Systemic Immune Profile Predicts the Development of Infections in Patients with Spinal Cord Injuries

Abstract

Patients with spinal cord injury (SCI) frequently develop infections that may affect quality of life, be life-threatening, and impair their neurological recovery in the acute and subacute injury phases. Therefore, identifying patients with SCI at risk for developing infections in this stage is of utmost importance. We determined the systemic levels of immune cell populations, cytokines, chemokines, and growth factors in 81 patients with traumatic SCI at 4 weeks after injury and compared them with those of 26 age-matched healthy control subjects. Patients who developed infections between 4 and 16 weeks after injury exhibited higher numbers of neutrophils and eosinophils, as well as lower numbers of lymphocytes and eotaxin-1 (CCL11) levels. Accordingly, lasso logistic regression showed that incomplete lesions (American Spinal Injury Association Impairment Scale [AIS] C and D grades), the levels of eotaxin-1, and the number of lymphocytes, basophils, and monocytes are predictive of lower odds for infections. On the other hand, the number of neutrophils and eosinophils as well as, in a lesser extent, the levels of IP-10 (CXCL10), MCP-1 (CCL2), BDNF [brain-derived neurotrophic factor], and vascular endothelial growth factor [VEGF]-A, are predictors of increased susceptibility for developing infections. Overall, our results point to systemic immune disbalance after SCI as predictors of infection in a period when infections may greatly interfere with neurological and functional recovery and suggest new pathways and players to further explore novel therapeutic strategies.

Introduction

Patients with spinal cord injury (SCI) face life-long recurrence of infections, mainly of urinary and respiratory tracts, that have a great impact on their quality of life. In addition to being the leading cause of morbidity and mortality after SCI,^1
–3 infections during the acute phase are a predictive factor of impaired neurological recovery and reduced functional independence.^4,5 The high rate of infections is directly linked to the impaired adaptive and innate immune responses that results from SCI.^{6

–16} Among the reasons why this occurs, autonomic nervous system dysfunction (and subsequent increase of ß2-adrenergic signaling in immune cells) as well as activation of the hypothalamic–pituitary–adrenal axis (and subsequent increase of cortisol levels) are the major contributors.^17

–20 However, in contrast to the increasing knowledge about the mechanisms mediating SCI-induced immune depression, little is known about whether the individual immune profile could be used to predict infections.

Recently, Jogia and colleagues²¹ reported that the neutrophil to lymphocyte ratio (NLR) is higher at 3 days after injury among patients who suffer from respiratory infections afterwards. Further, the NLR serves as a predictive marker of infection occurrence.²¹ However, to our knowledge, there is a lack of a more detailed immune profile including other markers of immune status, such as cytokines and other inflammation-related molecules, that would predict infections.

Our aim was to specifically address whether the immune profile of patients with SCI during the subacute period could be used to predict infections. To achieve this, we studied the peripheral blood immune cell populations as well as the levels of cytokines, chemokines, and growth factors from a blood sample that was collected 4 weeks after injury. Further, we explored the association of these markers with occurrence of infections in the following 12 weeks.

Methods

Patients and healthy control subjects

Eighty-one patients with traumatic SCI were recruited at the Trauma Center Murnau (Bavaria, Germany) or at the Hospital Nacional de Paraplejicos (Toledo, Spain) under the development of the Autoantibodies in Spinal Cord Injury study.²² The study protocol and the informed consent sheet of the Autoantibodies in Spinal Cord Injury study were evaluated and approved by the Ethics Committee of Toledo Health Care Area and by the Ethics Committee of the Bavarian Medical Board (registry number 15046). The study follows and adheres to the World Medical Association Declaration of Helsinki and is registered at the public database Clinicaltrials.gov (NCT02493543).

All patients fulfilled the inclusion and exclusion criteria and gave their informed consent to participate.

Inclusion criteria were: males and females; at least 18 years old; any neurological level of injury, except cauda equina syndrome; complete and incomplete lesions; and if the patient was treated with glucocorticoids, the last dose should have been administered at least 7 days before study onset.

Exclusion criteria were: a diagnosed autoimmune disorder; a diagnosed tumor; and preexisting neurodegenerative disease.

In addition, patients with traumatic brain injury (Glascow Coma Scale score <14) were excluded from this study. Polytrauma patients were recruited as long as their injuries did not interfere with neurological examination.

Sensorimotor function of patients was evaluated following the International Standards for Neurological Classification of Spinal Cord Injury scale (ISNCSCI) at an average time of 31 ± 1 days after injury (from now on described as 4 weeks after injury). All evaluations were performed by experienced personnel who attended the ISNCSCI training course at Heidelberg University Hospital. On the same dates of sensorimotor function evaluation, a blood sample was obtained from each patient. Age, gender, lesion level, and severity of patients are summarized in Table 1.

Table 1.

Demographic and Clinical Characteristics of Participants

	SCI	CTL
Age^a,b
Years mean ± SEM (range)	42.1 ± 3.4 (19-79)	40.8 ± 4.3 (22-61)
≤35: n (%)	33 (41)	9 (34.6)
36-55: n (%)	28 (34.5)	14 (54)
≥56: n (%)	20 (24.7)	3 (11.5)
Gender^c
Female: n (%)	9 (11.1)	8 (30.8)
Male: n (%)	72 (88.9)	18 (69.2)
AIS grade
AIS A: n (%)	47 (58)
AIS B: n (%)	8 (9.9)
AIS C: n (%)	12 (14.8)
AIS D: n (%)	14 (17.3)
Level of lesion
C1-C4: n (%)	25 (30.9)
C5-T1: n (%)	22 (27.2)
T2-T10: n (%)	24 (29.6)
T11-L2: n (%)	10 (12.3)
Below L2: n (%)	0 (0)

t-test mean values SCI versus CTL p-value = 0.65.

Frequencies of age ranges, Fisher's exact test p-value = 0.18.

Frequencies of genders, Fisher's exact test p-value = 0.028.

AIS, American Spinal Injury Association Impairment Scale; CTL, control subjects; SCI, spinal cord injury; SEM, standard error of the mean.

Twenty-six healthy volunteers were recruited at the Hospital Nacional de Paraplejicos. A single blood sample was taken after the signing of the informed consent and fulfilling the inclusion and exclusion criteria detailed above was confirmed (with the obvious exception of not suffering an SCI). Age and gender of these individuals are summarized in Table 1.

Blood sera

Peripheral blood was collected by venipuncture. A blood clot was allowed to form by maintaining the tubes for 45 min at room temperature (RT) followed by 1 h at 4°C. Blood was centrifuged at 4°C, at 1,500g for 20 min and serum was aliquoted and stored at −80°C until used.

Whenever a blood sample was taken, a standard hematological analysis was performed in parallel by the clinical laboratories at Murnau Trauma Center (Bavaria, Germany) or at the Hospital Nacional de Paraplejicos (Toledo, Spain).

Documentation of infections

Pneumonias, upper respiratory tract infections (URIs), and urinary tract infections (UTIs) occurring between the day after the first examination date (4 weeks after injury) up to the second examination date (16 weeks after injury) were annotated in the medical records. The most frequent infections were UTIs (54 patients), followed by URIs (9 patients), and pneumonias (6 patients).

Patients were in the hospital during the duration of the study; all infections were acquired during the inpatient period. Distribution of dates of first infection after 4 weeks from injury are shown in Figure 1.

FIG. 1.

Dates of infections after SCI. (A) Distribution of first episode of infection along the period of study (4 to 16 weeks after injury). (B) Infections before 4 weeks. Black dots: patients with infections between 4 and 16 weeks. Gray dots: patients without infections between 4 and 16 weeks. 4 w, 4 weeks after injury; SCI, spinal cord injury.

Pneumonias were diagnosed by clinical examination and confirmed by chest X-ray. UTIs were diagnosed by clinical examination and confirmed by laboratory findings in urine (bacteriuria and/or high number of leukocytes). URIs were diagnosed by clinical examination.

The 55 patients from Murnau Trauma Center (Germany) were admitted to the hospital in the acute phase after injury, whereas the 21 patients from the National Hospital for Paraplegics (Spain) received acute medical management at other medical centers. Thus, infections before inclusion in the study (4 weeks) were available only for patients from the Murnau Trauma Center. Pneumonias, URIs, and UTIs before inclusion were defined as stated above and extracted from medical records.

Determination of serum inflammation-related factor levels

Serum cytokines, chemokines, and growth factors related with distinct types of immune responses were measured with a custom human 16-plex panel (ProcartaPlex Immunoassay, Thermo Fisher Scientific; interleukin [IL]-1β, IL-6, IL-10, IL-17A, IL23, IL-37, tumor necrosis factor [TNF]-α, IL-8 [CXCL8], IP-10 [CXCL10], eotaxin-1 [CCL11], MCP-1 [CCL2], GRO-α [CXCL1], PD-L1, vascular endothelial growth factor [VEGF]-A, brain-derived neurotrophic factor [BDNF], tPA) and a transforming growth factor (TGF)-β1 single-plex panel (ProcartaPlex Immunoassay, Thermo Fisher Scientific), according to the manufacturer's protocol on a Luminex MAGPIX instrument. The 7-point standard curves were created using the Procarta Plex Analyst software. The lower limit of quantitation (LLOQ) was 2.0 pg/mL for IL-1β, 9.2 pg/mL for IL-6, 2.2 pg/mL for IL-10, 2.0 pg/mL for IL-17A, 16.7 pg/mL for IL-23, 4.2 pg/mL for IL-37, 7.2 pg/mL for TNF-α, 2.4 pg/mL for IL-8, 1.8 pg/mL for IP-10, 2.1 pg/mL for eotaxin-1, 3.6 pg/mL for MCP-1, 2.0 pg/mL for GRO-α, 4.0 pg/mL for PD-L1, 5.7 pg/mL for VEGF-A, 2.1 pg/mL for BDNF, 38.3 pg/mL for tPA, and 1.4 pg/mL for TGF- β1.

Among the studied cytokines, IL-1ß, IL-6, IL-8, IL-10, IL-17A, IL-23, IL-37, TNF-α, and PDL-1 were excluded from the analysis due to their absence or low proportion of detection. IL-1ß, IL-10, IL-23, IL17A, and TNF-α levels were below the LLOQ in all samples, including patients and control subjects. IL-37 was over the LLOQ in only 1 patient (1.2%) and IL-6 in 3 patients (3.7%), whereas both cytokines were below the LLOQ in all control subjects. IL-8 was above the LLOQ in 9 patients (11.1%) and in 4 healthy control subjects (15%). PD-L1 was above the LLOQ in 15 patients (18.5%) and in 3 control subjects (11%). Additionally, this protein was above the upper level of quantification (ULOQ) of 233.0 pg/mL in 1 patient.

Statistical analysis

All statistical analyses were performed in R statistical programming language²³ using RStudio.²⁴ For comparisons of three groups by one-way analysis of variance (ANOVA; results shown in Fig. 2), we first checked normality of residuals by Shapiro's test and homoscedasticity by Levene's test. When both conditions were fulfilled, ANOVA and Tukey honest significant difference post hoc test were performed. Otherwise, we performed a Kruskal-Wallis rank sum test followed by Dunn's test. The precise statistical procedure used in each case is detailed in the legend of Figure 2. Levene's test was applied using the car package²⁵ and Dunn's test using the FSA package.²⁶

FIG. 2.

SCI-induced changes in the number of systemic immune cell populations and inflammatory mediators at 4 weeks after injury were more prominent in patients who suffered from infections afterwards. The number of peripheral blood total leukocytes was higher at 4 weeks after injury in patients who suffered from infections in the following 12 weeks compared with control subjects (A). This higher number was paralleled by the significant increase in the number of neutrophils (B) and the decrease in that of lymphocytes (C). Accordingly, the NLR was increased at 4 weeks after injury in both patients with and without infections (D). The number of monocytes was similar in patients and control subjects (E), whereas those of eosinophils (F) and basophils (G) were increased at 4 weeks after injury among the patients with infections in the following 12 weeks. The number of platelets (H) was also significantly increased at 4 weeks after injury in both groups of patients compared with control subjects. Serum levels of the cytokine TGF-ß1 were lower at 4 weeks after injury in both groups of patients compared with control subjects (I). Among chemokines, eotaxin-1 levels were lower among the group of patients with infections compared with control individuals (J), whereas those of MCP-1 (K), GRO-α (L), and IP-10 (M) were similar in patients and control subjects. The levels of the growth factor BDNF were significantly lower in patients compared with control subjects (N), whereas those of VEGFA (O) and tPA (P) were similar between patients and control subjects. Kruskal-Wallis test was applied in all cases with the exception of leukocytes number (A) and TGF-ß level (I), where one-way ANOVA was performed. When Kruskal-Wallis or ANOVA p-values were <0.05, Dunn's or Tukey's post hoc tests, respectively, were conducted. In all panels but (A) and (I): *Dunn's test p < 0.05 versus CTL, ***Dunn's test p < 0.001 versus CTL. In (A) and (I): *Tukey's p < 0.05 versus CTL, ***Tukey's p < 0.001 versus CTL. ANOVA, analysis of variance; BDNF, brain-derived neurotrophic factor; CTL, control healthy subjects; w/o INF, without infections; INF, infections; NLR, neutrophil to lymphocyte ratio; SCI, spinal cord injury; TGF, transforming growth factor; VEGF, vascular endothelial growth factor.

Lasso logistic regression was performed using the glmnet package.²⁷ Shrinkage parameter, lambda, was estimated by 10-fold cross-validation. The model was trained with the number of cells and levels of inflammatory mediators shown in Figure 2, as well as with age, sex, lesion level (paraplegic or tetraplegic), and severity (American Spinal Injury Association Impairment Scale [AIS] grade) of each participant.

To build a receiver operating characteristic (ROC) curve and to calculate the area under the ROC curve (AUC) we employed the library pROC.²⁸ The decision threshold of the lasso regression outcome for classifying patients as with or without infections was determined from the ROC curve according to the maximum Youden's J statistic, useful for imbalanced classification.²⁹ Then, accuracy, sensitivity, specificity, positive predictive value, and negative predictive value of the lasso regression model were estimated. A precision-recall (PR) curve was built employing the library PRROC.³⁰

Results

Changes in immune cell populations and inflammatory mediators at 4 weeks after injury are more prominent in patients who develop infections afterwards

The 81 patients with SCI included in the study were divided into two groups according to whether they suffered from infections in the following 12 weeks (57 patients) or they were infection free within the observation period (24 patients; Table 2). Except for the occurrence of infections, both groups of patients exhibited similar clinical and demographic characteristics at the baseline (Table 2). A first infection episode occurred across the whole period of the study, although skewness toward earlier times was observed: median time of infection was 33 days after baseline (Fig. 1A). Infections before 4 weeks occurred in 6 of 40 patients (13%) who developed them later on and in 3 of 15 patients (20%) who remained free of them (Fig. 1B). Thus, frequency of infections before 4 weeks does not seem to affect frequency afterwards (Fisher's exact test p-value = 0.69).

Table 2.

Relationship between Infections from 4 to 16 Weeks and Demographic and Clinical Variables of Patients with Spinal Cord Injury

	No infections (n = 24)	Infections (n = 57)	P-value^a
Age			0.43
≤35: n (%)	12 (50)	21 (36.8)
36-55: n (%)	6 (25)	22 (38.6)
≥56: n (%)	6 (25)	14 (24.6)
Gender			0.26
Female: n (%)	1 (4.2)	8 (14)
Male: n (%)	23 (95.8)	49 (86)
AIS grade			0.14
AIS A: n (%)	11 (45.8)	36 (63.2)
AIS B: n (%)	1 (4.2)	7 (12.3)
AIS C: n (%)	6 (25)	6 (10.5)
AIS D: n (%)	6 (25)	8 (14)
Level of lesion			0.24
C1-C4: n (%)	8 (33.3)	17 (29.8)
C5-T1: n (%)	7 (29.2)	15 (26.3)
T2-T10: n (%)	4 (16.7)	20 (35.1)
T11-L2: n (%)	5 (20.8)	5 (8.8)

Fisher's exact test p-value.

AIS, American Spinal Injury Association Impairment Scale.

We compared the immune cell populations and inflammatory mediators of both groups of patients in the subacute phase (4 weeks after injury) with those of healthy subjects as the reference. As shown in Figure 2, changes in immune cell populations and inflammatory mediators were more prominent among patients who suffered from infections in the follow-up period, whereas patients without infections had intermediate levels between those of controls and patients with infections. Indeed, the numbers of total leukocytes, neutrophils, and eosinophils were significantly higher, whereas those of lymphocytes were significantly lower only in patients suffering from infections compared with control subjects (Fig. 2). Neutrophil to lymphocyte ratio and platelet numbers were also higher in patients with infections than in those without infections, and the values of both groups of patients were significantly higher than those of control subjects (Fig. 2). Among the inflammatory mediators, the levels of the chemokine eotaxin-1 (CCL11) were significantly lower only among patients who afterwards suffered from infections (Fig. 2J), whereas the levels of TGF-ß1 and BDNF were significantly lower in both groups of patients compared with control subjects (Fig. 2D,H). No statistically significant differences were reached between patients with infections and patients without them.

The systemic immune profile at 4 weeks after injury may predict the occurrence of infections

Lasso logistic regression was applied to automatically select the cell populations, inflammatory mediators, and clinical and demographic variables that best predict occurrence of infections. As shown in Table 3, the final model includes motor incomplete lesions (AIS C or D grades) as factors associated with lower odds of developing infections. Among the inflammatory mediators, eotaxin-1 levels are associated with lower odds of infections, whereas IP-10 and, to a lesser extent, MCP1, BDNF, and VEGF-A levels are associated with higher odds. Regarding the cell populations, the numbers of neutrophils and eosinophils are associated with higher odds of developing infections, whereas the opposite is observed for the numbers of lymphocytes, basophils, and monocytes. The decision threshold to discriminate whether the output of the model for a patient corresponds to occurrence of infections was estimated by the Youden's J statistic from the ROC curve. As shown in Figure 3A, the AUC-ROC was 0.83. Applying the optimum threshold value results in a model accuracy of 0.80 (p-value = 0.046), with a sensitivity (recall) of 0.68, a positive predictive value (precision) of 0.65, a specificity of 0.85, and a negative predictive value of 0.87. To further assess confidence on prediction of infections, we calculated the F1-score, 0.67, and the area under the PR curve, 0.93 (Fig. 3B).

FIG. 3.

ROC and precision-recall curves for the lasso regression model. AUC, area under the curve; FPR, False positive rate; ROC, receiver operating characteristic.

Table 3.

Variables Automatically Selected by Lasso Logistic Regression as Significant Predictors of Infections

	Odds
Intercept	2.72
AIS C (vs. AIS A)	0.42
AIS D (vs. AIS A)	0.96
Eotaxin-1 (pg/mL)	0.95
IP-10 (pg/mL)	1.03
MCP1 (pg/mL)	1.007
BDNF (pg/mL)	1.0008
VEGF-alpha (pg/mL)	1.0003
Neutrophils number (10⁶ cells/mL)	1.08
Eosinophils number (10⁶ cells/mL)	3.3
Basophils number (10⁶ cells/mL)	0.004
Lymphocytes number (10⁶ cells/mL	0.78
Monocytes number (10⁶ cells/mL)	0.94

AIS, American Spinal Injury Association Impairment Scale; BDNF, brain-derived neurotrophic factor; IP-10, Interferon-γ-Inducible Protein 10; MCP1, Monocyte chemoattractant protein-1; VEGF, vascular endothelial growth factor.

Discussion

Our results suggest that the alterations in the systemic immune profile are more pronounced in patients who will suffer from infections and may be useful to predict the occurrence of infections in the subacute phase (4 weeks after injury and onwards), when most patients are hospitalized to undergo their physical rehabilitation program.

Previous studies have shown that patients experience transient neutrophilia and lymphopenia during the first week after injury.^11,21 Our results expand the period at which these alterations are still observable in the subgroup of patients who will suffer from infections. Indeed, although only a proportion of patients reached at 4 weeks after injury the clinical thresholds established for lymphopenia (≤10⁶ cells/mL) or neutrophilia (≥8 × 10⁶ cells/mL), the number of lymphocytes was significantly lower among the group of patients who suffered from infections compared with control subjects, whereas the numbers of neutrophils was significantly higher (Fig. 2). This disbalance toward neutrophils is reflected by the increase in the NLR (Fig. 2), which has been previously reported to be higher at earlier time-points (in the first week after injury) among patients who develop infections afterwards.²¹ From a mechanistical perspective, NLR is considered an indicator of physical stress because it is increased by the rise in cortisol and catecholamines levels.³¹ Thus, the association between more neutrophils and less lymphocytes with higher susceptibility to infections might reflect larger autonomic nervous system disbalance and/or more potent activation of the hypothalamic–pituitary–adrenal axis, the proposed two main causes of immunodepression after SCI.^17

–20

In this regard, the potential effect of pharmacological treatments with glucocorticoids should be also explored. In our cohort there were only two patients treated with glucocorticoids: one from days 24 to 13 before inclusion in the study and the other one from days 79 to 90 after the inclusion. Nevertheless, these patients did not exhibit a higher number of infections or present abnormal numbers of peripheral blood cell populations or levels of inflammatory mediators. Further studies are required to answer whether exogenous glucocorticoids may interfere with susceptibility to infections.

Noticeably, the number of eosinophils was also higher among patients who developed infections, whereas (as occurred with neutrophils and lymphocytes) patients free of infections showed intermediate levels between those of patients suffering infections and those of healthy controls (Fig. 2). Focusing on inflammatory mediators, the same tendency was also detected for eotaxin-1 (Fig. 2). Levels of this chemokine were significantly lower in patients who suffered from infections, whereas those without infections showed intermediate levels between the former and control subjects.

Eotaxin-1 is involved in the release of eosinophils from the bone marrow to blood and their homing into tissues.³² Indeed, intravenous injection of eotaxin-1 rapidly produces eosinophilia.³³ However, our results show that although eotaxin-1 levels are significantly lower in patients with infections, the number of eosinophils goes in the opposite direction, suggesting that the increase in eosinophils is independent from eotaxin-1-induced release of these cells from the bone marrow. A potential explanation might be found in the upregulation of cytokines associated classically to be produced under allergic eosinophilic responses, such as eosinophilic esophagitis, in which IL-5 and IL-13 produced by pathogenic effector Th2 cells, innate lymphoid cells, or invariant natural killer T cells act as potent chemoattractants for eosinophils.^34
–36

We did determine the levels of Th2-related cytokines (except IL-10, whose levels were below the limit of detection in all samples), but catecholamines and glucocorticoids, elevated after injury and, as previously discussed, main effectors of immunodepression after SCI,^17

–20 are also well-known inducers of skewness toward Th2 activity.³⁷ To note, Th2 responses have been classically considered as downregulators of Th1-mediated cellular immune responses, essential to fighting against intracellular pathogens and other pathogens. Further work is warranted to study whether an imbalance on immune activity toward a Th2-like profile underlies eotaxin-1-independent increase of eosinophil numbers and whether this activity may enhance the susceptibility to infections in patients with SCI.

We employed lasso logistic regression to automatically select the variables that best predict the occurrence of infections. The final model (Table 3) shows that the numbers of neutrophils and eosinophils are related with higher odds of occurrence of infections, whereas the number of lymphocytes has the inverse relationship (more lymphocytes, lower odds of infections; Table 3). Noticeably, among these cell populations, the one that most affects the odds of infections is the number of eosinophils: odds of infections increase by 3.3 per each increment in 10⁶ eosinophils/mL. Neutrophils present a weaker increment in the odds of infections (1.08 per 10⁶ cells/mL), but it should be taken into consideration that the number of neutrophils among the patients in our study ranged from 5 × 10⁵ to 9.7 × 10⁶ cells/mL, so the final contribution of neutrophil number to the odds of infection may be much more substantial. Indeed, between the two patients with the extreme neutrophil number values, keeping all the other variables constant, the odds of infection would be doubled in the patient with higher neutrophil numbers (1.08^9.7-0.5 = 2.03).

Something similar occurs with odds associated to lymphocytes (0.78), whose numbers in our cohort of patients varied from 7 × 10⁵ to 4.1 × 10⁶ cells/mL, so the effect of lymphocyte numbers in odds of infections would be reduced almost 60% between the two patients discussed above (keeping all the other variables constant). Monocyte number is also a variable included in the model, but its contribution was more modest (odds = 0.94) and close to its maximum potential effect because in our cohort the highest monocyte number was 1.1 × 10⁶ cells/mL. On the other hand, odds of infection decreased dramatically (0.004) for each increment on 10⁶ basophils/mL, but it should be noted that this concentration of basophils was not achieved in our cohort (basophils mean value, 2 × 10⁴ cells/mL; maximum value, 2 × 10⁵ cells/mL).

Among inflammation-related molecules, eotaxin-1 has the greatest impact on odds of infections (0.95), followed closely by IP-10 (1.03). But in our cohort, the range of values for eotaxin-1 was wider, from 0 to 84.3 pg/mL, whereas for IP-10 the range was from 0 to 60 pg/mL. MCP-1, BDNF, and VEGF-alpha were also included in the model, but their contribution was much weaker (Table 3).

The association between the immunological profile of patients with SCI and their susceptibility to infections could be further supported by observing that a particular immunological profile is related to the number of infections or their duration. Unfortunately, our cohort was not suitable for exploring these issues because the number of patients with 0, 1, 2, 3, or 4 infections were (respectively) 24, 26, 19, 7, and 5; too few patients had 3 or 4 infections in order to conduct a reliable statistical analysis. Regarding the duration of infections, our cohort was quite homogenous and long-term infections did not occur. Indeed, the 95% confidence interval of infection duration was 8.1 ± 0.56 days, with an interquartile range from 6 to 10 days, and a maximum length of infection of 16 days.

The lasso logistic regression model also incorporates AIS C or AIS D lesions as factors that decrease the odds of infections. A potential explanation, in line with the discussion above, may be that less severe spinal cord lesions have a less severe effect on the immune system. In this regard, the association of infection occurrence with neurological and functional recovery is statistical, not causal: it still remains uncertain as to whether it is infection (or the immunodepression that underlies higher infection susceptibility) that impairs neurological recovery, or rather that lesion severity determines recovery, immunodepression, and susceptibility to infections.

A limitation of our study is the absence of a validation cohort to test the applicability of the model to predict infection occurrence in other data sets. But overall, our results show that the alteration of the systemic immune profile at 4 weeks after SCI is more pronounced in those patients who afterwards suffer from infections and it may serve to predict the occurrence of infections. Further work is needed to precisely identify the mechanisms underlying these differences and explore whether therapeutic strategies could be developed to harness the immune system in order to prevent infections.

Funding Information

This work was funded by the Ministry of Science and Innovation of Spain (MCIN)/Agencia Estatal de Investigacion (AEI)/DOI 10.13039/501100011033, grant ID PID2020-120652RB-I00 (to DGO and AAM).

Footnotes

Author Disclosure Statement

No competing financial interests exist.

References

DeVivo

M.J.

, Kartus

P.L.

, and Stover ,S.L., et al. (1989). Cause of death for patients with spinal cord injuries. Arch. Intern. Med. 149, 1761–1766.

Soden

R.J.

, Walsh

, and Middleton

J.W.

, et al. (2000). Causes of death after spinal cord injury. Spinal Cord, 38, 604–610.

Thietje

, Pouw

M.H.

, and Schulz

A.P.

, et al. (2011). Mortality in patients with traumatic spinal cord injury: descriptive analysis of 62 deceased subjects. J. Spinal Cord Med. 34, 482–487.

Failli

, Kopp

M.A.

, and Gericke

, et al. (2012). Functional neurological recovery after spinal cord injury is impaired in patients with infections. Brain, 135, 3238–3250.

Kopp

, Watzlawick

, and Martus

, et al. (2017). Long-term functional outcome in patients with acquired infections after acute spinal cord injury. Neurology, 88, 892–900.

Cruse

J.M.

, Lewis

R.E.

, and Bishop

G.R.

, et al. (1992). Neuroendocrine-immune interactions associated with loss and restoration of immune system function in spinal cord injury and stroke patients. Immunol. Res. 61, 183–192.

Cruse

J.M.

, Lewis

R.E.

, Bishop, and

G.R.

, et al. (1993). Decreased immune reactivity and neuroendocrine alterations related to chronic stress in spinal cord injury and stroke patients. Pathobiology, 61, 183–192.

Kliesch

W.F.

, Cruse

J.M.

, and Lewis

R.E.

, et al. (1996). Restoration of depressed immune function in spinal cord injury patients receiving rehabilitation therapy. Paraplegia, 34, 82–90.

Campagnolo

D.I.

, Bartlett

J.A.

, and Keller

S.E.

, et al. (1997). Impaired phagocytosis of Staphylococcus aureus in complete tetraplegics. Am. J. Phys. Med. Rehabil. 76, 276–280.

10.

Iversen

P.O.

, Hjeltnes

, and Holm

, et al. (2000). Depressed immunity and impaired proliferation of hematopoietic progenitor cells in patients with complete spinal cord injury. Blood, 96, 2081–2084.

11.

Riegger

, Conrad

, and Schluesener

H.J.

, et al. (2009). Immune depression syndrome following human spinal cord injury (SCI): a pilot study. Neuroscience, 158, 1194–1199.

12.

Oropallo

M.A.

, Goenka

, and Cancro

M.P.

(2014). Spinal cord injury impacts B cell production, homeostasis, and activation. Semin. Immunol. 26, 421–427.

13.

Oropallo

M.A.

, Held

K.S.

, and Goenka

, et al. (2012). Chronic spinal cord injury impairs primary antibody responses but spares existing humoral immunity in mice. J. Immunol. 188, 5257–5266.

14.

Held

K.S.

, Steward

, and Blanc

, et al. (2010). Impaired immune responses following spinal cord injury lead to reduced ability to control viral infection. Exp. Neurol. 226, 242–253.

15.

Held

K.S.

, and Lane

T.E.

(2014). Spinal cord injury, immunodepression, and antigenic challenge. Semin. Immunol. 26, 415–420.

16.

Ibarra

, Jiménez

, and Cortes

, et al. (2007). Influence of the intensity, level and phase of spinal cord injury on the proliferation of T cells and T-cell-dependent antibody reactions in rats. Spinal Cord, 45, 380–386.

17.

Lucin

K.M.

, Sanders

V.M.

, and Popovich

P.G.

(2009). Stress hormones collaborate to induce lymphocyte apoptosis after high level spinal cord injury. J. Neurochem. 110, 1409–1421.

18.

Zhang

, Guan

, and Reader

, et al. (2013). Autonomic dysreflexia causes chronic immune suppression after spinal cord injury. J. Neurosci. 33, 12970–12981.

19.

Ueno

, Ueno-Nakamura

, and Niehaus

, et al. (2016). Silencing spinal interneurons inhibits immune suppressive autonomic reflexes caused by spinal cord injury. Nat. Neurosci. 19, 784–787.

20.

Lucin

, Sanders

, and Jones

, et al. (2007). Impaired antibody synthesis after spinal cord injury is level dependent and is due to sympathetic nervous system dysregulation. Exp. Neurol. 207, 75–84.

21.

Jogia

, Lübstorf

, and Jacobson

, et al. (2021). Prognostic value of early leukocyte fluctuations for recovery from traumatic spinal cord injury. Clin. Transl. Med. 11, e272.

22.

Arevalo-Martin

, Grassner

, and and Garcia-Ovejero

, et al. (2018). Elevated autoantibodies in subacute human spinal cord injury are naturally occurring antibodies. Front. Immunol. 9, 2365.

23.

R Core

Team

. (2018). R: a language and enviroment for statistical computing. R Found. Stat. Comput. Vienna

Austria

. https//www.R-project.org/ (Last accessed September 18, 2019 ).

24.

Rstudio Team. (2019). RStudio: integrated development for R. RStudio, Inc.: Boston, MA.

25.

Fox

, and Weisberg

(2019). An R Companion to Applied Regression, 3rd ed. Sage Publications: Los Angeles, CA.

26.

Ogle

D.H.

, Wheeler

, and Dinno

(2022). FSA: Fisheries stock analysis. R package version 0.9.0. https://github.com/fishR-Core-Team/FSA (Last accessed June 10, 2021 ).

27.

Friedman

, Hastie

, and Tibshirani

(2010). Regularization paths for generalized linear models via coordinate descent. J. Stat. Soft. 33, 1–22.

28.

Robin

, Turck

, and Hainard

, et al. (2011). pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics, 12, 77.

29.

Youden

(1950). Index for rating diagnostic tests. Cancer, 3, 32–35.

30.

Grau

, Grosse

, and Keilwagen

(2015). PRROC: computing and visualizing precision-recall and receiver operating characteristic curves in R. Bioinformatics, 31, 2595–2597.

31.

Zahorec

(2001). Ratio of neutrophil to lymphocyte counts—rapid and simple parameter of systemic inflammation and stress in critically ill. Bratisl. Lek. Listy, 102, 5–14.

32.

Mould

A.W.

, Matthaei

K.I.

, and Young

I.G.

, et al. (1997). Relationship between interleukin-5 and eotaxin in regulating blood and tissue eosinophilia in mice. J. Clin. Invest. 99, 1064–1071.

33.

Palframan

, Collins

, and Williams

, et al. (1998). Eotaxin induces a rapid release of eosinophils and their progenitors from the bone marrow. Blood, 91, 2240–2248.

34.

Mitson-Salazar

, Yin

, and Wansley

, et al. (2016). Hematopoietic prostaglandin D synthase defines a proeosinophilic pathogenic effector human T(H)2 cell subpopulation with enhanced function. J. Allergy Clin. Immunol. 137, 907–918.

35.

Doherty

, Baum

, and Newbury

, et al. (2015). Group 2 innate lymphocytes (ILC2) are enriched in active eosinophilic esophagitis. J. Allergy Clin. Immunol. 136, 792–794.

36.

Rayapudi

, Rajavelu

, and Zhu

, et al. (2014). Invariant natural killer T-cell neutralization is a possible novel therapy for human eosinophilic esophagitis. Clin. Transl. Immunol. 3, e9.

37.

Spellberg

, and Edwards

(2001). Type 1/Type 2 immunity in infectious diseases. Clin. Infect. Dis. 32, 76–102.