. 2002. Vector algebra in the analysis of genome-wide expression data. Genome Biol. 3:1-11.
16. Lin, M. T., and T. E. Albertson. 2004. Genomic polymorphisms in sepsis. Crit. Care Med. 32:569-579.[CrossRef][Medline]
17. Majetschak, M., B. Christensen, U. Obertacke, C. Waydhas, A. E. Schindler, D. Nast-Kolb, and F. U. Schade. 2000. Sex differences in posttraumatic cytokine release of endotoxin-stimulated whole blood: relationship to the development of severe sepsis. J. Trauma Inj. Infect. Crit. Care 48:832-839.[CrossRef]
18. Martin, G. S., D. M. Mannino, S. Eaton, and M. Moss. 2003. The epidemiology of sepsis in the United States from 1979 through 2000. N. Engl. J. Med. 348:1546-1554.[Abstract/Free Full Text]
19. Michel, O., T. D. LeVan, D. Stern, M. Dentener, J. Thorn, D. Gnat, M. L. Beijer, P. Cochaux, P. G. Holt, F. D. Martinez, and R. Rylander. 2003. Systemic responsiveness to lipopolysaccharide and polymorphisms in the toll-like receptor 4 gene in human beings. J. Allergy Clin. Immunol. 112:923-929.[CrossRef][Medline]
20. Montgomery, D. C. 2001. Design and analysis of experiments, 5th ed. John Wiley and Sons, New York, N.Y.
21. Pinsky, M. R. 2001. Sepsis: a pro- and anti-inflammatory disequilibrium syndrome. Contrib. Nephrol. 132:354-366.
22. Reuss, E., R. Fimmers, A. Kruger, C. Becker, C. Rittner, and T. Höhler. 2002. Differential regulation of interleukin-10 production by genetic and environmental factors—a twin study. Genes Immun. 3:407-413.[CrossRef][Medline]
23. Schröder, J., V. Kahlke, K.-H. Staubach, P. Zabel, and F. Stüber. 1998. Gender differences in human sepsis. Arch. Surg. 133:1200-1205.[Abstract/Free Full Text]
24. Søorensen, T. I. A., G. G. Nielsen, P. K. Andersen, and T. W. Teasdale. 1988. Genetic and environmental influences on premature death in adult adoptees. N. Engl. J. Med. 318:727-732.[Abstract]
25. Tiberio, L., L. Fletcher, J. H. Eldridge, and D. D. Duncan. 2004. Host factors impacting the innate response in humans to the candidate adjuvants RC529 and monophosphoryl lipid A. Vaccine 22:1515-1523.[CrossRef][Medline]
26. van der Poll, T. 2001. Immunotherapy of sepsis. Lancet Infect. Dis. 1:165-174.[CrossRef][Medline]
27. Wanidworanun, C., and W. Strober. 1993. Predominant role of tumor necrosis factor-{alpha} in human monocyte IL-10 synthesis. J. Immunol. 151:6853-6861.[Abstract]
28. Wichmann, M. W., D. Inthorn, H. J. Andress, and F. W. Schildberg. 2000. Incidence and mortality of severe sepsis in surgical intensive care patients: the influence of patient gender on disease process and outcome. Intensive Care Med. 26:167-172.[CrossRef][Medline]

Infection and Immunity, July 2004, p. 4282-4285, Vol. 72, No. 7
0019-9567/04/$08.00+0     DOI: 10.1128/IAI.72.7.4282-4285.2004
Copyright © 2004, American Society for Microbiology. All Rights Reserved.




This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrowReprints and Permissions
Right arrow Copyright Information
Right arrow Books from ASM Press
Right arrow MicrobeWorld
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Duncan, D. D.
Right arrow Articles by Eldridge, J. H.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Duncan, D. D.
Right arrow Articles by Eldridge, J. H.

Home Help [Feedback] [For Subscribers] [Archive] [Search] [Contents]
J. Bacteriol. J. Virol. Eukaryot. Cell
Microbiol. Mol. Biol. Rev. Clin. Vaccine Immunol. All ASM Journals