Prospective evaluation of S100A12 and S100A8/A9 (calprotectin) in dogs with sepsis or the systemic inflammatory response syndrome

Abstract

Pattern recognition receptors (e.g., S100A12 or S100A8/A9) hold promise as inflammatory biomarkers. We prospectively determined and compared serum S100A12 and S100A8/A9 concentrations in dogs with sepsis (n = 11) or systemic inflammatory response syndrome (SIRS; n = 8) over a 3-d period with each other, healthy controls (n = 50), and other clinical and clinicopathologic variables. Serum S100A12 and S100A8/A9 concentrations were significantly higher in dogs with sepsis or SIRS (all p < 0.05) at the time of hospital admission (day 1) compared to healthy controls, with no differences between patient groups. However, septic dogs had significantly lower serum S100A12 concentrations on day 2 and day 3 (both p < 0.05) compared to dogs with SIRS. Likewise, dogs with sepsis had significantly lower S100A8/A9 concentrations on day 2 (p < 0.05). Neither serum S100A12 nor S100A8/A9 concentrations were associated with survival to discharge. Our results suggest a differential expression of the S100/calgranulins between dogs with sepsis and those with SIRS. Serum S100A12 or S100A8/A9 concentration at the time of hospital admission did not differentiate dogs with sepsis from those with SIRS, but the trend of S100/calgranulin concentrations during the following 24–48 h may be a useful surrogate marker for differentiating sepsis from SIRS.

Early diagnosis, differentiation, and prognosis of sepsis and systemic inflammatory response syndrome (SIRS) continue to challenge healthcare providers in both human and veterinary medicine.¹⁶ Despite advances in treatment strategies, a worldwide review stated that the overall mortality rate for human sepsis patients is ~50%.²¹ The pathogenesis of sepsis and SIRS involves a complex interplay of factors leading to dysregulated immunity. Once a local inflammatory response is initiated, complicating factors (e.g., dysregulation in pro- and anti-inflammatory cytokines, a state of immunosuppression, or an overwhelming infection) can worsen the existing condition with the potential to ultimately culminate in sepsis or SIRS.¹⁷

Proteins of the Ca²⁺-binding S100 family (including >20 proteins) are involved in intracellular signal transduction, cell differentiation, regulation of cell motility, transcription, and cell cycle progression.¹⁵ In addition to these cellular functions, the S100/calgranulins, such as S100A12 protein and the S100A8/A9 complex, can be secreted from cells and act as damage-associated patterns (DAMPs) with cytokine- and chemokine-like extracellular functions.¹⁵

Species-specific assays for the quantification of S100A12 and S100A8/A9 in canine specimens have been developed and analytically validated,^8,9 but serum concentrations of neither S100A12 nor S100A8/A9 have been evaluated in dogs with sepsis or noninfectious SIRS, to our knowledge. C-reactive protein (CRP), a positive acute-phase protein, is a useful marker of systemic inflammation in people,²⁰ and studies have shown that serum CRP concentrations are increased in dogs with inflammatory conditions.^5,10,13,23 We compared serum concentrations of S100A12 and S100A8/A9 in dogs with sepsis or noninfectious SIRS over a 3-d period as well as with serum concentrations of these markers in healthy control dogs. We hypothesized that serum S100/calgranulin concentrations are increased in dogs with sepsis or SIRS. We also hypothesized that changes in serum S100/calgranulins could mirror those of CRP, interleukin (IL)-6, and tumor necrosis factor–alpha (TNF-α), and that such changes correlate with the acute patient physiologic and laboratory evaluation (APPLE) scores.

We designed a prospective, observational clinical study. We included patients that were admitted to the Texas A&M University Small Animal Intensive Care Unit (ICU; College Station, TX) and that met previously published criteria for sepsis or SIRS^2,6 as well as healthy controls. Informed owner consent was obtained for all dogs enrolled. The study was reviewed and approved by the Texas A&M University Clinical Review Research Committee (CRRC 2010-18).

A total of 69 dogs were enrolled in the study: 8 dogs that met the SIRS criteria without evidence of an infection (SIRS group), 11 dogs with evidence of an infection along with SIRS criteria (SEPSIS group), and 50 apparently healthy control dogs (CONTROL group). Diagnostic criteria for SIRS were the presence of at least 2 of the following: hypo- or hyperthermia (<37.8°C or >39.4°C), tachycardia (heart rate >140/min), tachypnea (respiratory rate >20/min), leukocytosis (>16 × 10⁹ white blood cells [WBC]/L) or leukopenia (<6 × 10⁹ WBC/L), and/or >3% band neutrophils.^2,6 Beyond those, the criterion for inclusion in the SEPSIS group was the confirmation of an infection via histologic, microbiologic, cytologic, and/or viral antigen evaluation. Some of the data from these dogs have been reported previously.¹⁰

The full scoring system assessment of a dog with sepsis or SIRS during the first evaluation included the use of the diagnosis-independent APPLE score. This scoring system includes 10 parameters (plasma creatinine, albumin, total bilirubin, WBC count, blood oxygen saturation, mentation score, respiratory rate, age, blood lactate, and the presence of effusion); the cumulative APPLE_full score can range from 0 to 80.⁷ The abbreviated scoring system included 5 parameters (blood glucose, plasma albumin, blood lactate, platelet count, and mentation score); the cumulative APPLE_fast score can range from 0 to 50.⁷ Patients were all hospitalized in ICU and were monitored according to the standard hospital ICU protocol. Treatment protocols for each patient were determined by the attending clinician and were not monitored nor altered by any of the study personnel.

Primary outcome measure was survival to discharge from the hospital. All dogs that did not survive to discharge were euthanized (intravenous pentobarbital injection) given a grave prognosis and moribund condition. No dogs were euthanized because of financial constraints.

The CONTROL group was comprised of apparently healthy dogs with no evidence of an infection, inflammation, or other conditions (e.g., diabetes mellitus). Physical examination, complete blood count, serum biochemistry profile, and historical information were collected to determine each dog’s health status.

Blood samples were obtained on the day of study enrollment (day 1) from all dogs, and on days 2 and 3 from dogs in the SEPSIS or SIRS groups. Serum was separated and immediately stored at −80°C until further testing within 10 mo. Serum concentrations of S100A12 and S100A8/A9 were determined via species-specific immunoassays.^8,11 Intra- and inter-individual coefficients of variation of the serum S100A12 assay were ≤8.1% and ≤7.8%, respectively,⁸ and were ≤12.7% and ≤10.0% for the serum S100A8/A9 test.¹¹ Serum CRP, plasma IL-6, and TNF-α were measured as described previously.¹⁰

Data were analyzed with a Shapiro–Wilk test for normality and a Brown–Forsythe test for equality of variances. Differences or associations between groups were evaluated by a Wilcoxon rank-sum or Kruskal–Wallis test (continuous variables) or by the Fisher exact test (categorical variables). A Spearman correlation coefficient ρ served to assess potential relationships between continuous variables. A receiver operating characteristic curve was constructed, with a likelihood ratio used to determine optimum cutoff concentrations, for calculation of sensitivities and specificities. Statistical significance was set at p ≤0.05. Commercial statistical software packages (JMP Pro v.13.0, SAS Institute, Cary, NC; Prism v.7.0, GraphPad Software, San Diego, CA) were used for all statistical analyses.

More than 20 canine breeds were represented in our study. Sixteen of the 50 CONTROL dogs (32%) were mixed-breed, and 15 different breeds were represented in the SEPSIS and SIRS groups, including 2 each of the following breeds: German Shepherd dogs, Labrador Retrievers, Boxers, and Rottweilers. Sex (p = 0.485) or age (p = 0.146) distribution did not differ among the 3 groups (Table 1). Neither APPLE_full scores (p = 0.456) nor APPLE_fast scores (p = 0.431) differed between SEPSIS and SIRS. The diagnoses of dogs enrolled in our study varied in both disease groups (Table 2), but parvoviral enteritis and septic peritonitis accounted for 8 of 11 SEPSIS cases (73%). Outcome did not differ between SEPSIS and SIRS (p = 0.637); 5 SIRS dogs (62%) and 8 dogs with sepsis (73%) survived to discharge.

Table 1.

Demographic data and serum S100 protein concentrations for all dogs enrolled in the S100/calgranulin study (n = 69).

Characteristic	Unit	Sepsis	SIRS	Controls	p value*
Total number	n	11	8	50	NA
Median age	y	3 (1–7)	7 (3–11)	5 (3–8)	0.146
Sex ratio (male:female)		7:4	3:5	29:21	0.485
APPLEfull	median	34 (24–41)	35 (30–48)	NA	0.456
APPLEfast	median	24 (20–28)	28 (20–30)	NA	0.431
Median serum S100A8/A9†	mg/L
Day 1		20.3 (5.5–33.2)a	19.7 (7.7–22.1)a	6.6 (4.2–9.1)b	0.005
Day 2		13.2 (4.2–24.2)b	31.6 (12.1–53.2)a	6.6 (4.2–9.1)b	0.001
Day 3		8.6 (2.4–25.9)a,b	35.4 (23.1–74.3)a	6.6 (4.2–9.1)b	0.006
Median serum S100A12†	μg/L
Day 1		305 (92–614)a	260 (142–368)a	129 (84–186)b	0.019
Day 2		221 (29–307)b	525 (117–1292)a	129 (84–186)b	0.029
Day 3		104 (27–442)b	594 (400–1946)a	129 (84–186)b	0.006

APPLE = acute patient physiologic and laboratory evaluation score; NA = not applicable; SIRS = systemic inflammatory response syndrome. Numbers in parentheses are interquartile ranges.

^*For an association or difference among SIRS, SEPSIS, and CONTROL groups (p values in boldface indicate significance).

^†For each parameter, medians not sharing a common superscript (^a,b) are significantly different at p < 0.05.

Table 2.

Disease distribution and survival outcome of dogs in the SEPIS (n = 11) and SIRS (n = 8) groups.

SEPSIS group			SIRS group
Disease	Survival to discharge (n)	Death or euthanasia (n)	Disease	Survival to discharge (n)	Death or euthanasia (n)
Septic peritonitis	3	2	GI foreign body, surgery	2	0
Parvoviral enteritis	3	0	GI obstruction, surgery	0	1
Septicemia, pyelonephritis	1	0	GI ulceration, hematemesis	1	0
Septicemia, body wall abscess post-abdominal surgery	1	0	Pancreatic adenocarcinoma	1	0
RMSF	0	1	Primary IMHA	1	0
			Lymphoma	0	1
			MCT, acute pancreatitis	0	1

GI = gastrointestinal; IMHA = immune-mediated hemolytic anemia; MCT = mast cell tumor; RMSF = Rocky Mountain spotted fever; SIRS = systemic inflammatory response syndrome.

Serum S100A12 concentrations differed among the 3 groups of dogs (p = 0.019), being significantly higher in the SEPSIS (p = 0.032) and SIRS (p = 0.035) compared to the CONTROL group on day 1 (Table 1, Fig. 1, Supplementary Fig. 1), with no difference in serum S100A12 detected between disease groups on day 1 (p = 0.901). S100A8/A9 concentrations were also significantly higher in SEPSIS (p = 0.019) and SIRS (p = 0.035) compared to CONTROL dogs (Table 1, Fig. 2, Supplementary Fig. 2), without a significant difference between SEPSIS and SIRS on day 1 (p = 0.837). Serum S100A12 or S100A8/A9 deficiency was not detected in any dog. Day 1 serum S100/calgranulin levels were highly correlated with each other, correlated moderately with serum CRP and WBC counts, but did not correlate with systemic IL-6 or TNF-α concentration, nor with APPLE_full or APPLE_fast scores (Table 3). Serum S100/calgranulin concentrations were highly correlated with each other also on days 2 (ρ = 0.96, p < 0.0001) and 3 (ρ = 0.91, p < 0.0001). In addition to a moderate correlation with serum CRP on day 2 (ρ = 0.56, p = 0.015; and ρ = 0.59, p = 0.010, respectively) and for S100A8/A9 also on day 3 (ρ = 0.65, p = 0.011), serum S100A8/A9 correlated with IL-6 on day 2 (ρ = 0.50, p = 0.040). None of the remaining relationships among these markers reached significance, and there was no correlation with serum TNF-α on day 2 or 3 (all p >0.05).

Figure 1.

Serum S100A12 concentrations in dogs with sepsis (n = 11) or SIRS (n = 8) and in healthy control dogs (n = 50). Compared to healthy control dogs (median = 129 μg/L, IQR = 84–186 μg/L; n = 50), serum concentrations of S100A12 were increased in dogs with sepsis (median = 305 μg/L, IQR = 92–614 μg/L; p = 0.032) or SIRS (median = 260 μg/L, IQR = 142–368 μg/L; p = 0.035), but no difference was observed for both patient groups combined at hospital admission (day 1; p = 0.901). Serum S100A12 concentrations remained significantly higher in dogs with noninfectious SIRS on day 2 (median = 525 μg/L, IQR = 117–1,292 μg/L) and day 3 (median = 594 μg/L, IQR = 400–1,946 μg/L) compared to dogs with sepsis on day 2 (median = 221 μg/L, IQR = 29–307 μg/L; p = 0.046) and day 3 (median = 104 μg/L, IQR = 27–442 μg/L; p = 0.020).

Solid lines = medians; gray-shaded area between dashed lines = reference interval (33–225 μg/L); symbols (▼, •, ○, ■, and □) = serum S100A12 concentrations in individual dogs (• and ■ represent survivors; ○ and □ are non-survivors).

Figure 2.

Serum S100A8/A9 concentrations in dogs with sepsis (n = 11) or SIRS (n = 8) and in healthy control dogs (n = 50). Compared to healthy control dogs (median = 6.6 mg/L, IQR = 4.2–9.1 mg/L; n = 50), serum concentrations of S100A8/A9 were increased in dogs with sepsis (median = 20.3 mg/L, IQR = 5.5–33.2 mg/L; p = 0.012) or SIRS (median = 19.7 mg/L, IQR = 7.7–22.1 mg/L; p = 0.014), but no difference was observed between both patient groups at hospital admission (day 1; p = 0.837). Serum S100A8/A9 concentrations remained significantly higher in dogs with noninfectious SIRS on day 2 (median = 31.6 mg/L, IQR = 12.1–53.2 mg/L) and day 3 (median = 35.4 mg/L, IQR = 23.1–74.3 mg/L) compared to dogs with sepsis on day 2 (median = 13.2 mg/L, IQR = 4.2–24.2 mg/L; p = 0.037) and day 3 (median = 8.6 mg/L, IQR = 2.4–25.9 mg/L), but the difference on day 3 did not reach significance (p = 0.056).

Solid lines = medians; gray-shaded area between dashed lines = reference interval (0.9–11.9 mg/L); symbols (▼, •, ○, ■, and □) = serum S100A8/A9 concentrations in individual dogs (• and ■ represent survivors; ○ and □ are non-survivors).

Table 3.

Correlation of serum S100A12 and S100A8/A9 concentrations with clinical and clinicopathologic variables in dogs with sepsis or SIRS on the day of ICU admission (day 1).

Parameter	Spearman ρ correlation coefficient (p value)
	Serum S100A12 concentration	Serum S100A8/A9 concentration
APPLEfull score	0.32 (0.180)	0.38 (0.105)
APPLEfast score	0.12 (0.623)	0.11 (0.657)
WBC count	0.79 (<0.0001)	0.72 (<0.0001)
Serum S100A12 concentration	NA	0.92 (<0.0001)
Serum S100A8/A9 concentration	0.92 (<0.0001)	NA
Serum CRP concentration	0.28 (0.025)	0.32 (0.009)
Serum IL-6 concentration	0.04 (0.881)	0.19 (0.463)
Serum TNF-α concentration	−0.28 (0.286)	−0.27 (0.295)

APPLE = acute patient physiologic and laboratory evaluation score; CRP = C-reactive protein; IL-6 = interleukin-6; NA = not applicable; TNF-α = tumor necrosis factor–alpha; WBC = white blood cell. Cells highlighted in bold indicate a statistically significant correlation (p ≤ 0.05).

Dogs with sepsis had significantly lower serum S100A12 concentrations on day 2 (p = 0.046) and day 3 (p = 0.020) compared to dogs with SIRS (Table 1, Fig. 1, Supplementary Fig. 1). Dogs with sepsis also had significantly lower serum S100A8/A9 concentrations on day 2 (p = 0.037), whereas the difference on day 3 did not reach significance (p = 0.056; Table 1, Fig. 2, Supplementary Fig. 2). Opposite trends in serum S100/calgranulins from day 1 to day 2 were significantly associated with a diagnosis of sepsis (decrease) versus SIRS (increase; p = 0.048 and p = 0.013, respectively). A day 2 serum S100A12 concentration ≤458 µg/L or a day 2 serum S100A8/A9 concentration ≤24.4 mg/L distinguished SEPSIS from SIRS with a sensitivity of 100% (95% confidence interval [CI]: 72–100%) and 91% (59–100%), respectively, and a specificity of 57% (18–90%) and 71% (29–96%), respectively (Fig. 3). Specificities of ≥80% (S100A12 = 86% [95% CI: 42–100%]; S100A8/A9 = 86% [42–100%]) for separating SEPSIS from SIRS yielded sensitivities of 36% (11–69%) and 46% (17–77%), respectively, at a cutoff concentration of 84.7 µg/L and 10.6 mg/L, respectively.

Figure 3.

Receiver operating characteristic curves for A. serum S100A12 and B. S100A8/A9 on day 2 to distinguish dogs with sepsis from dogs with SIRS. A serum S100A12 concentration of 458 μg/L and a S100A8/A9 concentration of 24.4 mg/L were determined to be the optimal cutoff values between dogs with sepsis or noninfectious SIRS, with an area under the curve of 79% and 81%, respectively.

The dashed lines indicate the lines of identity.

Serum S100A8/A9 concentrations at admission (day 1) did not differ between survivors (median = 18.9 mg/L, interquartile range [IQR] = 7.0–27.9 mg/L; n = 13) and non-survivors (median = 21.7 mg/L, IQR = 6.1–64.6 mg/L; n = 6; p = 0.511). Serum S100A12 levels were also not different between survivors (median = 305 µg/L, IQR = 112–495 µg/L) and non-survivors (median = 325 µg/L, IQR = 146–930 µg/L; p = 0.693). The % change of serum S100A12 and S100A8/A9 from day 1 to 2 was also not significantly different in survivors versus non-survivors (p = 0.490 and p = 0.767, respectively), and likewise from day 1 to 3 (p = 0.411 and p = 0.523, respectively) and day 2 to 3 (p = 0.927 and p = 0.927, respectively). APPLE_fast, but not APPLE_full (p > 0.05) scores were significantly lower in survivors (median = 22, IQR = 19–27) than non-survivors (median = 30, IQR = 26–36; p = 0.011).

We found that serum S100A12 and S100A8/A9 concentrations on the day of patient admission (day 1) were significantly higher in dogs with sepsis or SIRS than in healthy controls. Our results mirror studies in human sepsis patients in which increased circulating concentrations of S100A12^1,14,22 and also S100A8/A9 heterodimers (calprotectin)²⁴ were found; these proteins are known to play a significant role in the pathogenesis of sepsis in humans.

The source of increased serum S100/calgranulin concentrations in dogs with sepsis or SIRS on the day of hospital admission and the differing trends of both biomarkers between these 2 disease groups during hospitalization cannot be determined by our results. One explanation could be the effect of an increased release of endogenous corticosteroids because of cytokine-mediated activation of the hypothalamic–pituitary–adrenal axis,³ given that high-dose steroid treatment can increase the expression of S100A8.¹² However, this would not explain the increased serum concentrations of S100A12 that was shown to be unaffected by exogenous steroid treatment.⁹ Contribution of a critical illness–related corticosteroid insufficiency¹⁸ could also explain our results, and an effect of individual treatment protocols cannot be entirely excluded.

Neither serum S100A12 nor S100A8/A9 concentration at the time of hospital admission was able to differentiate between sepsis and SIRS. However, lack of an increase in S100A12 and also S100A8/A9 concentrations within 24–48 h after hospital admission was associated with a diagnosis of sepsis, and day 2 serum S100A12 and S100A8/A9 concentrations distinguished dogs with sepsis from SIRS patients with high sensitivity and moderate specificity. Thus, changes in both biomarkers as early as the patients’ second day of hospitalization might allow tailoring the management of these patients early in the treatment process. However, further studies are warranted to determine if these values remain significantly increased in SIRS patients throughout the course of disease. Also, whether the difference in serum S100/calgranulin concentrations between dogs with sepsis versus those with SIRS results from a decreased expression during the recovery from severe sepsis¹⁹ or sepsis-induced immune dysfunction⁴ cannot be determined because we did not analyze reference markers for immunosuppression. Further research is also warranted to explore the concept to potentially use DAMPs (such as the S100/calgranulins) to modulate inflammatory processes as a therapeutic strategy.¹⁶

Given that serum S100/calgranulin concentrations were significantly higher in dogs with SIRS than dogs with sepsis on days 2 and 3, these biomarkers outperformed serum CRP concentrations, which could not distinguish both patient groups on any of the 3 d of hospitalization.¹⁰ However, the difference in serum CRP concentrations between healthy dogs and dogs with sepsis or SIRS agrees with previous studies reporting significantly increased serum CRP concentrations in dogs with sepsis or SIRS.^5,13,23 Further, we found moderate correlation between serum CRP and S100/calgranulin concentrations, suggesting that all 3 analytes reflect the systemic inflammatory response but with some differences (spatial, temporal, or both) in their expression.

We found no significant differences in serum S100/calgranulin concentrations between survivors and non-survivors. Thus, neither biomarker appears to accurately predict survival in dogs with sepsis or SIRS. Further studies including larger numbers of dogs and serial blood testing throughout a more prolonged patient follow-up will be needed to evaluate the potential of the S100/calgranulins as predictors of mortality in dogs with sepsis or SIRS. Also, the owners’ decisions to euthanize possibly affected our results because euthanasia might have been elected after consideration of factors other than the patient’s prognosis.

Serum concentrations of the S100/calgranulins were associated with the WBC counts on day 1, but their moderate correlation suggests that serum S100/calgranulin concentrations do not merely reflect the WBC count. Together with the differential trends of serum S100/calgranulins after 24 h in dogs with sepsis versus SIRS, this leads us to speculate that the S100/calgranulins reflect both the number and activation status of WBC, primarily neutrophils and monocytes.

We did not identify a correlation between serum S100/calgranulin concentrations and the type II acute-phase protein inducer IL-6¹⁷ in plasma, except for serum S100A8/A9, which was moderately correlated with plasma IL-6 concentrations on day 2. Serum S100/calgranulin concentrations were also not associated with the type I acute-phase protein inducer TNF-α in plasma.¹⁷ This suggests that S100/calgranulin expression, their release into the extracellular space, or both is regulated by other factors or by a combination of factors within the acute-phase response, and that the S100/calgranulins do not classify as type I or II acute-phase proteins.¹⁷

Our study had some limitations. First, our prospective evaluation of the inflammatory proteins S100A12 and S100A8/A9 in dogs with documented sepsis or SIRS limited the number of cases enrolled in the study. Second, we did not classify the degree of illness beyond determination of whether patients were strictly SIRS or also had a documentable infection. Systemic hypotension, organ dysfunction, or pressor dependency were not reported, and future studies would need to index S100/calgranulins and these furthering degrees of sepsis and SIRS. Third, determination of SIRS induced by sepsis versus noninfectious SIRS status was performed as accurately as possible with the same clinician involved in the case during the 3-d study period, but some dogs were treated with antibiotics before cultures (e.g., blood cultures) were obtained. Treatments were also not standardized, and the cases were not all treated by the same clinician. Further, the owners’ decision on how to proceed based on the information provided by the clinician may or may not have been affected by a subjective analysis of the case. Also, an effect of treatment on the results cannot be entirely excluded. Finally, serum S100/calgranulin levels were evaluated over a 3-d period starting on the day of hospital ICU admission, assuming that the day of enrollment in the study equates to the same day of the disease course or infection for each dog in the study.