Unilateral L4-dorsal root ganglion stimulation evokes pain relief in chronic neuropathic postsurgical knee pain and changes of inflammatory markers: part II whole transcriptome profiling

During initiation of the sampling processes 1 sample was lost, hence in the current study we have 11 subjects with CRPS who were fully analyzed at baseline and after 3 months of selective L4-DRG stimulation.

RNA was isolated from the frozen blood and whole transcriptome analysis was performed to identify changes in gene expression pre- and post-stimulation. Blood was collected in EDTA tubes and RNA was extracted with the help of NucleoSpin RNA Blood Midi Kit (Cat# 740210.20, Macherey–Nagel GmbH & Co. KG). 100 ng of RNA from each sample were subjected for gene expression analysis using Clariom-S-human array from Affymetrix^® (Santa Clara, CA). GeneChip^® Hybridization Oven 645 were used for hybridization, GeneChip Fluidics Station 450 for washing and GeneChip Scanner 7G was used for scanning. After scanning, CEL-Files were generated from DAT-Files by Affymetrix^® GeneChip^® Command Console^® software. Data was analyzed by using Affymatrix^® Power Tools Programm (version 1.15.1) first subtracting background, followed by normalization, and then summarizing expression. R-Packages Program was used to obtain the differential expression, q-values and quality check.

Pathway enrichment of differentially expressed genes, obtained from gene expression assay, was performed by two different prediction tools: GOrilla^® (Gene Ontology enRIchment anaLysis and visuaLizAtion tool) and PANTHER^® V. 14.0 (Protein ANalysis THrough Evolutionary Relationships; http://​www.​pantherdb.​org). This software was used for transcriptome profiling allowing rapid analysis and standardized interpretation of large data in a fully automated workflow [9].

The PANTHER database represents a classification of peptide-coding genes (104 sequenced genome) according to the evolutionary relationships of peptide function included in the Gene Ontology (GO) Phylogenetic Annotation Project and Biological Pathways databases. This includes a phylogenetic tree according to the ontological function, which in turns allows to identify ontologically related peptide subfamilies based upon the hidden Markov model (HMM). The PANTHER model also enables the use of distinct analysis tools (gene list enrichment, Reactome^® database) to characterize peptide-coding genes based on the molecular function, biological processes and pathways [10‐15]. Finally, the most significantly altered genes were screened and a heat map of upregulated (green) and down regulated genes (red) was prepared.

As reported in our previous study, after 3 months of selective L4-DRG_STIM pain levels significantly declined (mean NRS; pre-DRG_STIM: 74.90 ± 16.3 versus 1-week DRG_STIM: 42.50 ± 13.18 versus 3 months DRG_STIM: 46.65 ± 27.52; p = 0.003) with improved mood and sleep quality. Pro- and anti-inflammatory serum cytokines concentrations were higher in CRPS subjects compared to healthy controls and remained increased post L4-DRG_STIM (HMGB-1, TNF-α, IL-6, leptin) with the exception of circulating pro-inflammatory IL-1β. Serum anti-inflammatory IL-10 declined and saliva neuropeptide oxytocin concentrations increased after L4-DRG_STIM.

Table 1 summarizes baseline characteristics and applied DRG stimulation pattern and Table 2 the peripheral levels of inflammatory mediators assessed in our previous study [1].

Pathway enrichment analyses showed predominant involvement of inflammation mediated chemokines/cytokines pathway and gonadotropin releasing hormone receptor pathway (Figs. 1, 2).

Further screening of differentially regulated genes showed changes in immune response related genes and genes involved in signal transduction. We observed that after 3 months of L4-DRG_STIM, 21 genes with predominant involvement of immunoinflammatory responses showed significant differential expression (Fig. 3). Out of these 21 genes, seven [7] genes including toll-like receptor 1 (TLR1; − 1.6-fold, p = 0.01), free fatty acid receptor 2 (FFAR2; − 1.5-fold, p = 0.04), interleukin 1 accessory protein (IL1RAP; − 1.4-fold, p = 0.02), interleukin 1 receptor antagonist (ILRN; − 1.3-fold, p = 0.04), complement component 5 (C5; − 1.3-fold, p = 0.01), PKB (PDZ binding kinase; − 1.2-fold, p = 0.04) and interleukin 18 (IL18; − 1.2-fold, p = 0.05) were downregulated (Figs. 3, 4). The remaining fourteen [14] genes including chemokine (C-X–C motif) ligand 2 (CXCL2; 1.2-fold, p = 0.001), chemokine (C–C motif) ligand 11 (CCL11; 1.2-fold, p = 0.01), interleukin 36 gamma (IL36G; 1.2-fold, p = 0.02), C-reactive protein (CRP; 1.2-fold, p = 0.02), secretoglobin, family 1A, member 1 (SCGB1A1; 1.2-fold, p = 0.03), interleukin 17F (IL-17F; 1.2-fold, p = 0.002), tumor necrosis factor receptor superfamily, member 4 (TNFRSF4; 1.3-fold, p = 0.006), phospholipase A2, group IIA (PLA2G2A; platelets, synovial fluid; 1.2-fold, p = 0.02), cAMP responsive element binding protein 3-like 3 (CREB3L3; 1.3-fold, p = 0.006), ADAM metallopeptidase with thrombospondin type 1 motif 12 (ADAMTS12; 1.3-fold, p = 0.001), interleukin 1 family, member 10 (theta) (IL1F10; 1.3-fold, p = 0.001), NADPH oxidase 1 (NOX1; 1.3-fold, p = 0.006), acidic chitinase (CHIA) and bradykinin receptor B1 (BDKRB1; 1.3-fold, p = 0.003) were upregulated (Figs. 3, 5, 6). The most prominently downregulated gene was TLR1, with a − 1.6-fold downregulation (p = 0.01) in comparison to pre-L4 DRG_STIM. TLR1 is known to be involved in mediating the inflammatory response through toll-like receptors signaling.

On the other hand, the most prominent upregulated (1.3-fold; p = 0.003) gene was BDKRB1 that is involved in leukocyte migration and immunoinflammatory response through G-protein coupled receptor signaling. The details of intensity of up or downregulation of each gene have been summarized in Table 3.

Orlova and colleagues examined small-noncoding RNAs (miRNAs; micro-RNAs) in blood and compared CRPS subjects versus healthy controls. Significantly increased levels were observed for ILRa, vascular endothelial growth factor and monocyte chemotactic protein (MCP-1). In addition, miRNAs expression correlated with co-morbidities frequently observed in CRPS such as headache, thyroid disorders and/or use of narcotics/antiepileptic drugs indicating biomarkers having a potential for patient stratification in CRPS associated pain [16]. However, sufficiently powered clinical studies relating the effects of DRG_STIM in CRPS patients assessing potentially therapeutic changes in gene expression (e.g. alterations in cytokine and chemokines levels and ratios) have been lacking.

SCS impacts pain signaling in CRPS via inhibiting GABA-ergic and cholinergic interneurons of the spinal cord organized in functional compartments (Laminae I–V). The pathological vascular state in CRPS has been associated by orthodromic (inhibition of the sympathetic tone) and antidromic (activation of spinal afferents) spinal routes [17, 18]. In a sophisticated hypothesis Russo and colleagues discussed possible relevant pathways addressed to the development of CRPS. Briefly summarized, the authors speculated, that in a time-dependent manner CRPS pathophysiology may be the consequence of a deteriorated complex interplay initially driven by local cellular and molecular players (T-cells, B-cells, dendritic cells, cytokines), ultimately leading to microglia activation in the CNS, which in turns promotes peripheral inflammation and drives the maintenance of CRPS. Dendritic cells are known to interact with some of the transcriptional factors of our exploratory study and represent a key cell population with multifunctional attributes including local inflammation synthesis and binding, migration via lymphatic nodes, systemic blood circulation, blood brain barrier penetration via meningeal, choroid plexus and periventricular space related pathways in the CNS. Based upon this, the described transcriptional alterations in our trial appear to have potential validity [18].

Our analysis of CRPS subjects revealed 21 immunoinflammatory-related genes were differentially expressed and significantly changed after L4-DRG_STIM. Seven (7) genes were significantly downregulated (Table 3): TLR1 (integral plasma membrane component; TLR4 signaling), FFAR2 (G-protein coupled receptor activity; cytokine activity; glucose hemostasis), IL1RAP (innate immune response; IL-1 receptor activity regulation), ILRN (inhibition of glutamate secretion; metabolic processing; response to starvation; insulin secretion; sensory perception of pain; chronic inflammatory signaling of IL-1), C5 (inflammatory response; cell chemotaxis; response to stress; leukocyte migration), PKB (negative regulation of immune response) and IL18 (regulation of gene expression; T-helper 1 type immune response; positive regulation of inflammation). Interestingly, interleukin 1 receptor related genes (IL1RAP and ILRN) associated with IL-1β pre-/post synaptic signaling were found to be downregulated. Of note, in line with previous preclinical and human studies we found significantly increased serum levels of pro-inflammatory IL-1β at baseline, but not after 3 months of L4-DRG_STIM compared to healthy controls in our previous trial [1]. TLR is richly expressed on dendritic cells surface processing noxious stimuli derived from bacteria/viruses (PAMP, pathogen associated molecular patterns) or with injury-related molecules. Noxious stimuli activating TLR leads to upregulation of pro-inflammatory mediators evoking a host response from the innate and adaptive immune system. The transcriptional downregulation in our cohort may reflect the observed L4-DRG_STIM responsiveness. Relational conclusions between such changes and clinical responsiveness are limited in view of the uncontrolled study design, but could be answered in future prospective comparative studies comparing different spinal modulation wave forms.

On the other hand, fourteen (14) genes were upregulated (Table 3). These included CXCL2 (regulation of cell proliferation extracellular; cell chemotaxis in inflammatory response), CCL11 (signal transduction; cellular response to IL-1; chemotaxis of neutrophil, mast cells, monocyte and lymphocyte), IL36G (positive regulation of IL-6; regulation cytokine synthesis; cell–cell signaling, innate immune response), CRP (regulation of gene expression, vasodilatation; aging; defense response to bacteria), SCGB1A1 (negative regulation of IL-4, IL-5, T-cell proliferation to inflammatory response), IL-17F (metabolic; cartilage development; cytokine synthesis of IL-2, IL-6, IL-8), TNFRSF4 (cellular defense/apoptosis; T-cell proliferation; regulation of transcriptional processes), PLA2G2A (lipid catabolic and small molecule metabolic function), CREB3L3 (transcription of RNA/DNA; positive regulation of acute inflammation), ADAMTS12 (cellular response to IL-1, TNF-α, VEGF; chondrocyte differentiation), IL1F10 (IL-1 receptor binding; cytokine mediated signaling pathways), NOX1 (angiogenesis; blood pressure control; metabolic function; signal transduction), CHIA (metabolic function; digestion; inflammatory response) and BDKRB1 (sensory perception of pain; response to mechanical stimuli). TNFR1 as the major player of TNF-α signaling has been associated with neuropathic pain related depressive-like behavior in animal NP model along with morphological hippocampal alterations such as impaired neurogenesis and reduced neuroplasticity. It is noteworthy, that some experimental studies observed a gender-dependency of TNF signaling possibly driven by hormonal factors [19, 20]. On a peptide level, we found significantly increased IL-6, TNF-α levels along with impaired mood in our previously published study, which in part reflect the upregulated genes IL36G, IL-17F and TNFRSF4. In addition, both upregulated IL36G and IL-17F genes are known to modulate IL-6 synthesis [1]. Relative to the body mass index (BMI), 75% of our study population was classified as pre-obese to obese class I-III. In addition, metabolic disorder associated impairment (hypertension, diabetes, cardiac ischemia) was diagnosed in every CRPS subject of our study [1]. Immunometabolism and its potential impact on the development of chronic pain has been recognized in the past 20 years. Possible relationships to specific therapeutic approaches such as electrical neuromodulation of the brain and the spinal cord remain an under-investigated issue. Immunometabolism markers such as leptin have been associated with cellular and molecular mechanisms linking obesity to white adipose tissue (WAT) dysfunction, which in turn promotes a broad spectrum of metabolic (cardiovascular, diabetes) and/or inflammatory associated disorders (chronic pain, neuropsychiatric disorders). However, the complex puzzle of reciprocal interactions remains incomplete so far and the authors strongly recommend, to include immune-metabolic assays among other neuroinflammatory mediators in future targeted neuromodulation research for chronic pain.

These changes of the assessed transcriptome are associated with impaired sensory processing of pain, metabolic disorders and impaired mood/sleep in our CRPS cohort. Gene expression of blood cells relevant for the inflammatory host response and associated cell signal transduction were altered in our current study. However, the findings indicate an ongoing pathophysiology. If DRG stimulation interacts with CRPS pathophysiology on non-neural circulating pathways involved in the underlying mechanism of CRPS is unknown.

Notably, only a few experimental studies exist indicating an impact of SCS/DRG patterns on genetic modification on the spinal cord/dorsal root ganglion level relevant for the neural transmission of nociception [5‐8].

Alteration of the membrane function (ion flux across neuronal and glial membranes), increased peptide expression leading to hyperexcitability in DRG/spinal cord neurons (T-type Ca²⁺ currents), decreased neuronal activity at laminae I/II and opioid receptor-dependent mechanisms may contribute to the genesis of chronic neuropathic pain [1, 2, 21]. Disbalance of T-type Ca²⁺ currents (Ca_v 3.2 T-type channel) leads to hyperexcitability of DRG and spinal cord neurons resulting in abnormal pain/sensory perception originating from peripheral nerve injury. Although the precise mechanism remains to be fully clarified, upregulation of T-type Ca²⁺ currents may be mediated by an increase of distinct receptors subtypes (e.g. Ni²⁺ insensitive current component) leading to reduced action potential thresholds, increased action potential frequency and ectopic discharges [21].

Experimental studies have assessed possible changes in gene expression of pain-associated inflammatory peptides evoked by SCS utilizing a current-controlled stimulation protocol (50 Hz, 20 µs, 70% motor threshold, 0.3–1 mA) [2‐4]. Interestingly, SCS at spine level L1 reversed experimentally induced increased RNA expression of IL-1β, IL-6, GABA B receptor 1 and sodium/potassium ATPase [5]. In another experimental setting the same working group confirmed some of the changes in gene expression in the spinal cord also occurred in the ipsilateral DRG suggesting a modulatory effect of SCS at both sites [6]. In an additional holistic gene assay approach, spinal cord genes encoding inflammatory peptides and ion channel regulatory proteins were found to respond to SCS in the presence and absence of an external noxious stimuli. Contrary, dorsal root ganglion inflammatory genes were only modified in the presence of noxious stimuli (injury) [7, 8].

Although transcriptomic assay in our study did not determine gene expression changes at the neural interface, the data presented in this study represent the first in-human DRG-SCS trial assessing changes in gene expression involved in neuroinflammatory host response in circulating blood cells [22]. Pro-inflammatory IL-1β and corresponding components (molecular/cellular) promote the development of chronic pain, metabolic disorders and neuropsychiatric disorders itself, but are frequently present in chronic pain patients, which in turn makes relational conclusion difficult due the multifunctional characteristics of cytokines/chemokines coding genes.

Given the multifunctional characteristics of genes assessed in our cohort, interpretation of the study findings is limited, and the application to clinical practice still requires additional work. Pro-inflammatory IL-1β and corresponding components promote the development of chronic neuropathic pain, metabolic disorders, cardiovascular diseases and neuropsychiatric disorders. On the other hand, impaired sensory perception, hypertension, heart failure, diabetes, obesity and mood disorders are highly prevalent in chronic pain patients. Even in the absence of pain-related co-morbidities, it may be useful to measure mediators associated with these co-morbidities in order to identify patients at risk to develop co-morbidities in the future. This study has several limitations including the uncontrolled design, the small-scale study cohort, lack of a sham control arm and short observation period. Furthermore, multiple variables and confounders include demographic differences (age/gender), lifestyle factors (physical activity, food intake, sleep quality), psychological issues (depression, catastrophizing) and pre-analytic variables (sample collection, storage, processing) [23]. Large-scale databank-driven phenotyping for DRG stimulation candidates is to establish a set of neuroinflammatory-based biomarkers and predictors for chronic pain amenable to this technique of neurostimulation therapy [24, 25].