Goracke‑Postle et al. BMC Neurol (2021) 21:384 https://doi.org/10.1186/s12883‑021‑02333‑2

R E S E A R C H A R T I C L E

Expression of and correlational patterns among neuroinflammatory, neuropeptide, and neuroendocrine molecules from cerebrospinal fluid in cerebral palsyCory J. Goracke‑Postle1, Chantel C. Burkitt2, Angela Panoskaltsis‑Mortari3, Michael Ehrhardt3, George L. Wilcox4, Patrick Graupman2, Michael Partington5 and Frank J. Symons6*

Abstract Background: The underlying pathogenesis of cerebral palsy (CP) remains poorly understood. The possibility of an early inflammatory response after acute insult is of increasing interest. Patterns of inflammatory and related biomark‑ers are emerging as potential early diagnostic markers for understanding the etiologic diversity of CP. Their presence has been investigated in plasma and umbilical cord blood but not in cerebrospinal fluid (CSF).

Methods: A clinical CP sample was recruited using a single‑time point cross‑sectional design to collect CSF at point‑of‑care during a standard‑of‑care surgical procedure (intrathecal pump implant). Patient demographic and clinical characteristics were sourced from medical chart audit.

Results: Significant (p ≤ 0.001) associations were found among neuroinflammatory, neuroendocrine, and nocicep‑tive analytes with association patterns varying by birth status (term, preterm, extremely preterm). When between birth‑group correlations were compared directly, there was a significant difference between preterm and extremely preterm birth subgroups for the correlation between tumour necrosis factor alpha (TNFα) and substance P.

Conclusion: This investigation shows that CSF can be used to study proteins in CP patients. Differences in inter‑correlational patterns among analytes varying by birth status underscores the importance of considering birth status in relation to possible mechanistic differences as indicated by biomarker signatures. Future work should be oriented toward prognostic and predictive validity to continue to parse the heterogeneity of CP’s presentation, pathophysiol‑ogy, and response to treatment.

Keywords: Cerebral palsy, Neuroinflammation, Neuroendocrine, Nociceptive, Biomarkers, CSF, Children

© The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

BackgroundCerebral palsy (CP) is a neurodevelopmental disorder of movement, muscle tone or posture caused by an insult to the developing brain before birth, at birth, or before the

age of two years. CP is the most common motor disabil-ity in childhood, affecting an estimated 500,000 Ameri-can children and occurring in approximately 2 children per 1,000 live births. The pathogenesis of CP is not fully understood. CP risk factors predominantly involve peri-natal factors such as anoxia and ischemia and prenatal factors such as young gestational age, intrauterine viral infections, and maternal thyroid abnormalities. Intrau-terine infection and inflammation are of particular

Open Access

*Correspondence: [email protected] Department of Educational Psychology, College of Education and Human Development, Minneapolis, MN 55455, USAFull list of author information is available at the end of the article

Page 2 of 13Goracke‑Postle et al. BMC Neurol (2021) 21:384

interest with both maternal response (chorioamnionitis) and fetal response (funicitis or elevated interleukin-6 in fetal plasma) being associated with white matter dam-age (WMD) and CP [1]. Injury and related inflammatory processes may persist for considerable periods of time (years) leading to hypotheses and emerging models of tertiary-like mechanism of damage including epigenetic regulation and inflammatory relevant changes [2].

Work investigating infection- and neuroinflammatory-related biomarkers related to WMD in neonates has generated evidence mostly supporting a pathway from intrauterine infection to placental inflammation then to systemic fetal circulation and the preterm newborn brain [3]. The specifics of the infection-inflammation-brain damage link have been described and documented and are an active area of investigation [4] particularly given cautions about what level of inference is or is not sup-ported by high quality evidence [5]. What is less clear, however, is the robustness of the inference supporting an infection-inflammation-WMD link with CP as an inevi-table outcome. There is some evidence for such a link [6], but it may be conditioned on whether the infant was born term or preterm. From a more general perspective, the emerging viewpoint on the role of neuro-inflammation as a pathophysiological contributor to CP has created the possibility of a new therapeutic window through which to view the condition [7, 8].

Specific findings from prior clinical work specifi-cally investigating inflammatory mediators have dem-onstrated differential biomarker patterns in umbilical cord serum from infants stratified by preterm and CP status [1]. Of the potential inflammation markers that differed between cases and controls, the markers were lower (based on medians) in the preterm CP cases rela-tive to controls but were higher, relative to controls, in the term CP cases. Sampling from a different compart-ment (plasma) and using a different approach, a study by Lin et al. (2010) comparing school-age children who were former preterm births reported higher cytokine responses (increased tumor necrosis factor-alpha [TNF-α] plasma levels and greater mRNA levels of toll-like receptor four [TLR4]) among those children with CP who were preterm relative to controls (children born preterm with normal development) [9]. Another study of blood, using a serial approach (i.e., repeated measures) in over 900 preterm infants, documented elevated concentration values of myriad inflammatory-relevant mediators which were associated with different risk profiles depending on the sampling day [10]. Using CSF from preterm infants with brain injury, Douglas-Escobar and Weiss (2012) documented combinations of biomarker concentration values that could be used to

inform clinical decision making [4]. Using blood from a sample of children with and without CP, Zareen (2020) [11] found significantly increased levels of erythropoie-tin at baseline in children with CP compared with chil-dren in the comparison group. In response to challenge (lipopolysaccharide), both groups had appropriate and comparable response profiles for interleukin-8, vascular endothelial growth factor (VEGF), TNF-α, and granu-locyte–macrophage colony-stimulating factor (GM-CSF) levels. The children with CP showed a statistically significant lipopolysaccharide hypo-responsiveness profile for interleukin-1a, interleukin-1b, interleukin-2, and interleukin-6 levels. Collectively, the work to date consistently shows immune and inflammatory differ-ences in children with CP.

There remain gaps in our knowledge about the spe-cific linkages among various immune and related mechanisms driving hypothesized persistent inflamma-tory states in children with CP, and the relation among the various biomarkers, risk factors, and specific out-comes. The overall generality of the findings to date for the CP population is limited by two kinds of problems, namely the relative difficulties in establishing valid pre-clinical models for this purpose [12] and the extreme paucity of clinically-relevant biomarker research within this high-need vulnerable patient group. For example, the relation among the CSF biomarkers investigated by Douglas-Escobar and Weiss (2012) and CP as an outcome was not clear [4]. More biomarker data of a comparable kind from the same and different compart-ments across the same and different age groups are needed. To the best of our knowledge, no comparable work has investigated inflammatory-relevant molecu-lar biomarkers in CSF from children, adolescents, and young adults with CP.

The purpose of this preliminary investigation was exploratory. The design was cross-sectional using a single time point for specimen collection from a clini-cal sample. There were two specific aims. The first aim was to document levels of inflammatory and related molecules in CSF in a sample primarily of school-age children with CP. To do so, for detectable analytes, participants were arrayed along each analyte’s con-centration gradient. The second aim was to examine clinically relevant grouping variables (e.g., CP sever-ity, term/preterm birth) to identify any potentially rel-evant correlational patterns among the molecules. Our intent was to extend the work initiated by Kaukola and Lin (described above) using a clinical sample in which standard-of-care surgical interventions were leveraged to gain access to CSF for future hypothesis-generating research purposes.

Page 3 of 13Goracke‑Postle et al. BMC Neurol (2021) 21:384

MethodsProtocol approvalThis study was approved by the Institutional Review Board (IRB, #0809M46301) of the University of Minne-sota. Written informed consent was obtained from each participant or legal representative (i.e., parent/guardian).

ParticipantsThis study utilized a single-time point cross-sectional design. Twenty-eight individuals (82% male) with CP participated (mean age = 9.74  years, SD = 4.36; range = 4–23). Specific CP diagnoses included: quad-riplegia (n = 17), diplegia (n = 5), and triplegia (n = 5). Participants were included in the study if they (a) had cerebral palsy, (b) were between 3–25  years of age, and (c) were scheduled for initial intrathecal baclofen (ITB) pump implant. Individuals were excluded if (a) they had an existing cerebral shunt; or (b) they had compounded dosing (i.e., opioid adjunctive to baclofen) through their pump. The participants were already characterized clini-cally using the Gross Motor Functional Classification System for Cerebral Palsy (GMFCS) to categorize gross motor function. The GMFCS is a 5-level classification system based on self-initiated movement with emphasis on truncal control and walking. The GMFCS is widely used clinically and in classifying individuals with CP for research studies [13].

For the subgroup comparisons, the breakdown of par-ticipant demographics was as follows: males (n = 23) and females (n = 5); non-quadriplegia (n = 11) and quad-riplegia (n = 17); Caucasian (n = 26) and other (n = 2); spastic CP (n = 13) and mixed tone CP (n = 12); term birth, defined as 37 weeks or later (n = 6), preterm birth, defined as 28–37  weeks (n = 12) and extremely preterm birth, defined as less than 28 weeks (n = 9); seizure (n = 6) and no seizure (n = 16) (Table 1).

CSF collectionPatients were consented in accordance with an approved IRB protocol. If consent was given, CSF was collected during a standard-of-care surgical procedure (ITB pump implant). In all cases, the surgery proceeded as usual until the spinal catheter had been placed. Then, the neurosur-geon collected 10–20 ml of CSF from the spinal catheter placed well above the spinal puncture site. This method avoided contaminating the collected CSF with blood.

Immediately following collection, the CSF was placed on wet ice (+ 4  °C) and transported to a cold room for processing, centrifuged at 3000  rpm × 5  min, pipetted into 100 and 250 µL aliquots, flash frozen in liquid nitro-gen and archived at -80  °C. After specimen collection,

the patient was monitored closely following routine operative and post-operative procedures. There were no adverse events.

CSF analyte analysisCSF was analyzed using conventional biochemical meth-ods based on commercially available enzyme-linked immunosorbent assay (ELISA) kits and expression lev-els of each marker were quantified. Specifically, samples were tested by the Cytokine Reference Laboratory (CRL, University of Minnesota). This is a CLIA’88-licensed facility (license #24D0931212). Samples were analyzed for adrenocorticotropic hormone (ACTH), agouti-related peptide (AgRP), brain-derived neurotrophic factor (BDNF), ciliary neurotrophic factor (CNTF), follicle-stimulating hormone (FSH), growth hormone (GH), luteinizing hormone (LH), prolactin (PRL), and thyroid stimulating hormone (TSH) using the “Human Brain-Derived Protein Panel” on the Luminex platform and done as a multi-plex (Luminex instrument—Bioplex 100 [Bio-Rad, 1000 Alfred Nobel Drive, Hercules, CA, 94547], Software: bio-plex Manager 4.0). The polysty-rene bead set (cat. # HPT-66  K-09) with kit lot number 1757143 was used. Kits/reagents were purchased from EMD Millipore Corporation, Billerica, MA. Interferon α2 (IFNα2), interleukin-1α (IL-1α), interleukin-1ra (IL-1ra), interleukin-6 (IL-6), interleukin-8 (IL-8), interleukin-10 (IL-10), interleukin-12p40 (IL-12p40), interleukin-12p70 (IL-12p70), interferon gamma-induced protein 10 (IP-10), monocyte chemotactic protein (MCP-1), mac-rophage inflammatory protein 1β (MIP1β), regulated on activation normal T expressed and secreted (RANTES), and tumor necrosis factorα (TNFα) were analyzed using the “Cytokine/Chemokine Panel 1” on the Luminex plat-form and done as a multi-plex (Luminex instrument—Bioplex 100 [Bio-Rad, 1000 Alfred Nobel Drive, Hercules, CA, 94547], Software: bio-plex Manager 4.0). The poly-styrene bead set (cat. # MPXHCYTO-60  K-14) with kit lot number 1757142 was used. Kits/reagents were pur-chased from EMD Millipore Corporation, Billerica, MA. Dynorphin A, neuropeptide Y, somatostatin, β endor-phin, cortisol, neurotensin, orexin A, substance P, mela-tonin, oxytocin, and melanocyte-stimulating hormone (α-MSH) were analyzed using the “Human Neuropeptide Panel” on the Luminex platform and done as a multi-plex (Luminex instrument—Bioplex 100 Bio-Rad, 1000 Alfred Nobel Drive, Hercules, CA, 94547], Software: bio-plex Manager 4.0). The polystyrene bead set (cat. # HNP-35 K-08) with kit lot number 1823005 was used. Kits/rea-gents were purchased from EMD Millipore Corporation, Billerica, MA.

Samples were assayed according to manufacturer’s instructions. ELISA employ the quantitative sandwich

Page 4 of 13Goracke‑Postle et al. BMC Neurol (2021) 21:384

enzyme immunoassay technique. The absorbance is measured on the microtiter plate reader (Bio-Rad model 550). The intensity of the color formed is proportional to the concentration of the sample. Fluorescent color-coded beads coated with a specific capture antibody were added to each sample. After incubation, and washing, biotinylated detection antibody was added, followed by phycoerythrin-conjugated streptavidin. The beads were read on a Luminex instrument (Bioplex 100) which is a dual-laser fluidics-based instrument. One laser deter-mines the analyte being detected via the color coding; the other measures the magnitude of the PE signal from the

detection antibody which is proportional to the amount of analyte bound to the bead. Samples were tested in duplicate and values were interpolated from 5 parameter-fitted (5PL) standard curves.

Statistical analysesData analysis was exploratory and relied on visual analy-sis, descriptive statistics, and correlational analyses. First, visual analysis of each analyte was conducted to under-stand its distributional form and to identify potential outliers. Further, measures of central tendency (means,

Table 1 Participant health information; M ± SD or n (%)

Note: gestational age was not available for one participant; Term birth = born 37 weeks gestation or later; preterm birth = born at 28–37 weeks gestation; extremely preterm birth = born at less than 28 weeks gestation; CP Cerebral Palsy, GMFCS Gross Motor Function Classification System, level I ambulant without assistance, level II ambulant without assistive devices, limitations outside the home, level III ambulant with assistive devices, wheelchair required outside the home, level IV non‑ambulatory, self‑mobile in wheelchair with limitations, level V non‑ambulatory, self‑mobility very limited

Complete sample(n = 28)

Term birth(n = 6)

Preterm birth(n = 12)

Extremely preterm birth(n = 9)

Male 23 (82.1) 3 (50.0) 11 (91.7) 9 (100)

Ethnicity

Caucasian 26 (92.9) 6 (100.0) 12 (100.0) 8 (88.9)

African American 1 (3.6) 0 (0) 0 (0) 1 (11.1)

Other (not specified) 1 (3.6) 0 (0) 0 (0) 0 (0)

Epilepsy

None 16 (57.1) 3 (50.0) 7 (58.3) 6 (66.7)

History of seizure/diagnosis of epilepsy 6 (21.4) 1 (16.7) 4 (33.3) 1 (11.1)

Questionable seizure activity 3 (10.7) 1 (16.7) 0 (0) 1 (11.1)

Missing 3 (10.7) 1 (16.7) 1 (8.3) 1 (11.1)

CP topography

Hemiplegia 1 (3.6) 0 (0) 1 (8.3) 0 (0)

Diplegia 5 (17.9) 2 (33.3) 3 (25.0) 0 (0)

Triplegia 5 (17.9) 1 (16.7) 0 (0) 4 (44.4)

Quadriplegia 17 (60.7) 3 (50.0) 8 (66.7) 5 (55.6)

GMFCS

Level I 3 (10.7) 0 (0) 3 (25.0) 0 (0)

Level II 3 (10.7) 1 (16.7) 0 (0) 2 (22.2)

Level III 7 (25.0) 3 (50.0) 2 (16.7) 2 (22.2)

Level IV 8 (28.6) 1 (16.7) 4 (33.3) 3 (33.3)

Level V 6 (21.4) 1 (16.7) 2 (16.7) 2 (22.2)

Missing 1 (3.6) 0 (0) 1 (8.3)

Tone

Spastic 13 (46.4) 4 (66.7) 5 (41.7) 4 (44.4)

Mixed tone 12 (42.9) 2 (33.3) 5 (41.7) 4 (44.4)

Missing 3 (10.7) 0 (0) 2 (16.7) 1 (11.1)

Current feeding tube

Yes 8 (28.6) 2 (33.3) 2 (16.7) 4 (44.4)

No 19 (67.9) 4 (66.7) 10 (83.3) 5 (55.5)

Missing 1 (3.6) 0(0) 0 (0) 0 (0)

Days in NICU at birth Range of NICU days

76.25 ± 67.04(0–300)

3.50 ± 7.00(0–14)

52.73 ± 29.72(10–120)

137.33 ± 64.76(90–300)

Page 5 of 13Goracke‑Postle et al. BMC Neurol (2021) 21:384

medians) and variation (standard deviations, coefficients of variation) were calculated for each analyte.

Second, to understand the associations between ana-lytes a series of pairwise scatterplots and Pearson prod-uct-moment correlations were computed between each possible pair of analytes for the entire sample. The corre-lations were tested for statistical significance against the null hypothesis of r = 0. Parallel analyses and plots were generated for the data set with missing data imputed with the lower limit / sensitivity number. Given the large number of correlations tested (528), Type 1 errors were controlled for by using the false discovery rate correction discussed by Benjamini & Hochberg (1995). Correlations were considered against an alpha = 0.05, after the false discovery rate correction was applied.

Third, to understand how the associations between analytes varied by subgroups, the correlational analyses described above were repeated for each of the gestational term subgroups. Given the large number of compari-sons involved in these analyses, an alpha = 0.001 was set

for each test after the false discovery correction rate was applied.

Finally, differences in correlations between subgroups were also tested. To be included in between-group analysis, each correlation first had to be statistically significant within the subgroups. Thirty-two pairs of correlations were statistically significant across all sub-groups. Then the identified correlations were tested against one another to check whether they were sig-nificantly different by birth status. To do this, Fisher’s r to z transformation was used to calculate the differ-ence in correlations that met the criteria for inclusion, and tested for statistical significance of that difference; p ≤ 0.05.

ResultsCSF analyte expression in CP subjectsDetectable CSF analytes were broken down, broadly, by the following categories: hormonal/endocrine brain-derived peptides or proteins: ACTH, AgRP,

Fig. 1 Visual representation of the direction and strength of the Pearson’s correlation coefficients between all 33 analytes assayed. Positive (blue), negative (red), strong (dark shading), and weak (light shading) correlations are depicted

Page 6 of 13Goracke‑Postle et al. BMC Neurol (2021) 21:384

BDNF, CNTF, FSH, GH, LH, PRL, and TSH; inflam-matory cytokines/chemokines: IL-1α, IL-1ra, IL-6, IL-8, IL-10, IL-12p40, IL-12p70, TNFα, IFN-α2, IP-10, MCP-1, MIP1β, and RANTES; and neurotransmitters/neuropeptides: Dynorphin A, neuropeptide Y, soma-tostatin, β endorphin, cortisol, neurotensin, orexin A, substance P, melatonin, oxytocin, melanocyte-stim-ulating hormone (α-MSH). Figure  1 illustrates the correlations among analytes (at p ≤ 0.001); a full list of analytes that demonstrated correlations (p ≤ 0.05) among the participants with CP is included in the Sup-plemental Information (Supplemental Table 1).

Considering this initial complete cohort, there were 35 analyte pairings with positive correlations as per our criteria from 21 distinct analytes (all p ≤ 0.001; Table  2). These 35 correlations represent combinations of endo-crine, inflammatory, and excitatory neuropeptides. The specific correlation pattern represents a novel approach to considering analytes that may shed light on the mech-anistic underpinnings of secondary processes that may be ongoing in CP and result in clinical signs and their manifestation.

To assess the potential of such protein signatures in CSF to distinguish differences relating to biological vari-ables underlying various subpopulations of CP patients, we assessed analyte correlations between various sub-groups. This analysis provided novel information, with analyte signature correlations becoming apparent for specific subsets of participant groups. This report focuses on birth term as a defining characteristic; how-ever, the Supplemental Information provides full analy-ses of various subgroups based on additional clinical characteristics.

Gestational age subgroup analysesTerm birthNo significant correlations between analytes were detected at the pre-determined significance level (p ≤ 0.001) used to report the rest of the subgroup findings. There were, however, two correlations at p ≤ 0.05 specific to Term Birth participants (correlations which were not present for Preterm Birth or Extremely Preterm Birth participants) specifically between IL-1ra and orexin A and between orexin A and substance P (identical data for both correla-tions: Correlation (Corr) = 0.99; 95% Confidence Interval (CI) = [0.89, 1]; Adjusted P (Adj P) = 0.041) (Fig. 2).

Preterm birthPreterm Birth status resulted in clusters of analytes that were highly correlated to one another (Fig.  3). There were 14 discrete correlations not found in the other ges-tational subgroup (bolded in Table  3); these correlations

represent unique correlations in this sample specific to individuals born preterm.

Extremely preterm birthExtremely Preterm Birth status also resulted in distinct analyte correlations different from both Term Birth and Preterm Birth participants (Fig.  4). There were 24 dis-crete analyte correlations unique in this sample of par-ticipants in the extremely preterm birth status (bold in Table 4).

Table 2 Significant Pearson’s correlations (p ≤ 0.001) between analyte pairs, representing 21 distinct analyte correlations (bold font)

Analyte 1 Analyte 2 Correlation 95% CI Adj. p

Dynorphin A AGRP 0.71 [0.34, 0.85] < .001Dynorphin A ACTH 0.79 [0.60, 0.90] < .001Dynorphin A IL-12p70 0.68 [0.41, 0.84] 0.001Dynorphin A TNFα 0.80 [0.60, 0.90] < .001Dynorphin A substance P 0.78 [0.57, 0.89] < .001AGRP IL-6 0.73 [0.49, 0.87] < .001AGRP IL-8 0.71 [0.46, 0.86] < .001AGRP IL-10 0.91 [0.81, 0.96] < .001AGRP IL-12p40 0.73 [0.50, 0.87] < .001AGRP MIP-1β 0.94 [0.87, 0.97] < .001AGRP TNFα 0.67 [0.40, 0.84] 0.001

FSH LH 0.89 [0.78, 0.95] < .001TSH ACTH 0.69 [0.42, 0.84] 0.001ACTH IL‑12p70 0.82 [0.65, 0.92] < .001

ACTH TNFα 0.92 [0.83, 0.96] < .001

ACTH β endorphin 0.67 [0.40, 0.83] 0.001ACTH substance P 0.91 [0.81, 0.96] < .001

IFNα2 IL‑12p70 0.73 [0.49, 0.87] < .001IL-1α IL‑12p40 0.78 [0.57, 0.89] < .001IL‑1α IP-10 0.79 [0.60, 0.90] < .001IL-1RA orexin A 0.75 [0.52, 0.88] < .001IL‑6 IL‑10 0.87 [0.74, 0.94] < .001

IL‑6 IL‑12p40 0.68 [0.41, 0.84] 0.001

IL‑8 MIP‑1β 0.73 [0.49, 0.87] < .001

IL‑10 IL‑12p40 0.74 [0.51, 0.87] < .001

IL‑10 MIP‑1β 0.81 [0.63, 0.91] < .001

IL‑12p40 IP‑10 0.81 [0.63, 0.91] < .001

IL‑12p70 TNFα 0.91 [0.81, 0.96] < .001

IL‑12p70 substance P 0.82 [0.64, 0.91] < .001

TNFα substance P 0.93 [0.86, 0.97] < .001

β endorphin orexin A 0.81 [0.63, 0.91] < .001

β endorphin substance P 0.81 [0.62, 0.91] < .001

β endorphin αMSH 0.70 [0.44, 0.85] 0.001

orexin A substance P 0.70 [0.44, 0.85] 0.001

orexin A αMSH 0.79 [0.58, 0.90] < .001

Page 7 of 13Goracke‑Postle et al. BMC Neurol (2021) 21:384

Between subgroup analysesDifferences in correlations between subgroups were also directly tested. As noted above, to be included in between-group analysis, each correlation first had to be statistically significant within the subgroups. Thirty-two pairs of correlations were statistically significant across all subgroups. Then the identified correlations were tested against one another to check whether they were significantly different by birth status. Based on this approach, there was a significant difference between the Preterm and Extremely Preterm Birth subgroups for the correlation between TNFα and substance P (Extremely Preterm r to z = 0.99, Preterm r to z = 0.82, Z Differ-ence = 1.49, p < 0.05).

DiscussionThere are many different ways that white matter and upper motor neurons can be damaged. It is gener-ally agreed that enough damage will interfere with and

ultimately impair motor control and increase risk for the clinical condition of CP (importantly there are, of course, other associated impairments including cognitive and sensory function). Of the many putative causal agents and pathways to CP, the role of neuroinflammation in perinatal brain damage has received considerable empiri-cal attention. There has also been considerable concep-tual emphasis challenging conventional wisdom about the nature of the threat in relation to inflammation and the developing brain with a distinct shift to a perspective that a core underlying feature may be less about static events (i.e., a one-time insult) and more about sustained states (i.e., an ongoing process). The significance of this shift is in widening the focus of inquiry to include tertiary mechanisms of brain damage, which, in turn, could shed new light on a very old problem – namely treating CP based on its underlying pathophysiology.

There is a working hypothesis that suggests that the relationship among key neuroendocrine hormones,

Fig. 2 Visual representation of the direction and strength of the Pearson’s correlation coefficients between all 33 analytes assayed within the Term Birth subgroup. Positive (blue), negative (red), strong (dark shading), and weak (light shading) correlations are depicted. *In instances where the same value was reported for each variable, no correlation was calculated

Page 8 of 13Goracke‑Postle et al. BMC Neurol (2021) 21:384

excitatory neuropeptides, and neuroinflammatory cytokines as markers may be a critical variable predict-ing outcome. The conceptual basis for this is the past two decades of work supporting the cross-talk among the endocrine, nociceptive, and immune systems. In healthy states there is an optimal balance between stress hormones and proinflammatory cytokines. In children with CP, the white matter damage may be ‘driving’ an immune-mediated inflammatory cascade. There is evi-dence supporting this possibility from three related stud-ies with CP patients, one using umbilical cord serum [1], one using plasma [9], and one using blood [11].

The current analysis expands the previous work and provides further description of the molecular milieu pre-sent in the CNS of CP patients. Such documentation pro-vides a unique opportunity to consider how differences in CNS concentrations of various inflammatory-relevant analytes between differing presentations of CP may be relevant for hypotheses about mechanisms underly-ing the differential outcomes in CP. Specifically, in this

sample, it was observed that Term Birth, Preterm Birth, and Extremely Preterm Birth status was associated with distinct patterns of analyte inter-correlations. If repro-ducibility of these findings could be established, there is an opportunity to better understand the mechanisms underlying various and varying outcomes specific to birth term in CP.

The correlations present in the participants suggest relationships among systems subserving arousal (orexin A), inflammation (TNFα), anti-inflammation (IL-1ra), and neuronal excitation (substance P), with distinctions depending on birth term. Orexin A (hypocretin 1) is an excitatory, hypothalamic neuropeptide that binds to both the orexin 1 and 2 receptors (OX1R and OX2R); the former is thought to act largely through the excitatory Gq protein and both are implicated in wakefulness and sleep cycle stability [14]. Narcoleptic symptoms in dogs and mice are associated with loss-of-function mutations in the gene encoding OX2R, although immunosuppres-sion can delay symptom onset, and deletion of the gene

Fig. 3 Visual representation of the direction and strength of the Pearson’s correlation coefficients between all 33 analytes assayed within the Preterm Birth subgroup. Positive (blue), negative (red), strong (dark shading), and weak (light shading) correlations are depicted

Page 9 of 13Goracke‑Postle et al. BMC Neurol (2021) 21:384

encoding both orexins in mice results in a full narcolepsy phenotype [14]. Substance P is another excitatory neu-ropeptide that binds to neurokinin 1 receptors (NK1R), which couple through the excitatory Gq protein. Sub-stance P is contained in and released by some C-fiber nociceptors and causes activation and internalization of NK1Rs on nociceptive spinal cord neurons, exciting them [15]. These neurons project to the thalamus carry-ing information on the sensory-discriminative aspects of pain ultimately to the somatosensory cortex [16]. Thus, Substance P is thought to be one of the important driv-ers of pain transmission in spinal cord and trigeminal nucleus. IL-1ra is an endogenous anti-inflammatory cytokine that binds unproductively to the IL-1 recep-tor 1 (IL1R1), thereby blocking signaling by two pro-inflammatory cytokines IL-1α and IL-1β. The FDA has approved an altered form of human IL-1ra in the form of anakinra for use in peripheral inflammatory disorders like rheumatoid arthritis [17]. In the CNS, both of these cytokines, acting through IL1R1s on astrocytes, endothe-lial cells, and neurons, initiate transcription of multiple pro-inflammatory cytokines, including TNFα, leading to reactive gliosis and enhanced neuronal excitability fol-lowing such insults as traumatic brain injury.

In clinical investigations, evidence suggests an auto-immune destruction of orexin neurons [14], and orexin A and IL-1ra in CSF have been associated with fatigue in Sjögren’s syndrome [18]. At least one study found

decreased levels of substance P in the CSF of patients with narcolepsy [19], a disorder for which orexin A defi-ciency is a well-known biomarker [20]. Sleep dysfunc-tion is well-established in CP; our observations suggest that specific inflammatory and nociceptive mediators may be implicated and future work should be designed to investigate in more detail the relation between sleep, inflammation, and neuronal excitation in CP based on biomarker-informed molecular signatures that may be specific to birth status.

Further, the relevance of the difference between the Preterm and Extremely Preterm subgroups appears particularly striking and significant given the litera-ture documenting the clinical differences in phenotypic presentation between these two groups. Understanding the relationship between TNFα and substance P expres-sion and regulation may have implications related to the underlying pathophysiology with possible prognostic and treatment relevance. One study of cultured human astro-cytes demonstrated a functional interaction between these two analytes: substance P enhanced the stimula-tory effect of TNFα on production of two inflammatory mediators: IL-6 and PGE2 [21]. This birth term-associ-ated analyte difference may underscore the importance of our observed positive correlation to a functional interac-tion between a nociceptive neurotransmitter, substance P, and a pro-inflammatory cytokine, TNFα, as one of the inflammatory processes perhaps underlying the develop-ment and/or presentation of CP.

From a knowledge translation perspective, the correla-tion difference between Preterm and Extremely Preterm Birth subgroups for TNFα and substance P may represent a potential interaction and intervention point between inflammation and nociceptive neurotransmission spe-cifically in the CNS; however, demonstration of a spinal or trigeminal localization of this interaction would be required to underpin this interpretation. Such an under-standing may provide an opportunity to improve out-comes through earlier intervention that includes targeting these specific pathways and their unique mechanisms. One weakness of the current analysis is that the group sizes are not robust enough to distinguish larger signa-ture patterns/differences between groups that may be informational for identifying distinctions between differ-ent patient subgroup populations; ultimately, there may be different CP patient subgroups with specific, identifi-able protein signatures. Larger samples with confirmatory analyses are needed. Additionally, further consideration regarding appropriate comparison groups or control samples is warranted. Healthy controls could be useful (see below regarding a point about establishing norma-tive values) but their use could be limited in so much

Table 3 Significant Pearson’s correlations (p ≤ 0.001) between analyte pairs in the Preterm Birth subgroup are displayed. There were 14 unique positive analyte correlations (bold font) specific to the Preterm Birth subgroup that were not significant within the other gestational age subgroups

Analyte 1 Analyte 2 Correlation 95% CI Adj. p

AGRP IL-1ra 0.93 [0.75, 0.98] 0.001AGRP IL‑10 0.93 [0.75, 0.98] 0.001

IL-12.p40 IP-10 0.91 [0.70, 0.97] 0.001IL-12.p70 substance P 0.92 [0.73, 0.98] 0.001AGRP LH 0.93 [0.77, 0.98] < .001LH IL-1ra 0.94 [0.80, 0.98] < .001LH IL-10 0.94 [0.80, 0.98] < .001LH MIP-1β 0.94 [0.79, 0.98] < .001IL-1ra IL-10 1.00 [1.00, 1.00] < .001IL-1ra MIP-1β 0.97 [0.90, 0.99] < .001IL-10 MIP-1β 0.97 [0.90, 0.99] < .001IL-12.p70 TNFα 0.93 [0.78, 0.98] < .001RANTES β endorphin 0.97 [0.89, 0.99] < .001RANTES Melatonin 0.95 [0.81, 0.99] < .001β endorphin Melatonin 0.99 [0.97, 1.00] < .001

Page 10 of 13Goracke‑Postle et al. BMC Neurol (2021) 21:384

as it might be more important to consider inclusion of carefully defined samples from other neurodegenerative disorders in which more is known about the underlying immune pathology and disease progression (e.g., multiple sclerosis). Doing so would provide important points of comparison for similarities and differences in inflamma-tory analyte levels/profiles that would, in turn, increase our understanding of inflammatory mechanisms specific to the CP phenotype.

Our long-term goal is to establish clinical value in adopting a biomarker approach to understand clinical outcomes among patients with CP. A biological marker or biomarker is any characteristic that can be measured and evaluated as an indicator of normal biologic pro-cesses, pathologic processes, or pharmacologic responses to therapeutic intervention. Biomarkers hold the poten-tial of a better understanding of the etiology and pathol-ogy of a given disorder, providing valuable insight into diagnosis, treatment, and prognosis for many debilitating

disorders and diseases. Current diagnostic and therapeu-tic approaches to manage chronic disability among indi-viduals with neurodevelopmental disorders including CP are limited by our narrow understanding of the biological mechanisms underlying developmental disorders of vari-ous etiologies and confounded further by phenotypic and etiologic heterogeneity (e.g., there is not one cause of CP; clinical presentation varies widely) and the lack of bio-markers predictive of therapeutic outcome.

Part of the hope is that molecular biomarkers may pro-vide a useful ‘work around’ for closing the gap between clinical presentation and often unknown underlying pathophysiology in CP. To get there, however, there are enormous gaps in what is known from a normative per-spective within the population of CP (e.g., there are few to no referent values for expected concentration values of the vast majority of inflammatory mediators). The difficulty with any clinical sample is whether the detect-able analyte represents a biomarker for ‘exposure’ (to

Fig. 4 Visual representation of the direction and strength of the Pearson’s correlation coefficients between all 33 analytes assayed within the Extremely Preterm subgroup. Positive (blue), negative (red), strong (dark shading), and weak (light shading) correlations are depicted. *In instances where the same value was reported for each variable, no correlation was calculated

Page 11 of 13Goracke‑Postle et al. BMC Neurol (2021) 21:384

an inflammatory process) or ‘outcome’ (of brain dam-age). There is no easy solution to this dilemma absent an experimental model. With that point acknowledged, we believe there is value in continuing to adopt and adapt a ‘biomarker epidemiological’ perspective as outlined by Dammann [5]. Doing so may help facilitate the devel-opment, testing, and application of immunomodulatory therapies for CP (see Fleiss and Gressens [8] for general considerations on this and related topics specific to ter-tiary management of brain damage as well as Lee et  al. 2012 [22]).

In particular, it would be important to obtain compari-son values from healthy and/or non-inflammatory based controls to establish the utility of neurochemical profile patterns as prognostic (natural history-like outcomes) or predictive tools (what profile would be most respon-sive to immune modulatory therapy trial). With that said, as noted above, careful selection of other well-defined

diseases in which immune-mediated inflammation is a core feature would also be a valuable approach (e.g., mul-tiple sclerosis – for which stem cell treatment trials are underway with a particular focus on patients with per-sistent inflammation [23]). A third approach – in a sense the strategy used for this preliminary work – is sampling within group (patient), but it would be strengthened con-siderably by larger samples and if/when ethically feasi-ble, repeated measurement. Such a repeated measures approach was used by Koh et  al. [24] in their investiga-tion of cytokine changes in children with CP receiving intravenous granulocyte-colony stimulating factor fol-lowed by autologous mobilized peripheral blood mono-nuclear cells.

ConclusionGiven the preliminary data presented here: 1) that endocrine, neuropeptide, and inflammatory markers are detectable in CSF from pediatric patients with CP and 2) that significant correlations exist among mark-ers for endocrine hormones, nociceptive neuropeptides and inflammatory mediators that are distinct in various subgroups of individuals with CP, we think it is critical to continue this line of research to consider further the functional consequences of altered inflammatory pro-cesses and responses in children with CP and to consider the potential for mechanism-driven intervention based on protein/peptide/neurotransmitter signatures in CP patients. Increasing our scientific understanding of the neurochemical milieu involved in distinct subgroups of CP may shed light on potential targets for earlier inter-vention to perhaps prevent transition to more severe presentation. Clarifying a mechanistic understanding of developmental differences may be what the analyte pat-terns are pointing to and offer a specific starting domain for further exploration. Our hope is that these prelimi-nary findings lay the groundwork for additional studies to confirm and expand on the potential for protein signa-tures in CP to be valuable clinical tools in the future.

AbbreviationsCP: Cerebral palsy; CSF: Cerebral spinal fluid; WMD: White matter damage; TNFα: Tumor necrosis factor α; mRNA: Messenger ribonucleic acid; TLR4: Toll‑like receptor 4; IRB: Institutional review board; ITB: Intrathecal baclofen; ELISA: Enzyme‑linked immunosorbent assay; CRL: Cytokine reference labora‑tory; AgRP: Agouti‑related peptide; FSH: Follicle‑stimulating hormone; BDNF: Brain‑derived neurotrophic factor; LH: Luteinizing hormone; TSH: Thyroid stimulating hormone; GH: Growth hormone; ACTH: Adrenocorticotropic hormone; PRL: Prolactin; CNTF: Human ciliary neurotropic factor; IFNα2: Interferon alpha 2; IL‑1α: Interleukin 1 alpha; IL‑1ra: Interleukin 1 receptor antagonist; IL‑6: Interleukin 6; IL‑8: Interleukin 8; IL‑10: Interleukin 10; IL‑12p40: Interleukin 12 subunit 40; IL‑12p70: Interleukin 12 subunit 70; IP‑10: Interferon gamma‑induced protein 10; MCP‑1: Monocyte chemotactic protein 1; MIP1β: Macrophage inflammatory protein 1 beta; RANTES: Regulated on activation normal T expressed and secreted; TNFα: Tumor necrosis factor alpha; α‑MSH: Melanocyte‑stimulating hormone; DynA: Dynorphin A; PE: Phycoerythrin; Corr:

Table 4 Significant Pearson’s correlations (p ≤ 0.001) between analyte pairs in the Extremely Preterm Birth subgroup. There were 24 unique positive analyte correlations (bold font) specific to the Extremely Preterm Birth subgroup that were not significant within the other gestational age subgroups

Analyte 1 Analyte 2 Correlation 95% CI Adj. p

AGRP IL-6 0.96 [0.83, 0.99] 0.001AGRP IP‑10 0.96 [0.81, 0.99] 0.001

AGRP MIP-1β 0.95 [0.79, 0.99] 0.001ACTH β endorphin 0.96 [0.81, 0.99] 0.001IP-10 MIP-1β 0.96 [0.80, 0.99] 0.001TNFα β endorphin 0.96 [0.81, 0.99] 0.001AGRP IL-1α 0.98 [0.92, 1.00] < .001AGRP IL-10 0.98 [0.92, 1.00] < .001AGRP RANTES 0.99 [0.95, 1.00] < .001AGRP Oxytocin 0.98 [0.92, 1.00] < .001TSH Prolactin 0.97 [0.86, 0.99] < .001ACTH TNFα 0.98 [0.92, 1.00] < .001ACTH substance P 0.97 [0.88, 0.99] < .001IL-1α IL-6 0.98 [0.92, 1.00] < .001IL-1α IL-10 1.00 [1.00, 1.00] < .001IL-1α RANTES 0.99 [0.96, 1.00] < .001IL-1α Oxytocin 1.00 [1.00, 1.00] < .001IL-6 IL-10 0.98 [0.92, 1.00] < .001IL-6 RANTES 0.97 [0.87, 0.99] < .001IL-6 Oxytocin 0.98 [0.92, 1.00] < .001IL-10 RANTES 0.99 [0.96, 1.00] < .001IL-10 Oxytocin 1.00 [1.00, 1.00] < .001IL-12.p70 β endorphin 0.98 [0.92, 1.00] < .001RANTES Oxytocin 0.99 [0.96, 1.00] < .001TNFα substance P 0.99 [0.97, 1.00] < .001

Page 12 of 13Goracke‑Postle et al. BMC Neurol (2021) 21:384

Correlation; CI: Confidence Interval; Adj P: Adjusted P; M: Mean; SD: Standard Deviation; rpm: Revolutions per minute; µL: Microliters; GM‑CSF: Granulocyte–macrophage colony‑stimulating factor; PGE2: Prostaglandin E2; FDA: Food and Drug Administration; IL1R1: IL‑1 receptor 1; OX1R: Orexin 1 receptor; OX2R: Orexin 2 receptor; NK1R: Neurokinin 1 receptor; Gq: Gq protein‑coupled alpha subunit receptor; °C: Degrees Celsius.

Supplementary InformationThe online version contains supplementary material available at https:// doi. org/ 10. 1186/ s12883‑ 021‑ 02333‑2.

Additional file 1: Table 1. A Pearson’s correlation test against the null hypothesis of 0 was conducted for all possible pairs of analytes assayed, controlling for familywise Type I errors using the Benjamini & Hochberg (1995) false discovery rate correction. As there were 528 correlation tests, only those retaining a significant correlation (p ≤ 0.05) after the false discovery rate correction are presented. Table 2. Pearson’s correlation coefficients between pairs of analytes within the Term Birth gestational age subgroup. Only significant correlations (p ≤ 0.05) after controlling for false discovery rate are presented. There were no unique analyte correla‑tions specific to those with Term Birth that were not also significant within the other gestational age subgroups. Table 3. Pearson’s correlation coef‑ficients between pairs of anlytes within the Preterm Birth gestational age subgroup. Only significant correlations (p ≤ 0.05) after controlling for false discovery rate are presented. Table 4. Pearson’s correlation coefficients between pairs of analytes within the Extremely Preterm Birth gestational age subgroup. Only significant correlations (p ≤ 0.05) after controlling for false discovery rate are presented. Figure 1. Visual representation of the direction and strength of the Pearson’s correlation coefficients between analytes assayed within the subgroup with spastic CP. Positive (blue), negative (red), strong (dark shading), and weak (light shading) correla‑tions are depicted. Table 5A. Pearson’s correlation coefficients between pairs of analytes within the subgroup with spastic CP. Complete dataset of significant correlations (p ≤ 0.05) after controlling for false discovery rate are presented. Table 5B. Significant Pearson’s correlations (p ≤ 0.001) between analyte pairs in the subgroup with spastic CP. There were 8 unique positive analyte correlations (bold font) specific to spastic CP that were not significant within the subgroup with mixed tone CP. Figure 2. Visual representation of the direction and strength of the Pearson’s cor‑relation coefficients between analytes assayed within the subgroup with mixed tone CP. Positive (blue), negative (red), strong (dark shading), and weak (light shading) correlations are depicted. Table 6A. Pearson’s cor‑relation coefficients between pairs of analytes within the subgroup with mixed tone CP. Complete dataset of significant correlations (p ≤ 0.05) after controlling for false discovery rate are presented. Table 6B. Significant Pearson’s correlations (p ≤ 0.001) between analyte pairs in the subgroup with mixed tone CP. There were 7 unique positive analyte correlations (bold font) specific to mixed tone CP that were not significant within the subgroup with spastic CP. Figure 3. Visual representation of the direction and strength of the Pearson’s correlation coefficients between analytes assayed within the subgroup with seizures present. Positive (blue), nega‑tive (red), strong (dark shading), and weak (light shading) correlations are depicted. Table 7A. Pearson’s correlation coefficients between pairs of analytes within the subgroup with seizures present. Complete dataset of significant correlations (p ≤ 0.05) after controlling for false discovery rate are presented. Table 7B. Significant Pearson’s correlations (p ≤ 0.001) between analyte pairs in the subgroup with seizures. There were no unique analyte correlations specific to those with seizures that were not also significant within the subgroup without seizures. Figure 4. Visual representation of the direction and strength of the Pearson’s correlation coefficients between analytes assayed within the subgroup without sei‑zures. Positive (blue), negative (red), strong (dark shading), and weak (light shading) correlations are depicted. Table 8A. Pearson’s correlation coef‑ficients between pairs of analytes within the subgroup without seizures. Complete dataset of significant correlations (p ≤ 0.05) after controlling for false discovery rate are presented. Table 8B. Significant Pearson’s correla‑tions (p ≤ 0.001) between analyte pairs in the subgroup without seizures. There were 30 unique positive analyte correlations (bold font) specific to

those without seizures that were not significant within the subgroup with seizures. Figure 5. Visual representation of the direction and strength of the Pearson’s correlation coefficients between analytes assayed within the subgroup with quadriplegia. Positive (blue), negative (red), strong (dark shading), and weak (light shading) correlations are depicted. Table 9A. Pearson’s correlation coefficients between pairs of analytes within the subgroup with quadriplegia. Complete dataset of significant correlations (p ≤ 0.05) after controlling for false discovery rate are presented. Table 9B. Significant Pearson’s correlations (p ≤ 0.001) between analyte pairs in the subgroup with quadriplegia. There were 16 unique positive analyte correlations (bold font) specific to those with quadriplegia that were not significant within the subgroup without quadriplegia. Figure 6. Visual representation of the direction and strength of the Pearson’s correlation coefficients between analytes assayed within the subgroup without quad‑riplegia. Positive (blue), negative (red), strong (dark shading), and weak (light shading) correlations are depicted. Table 10A. Pearson’s correla‑tion coefficients between pairs of analytes within the subgroup without quadriplegia. Complete dataset of significant correlations (p ≤ 0.05) after controlling for false discovery rate are presented. Table 10B. Significant Pearson’s correlations (p ≤ 0.001) between analyte pairs in the subgroup without quadriplegia. There were 2 unique positive analyte correlations (bold font) specific to those without quadriplegia that were not significant within the subgroup with quadriplegia.

AcknowledgementsNot applicable.

Authors’ contributionsCGP: statistical design, data analysis, manuscript writing. CB: study design, implementation, manuscript writing. APM: study design, analyses, manuscript review. ME: specimen processing, specimen analyses, manuscript writing. GW: manuscript writing, manuscript review. PG: study implementation, recruitment/specimen procurement oversight, manuscript review. MP: study conception, recruitment/specimen procurement oversight, manuscript review. FS: study conception, design, manuscript writing. The author(s) read and approved the final manuscript.

FundingThis study was funded, in part, by Eunice Kennedy Shriver NICHD Grant No. 73126.

Availability of data and materialsAll data generated or analyzed during this study are included in this published article [and its Supplementary information file]. The datasets analyzed are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participateAll work described was approved by the Institutional Review Board (IRB) of the University of Minnesota. All participants provided written consent in accordance with IRB consent form review and approval. Committee reference number 0809M46301.

Consent for publicationNot applicable. All data is aggregate.

Competing interestsThe authors declare that they have no competing interests.

Author details1 College of Pharmacy, University of Minnesota, Minneapolis, MN 55455, USA. 2 Gillette Children’s Specialty Healthcare, Saint Paul, MN 55101, USA. 3 Depart‑ment of Pediatrics, University of Minnesota Medical School, Minneapolis, MN 55455, USA. 4 Departments of Neuroscience, Pharmacology, Dermatology, Uni‑versity of Minnesota Medical School, Minneapolis, MN 55455, USA. 5 Children’s Mercy, Kansas City, MO 64108, USA. 6 Department of Educational Psychology, College of Education and Human Development, Minneapolis, MN 55455, USA.

Page 13 of 13Goracke‑Postle et al. BMC Neurol (2021) 21:384

• fast, convenient online submission

•

thorough peer review by experienced researchers in your field

• rapid publication on acceptance

• support for research data, including large and complex data types

•

gold Open Access which fosters wider collaboration and increased citations

maximum visibility for your research: over 100M website views per year •

At BMC, research is always in progress.

Learn more biomedcentral.com/submissions

Ready to submit your researchReady to submit your research ? Choose BMC and benefit from: ? Choose BMC and benefit from:

Received: 29 May 2020 Accepted: 23 July 2021

References 1. Kaukola T, et al. Cerebral palsy is characterized by protein mediators in

cord serum. Ann Neurol. 2004;55(2):186–94. 2. Hagberg H, Gressens P, Mallard C. Inflammation during fetal and neonatal

life: implications for neurologic and neuropsychiatric disease in children and adults. Ann Neurol. 2012;71(4):444–57.

3. Volpe JJ. Brain injury in premature infants: a complex amalgam of destructive and developmental disturbances. Lancet Neurol. 2009;8(1):110–24.

4. Douglas‑Escobar M, Weiss MD. Biomarkers of brain injury in the prema‑ture infant. Front Neurol. 2012;3:185.

5. Dammann O, Leviton A. Inflammatory brain damage in preterm newborns–dry numbers, wet lab, and causal inferences. Early Hum Dev. 2004;79(1):1–15.

6. Neufeld MD, et al. Maternal infection and risk of cerebral palsy in term and preterm infants. J Perinatol. 2005;25(2):108–13.

7. Dammann O. Persistent neuro‑inflammation in cerebral palsy: a thera‑peutic window of opportunity? Acta Paediatr. 2007;96(1):6–7.

8. Fleiss B, Gressens P. Tertiary mechanisms of brain damage: a new hope for treatment of cerebral palsy? Lancet Neurol. 2012;11(6):556–66.

9. Lin CY, et al. Altered inflammatory responses in preterm children with cerebral palsy. Ann Neurol. 2010;68(2):204–12.

10 Leviton A, et al. The relationship between early concentrations of 25 blood proteins and cerebral white matter injury in preterm newborns: the ELGAN study. J Pediatr. 2011;158(6):897‑903.e1‑5.

11. Zareen Z, et al. Cytokine dysregulation in children with cerebral palsy. Dev Med Child Neurol. 2021;63(4):407–12.

12. Roberson R, et al. Postnatal inflammatory rat model for cerebral palsy: too different from humans. Am J Obstet Gynecol. 2006;195(4):1038–44.

13. Palisano R, et al. Development and reliability of a system to classify gross motor function in children with cerebral palsy. Dev Med Child Neurol. 1997;39(4):214–23.

14. de Lecea L. Cortistatin: a natural somatostatin analog. J Endocrinol Invest. 2005;28(11 Suppl International):10–4.

15. De Koninck Y, Henry JL. Substance P‑mediated slow excitatory postsyn‑aptic potential elicited in dorsal horn neurons in vivo by noxious stimula‑tion. Proc Natl Acad Sci U S A. 1991;88(24):11344–8.

16. Todd AJ. Anatomy of primary afferents and projection neurones in the rat spinal dorsal horn with particular emphasis on substance P and the neurokinin 1 receptor. Exp Physiol. 2002;87(2):245–9.

17. Thome JG, et al. Contributions of interleukin‑1 receptor signaling in traumatic brain injury. Front Behav Neurosci. 2019;13:287.

18. Bårdsen K, et al. Interleukin‑1‑related activity and hypocretin‑1 in cerebrospinal fluid contribute to fatigue in primary Sjögren’s syndrome. J Neuroinflammation. 2019;16(1):102.

19. Strittmatter M, et al. CSF substance P somatostatin and monoaminer‑gic transmitter metabolites in patients with narcolepsy. Neurosci Lett. 1996;218(2):99–102.

20. American Academy of Sleep Medicine. The international classification of sleep disorders : diagnostic and coding manual. 2nd ed. 2005, Westches‑ter: American Academy of Sleep Medicine. xviii, p. 297.

21. Palma C, et al. Functional characterization of substance P receptors on cultured human spinal cord astrocytes: synergism of substance P with cytokines in inducing interleukin‑6 and prostaglandin E2 production. Glia. 1997;21(2):183–93.

22. Lee YH, et al. Safety and feasibility of countering neurological impairment by intravenous administration of autologous cord blood in cerebral palsy. J Transl Med. 2012;10:58.

23. Uccelli A, et al. MEsenchymal StEm cells for Multiple Sclerosis (MESEMS): a randomized, double blind, cross‑over phase I/II clinical trial with autolo‑gous mesenchymal stem cells for the therapy of multiple sclerosis. Trials. 2019;20(1):263.

24. Koh H, et al. Serial changes of cytokines in children with cerebral palsy who received intravenous granulocyte‑colony stimulating factor fol‑lowed by autologous mobilized peripheral blood mononuclear cells. J Korean Med Sci. 2018;33(21):e102.

Publisher’s NoteSpringer Nature remains neutral with regard to jurisdictional claims in pub‑lished maps and institutional affiliations.

BioMed Central publishes under the Creative Commons Attribution License (CCAL). Underthe CCAL, authors retain copyright to the article but users are allowed to download, reprint,distribute and /or copy articles in BioMed Central journals, as long as the original work isproperly cited.