Background
The complement system, a part of the innate arm of immunity, is activated in various inflammatory conditions [
1]. A number of studies have also documented complement activity in the CNS both in acute conditions like stroke [
2] and traumatic brain injury (TBI) [
3], as well as in chronic diseases like multiple sclerosis (MS) [
4,
5] and Alzheimer’s disease (AD) [
6], all characterized by a varying degree of inflammatory features present in the tissue. The complement system consists of a large number of components, many of which are primarily produced in the liver, but some are also expressed within the central nervous system CNS [
7,
8]. Activation of the complement cascade is needed for a wide range of important defense functions, including cell-lysis, chemotaxia, opsonisation, and immune cell stimulation [
9,
10] but may also aggravate tissue damage [
3,
5,
6,
11]. The mechanisms underlying complement activation in the CNS are complex and still not fully understood. The intricate structure of the system not only offers flexibility and efficiency, but also provides a basis for multiple points of possible dysregulation [
12].A prerequisite for the agility of complement responses is a wide distribution of complement receptors, present on a range of cell types including macrophages [
13], T cells [
14], B cells [
15], microglia, and astrocytes [
16,
17]. Furthermore, many of the complement receptors exist in both membrane-bound and secreted forms, for instance, both complement receptors 1 (CR1) and 2 (CR2) exist in soluble forms: sCR1 [
18] and sCR2 [
19], respectively. Thus, the functional consequences of complement receptor expression depends not only on cellular localization, but also if the protein is anchored to the membrane or released into the extracellular space. Thus, soluble complement receptors can inhibit the cascade by binding complement proteins [
12], a fact which has been explored in complement directed therapies [
20].
We have previously described strain-dependent differences in the local expression of the upstream complement components C1q and C3 in the spinal cord of the inbred DA and PVG rat strains following a standardized nerve injury, where differences in expression of C1q correlated with subsequent nerve cell loss [
21]. Furthermore, we recently identified distinct regulatory pathways of several complement components, including C1q and C3, in a F2(DAxPVG) intercross using the same nerve injury model [
22]. The aim of the current study was to characterize and genetically map and characterize any possible differences in the local expression of complement receptors in this F2(DAxPVG) intercross.
Methods
Animals and operations
The inbred MHC congenic rat strain Dark Agouti-
RT1
av1
(DA) and the inbred MHC congenic Piebald Viral Glaxo-
RT1
av1
hereafter called PVG were bred and maintained in the in-house breeding facility under specific pathogen-free conditions and climate-controlled environment with 12 h light/dark cycles and fed standard rodent chow and water ad libitum. The F2(DAxPVG) intercross has been described previously [
22‐
24]. In brief, DA/PVG males and females were crossed generating two groups of offspring (F1), in turn mated reciprocally generating four groups of F2 progeny from which a total of 144 animals were used. Both female and male rats and an equal number of rats from each of the four groups were studied.
All animals from the F2 intercross were at an age of 9–12 weeks subjected to unilateral avulsion of the left lumbar L3–L5 ventral roots, as described in Additional file
1. Five days post-operatively, the animals were euthanized with CO
2 and perfused via the ascending aorta with ice-cold PBS containing heparin (LEO Pharma AB, Malmö, Sweden) (10 IE/ml). Spinal cords were dissected and examined using a dissection microscope to verify the completeness of the lesion and exclude any visible signs of hemorrhage or necrotic zones. After removal of the scar on the superficial part of the spinal cord, the ipsilateral ventral quadrant of L3, L4, and L5 was dissected out, and then snap-frozen for subsequent mRNA extraction.
The G12 (DAxPVG) advanced intercross line (AIL) was developed by continued structured breeding from a G10 AIL previously established [
25]. A cohort of 161 G12 animals were subjected to ventral root avulsion (VRA), with a 5-day post-operative survival and the L3 segment used in the expression studies.
An additional cohort consisting of 72 DA and PVG rats was used to analyze the kinetics following VRA. The animals were divided into five experimental groups of 5–7 individuals with 1, 3, 5, 7, or 14 days post VRA survival and one naïve (un-operated) control group. The L3 segments were taken for mRNA preparation and the L4–L5 segments were taken en bloc and snap-frozen for further preparation/sectioning.
DA and PVG animals (n = 22) were operated with left sided sciatic nerve transection (SNT) below the obturator tendon. Half the group was euthanized and perfused with ice-cold PBS with heparin 5 days after operation, and the left-sided L4 segment was taken for further RT-PCR analysis, with the right L4 segment as control. The remaining group was perfused with Lana’s fixative at the same post-operative survival. These spinal cords were dissected, kept in 4 % PFA overnight, and then transferred to 10 % sucrose in PBS before mounting and sectioning of the L3–L5 segments.
Cr2
−/−
mice (Balb/c background) were kindly provided by Professor Birgitta Heyman, Department of Medical Biochemistry and Microbiology, Uppsala University, and have previously been described [
26]. Control Balb/c mice were purchased from Charles River (Wilmington, MA).
Cr2
−/−
mice and Balb/c mice (
n = 27) were operated with left-sided SNT using the same surgical procedure as in the rats. At 5 days post-operative survival, 11 animals were euthanized with CO
2 and perfused with PBS, with dissection of the L4 segments for RT-PCR analysis, as in the rats. At the same post-operative survival, the remaining 16 animals were perfused with Lana’s fixative and the L3–L5 segments taken as described for the SNT rats.
CSF samples were obtained from seven PVG animals, three naïve and four animals with a 5-day post-operative survival after VRA. The posterior cranial fossa was defined, and a fine needle (BD Valu-set 23G, BD Pharmingen) attached to a 1-ml syringe was used to carefully aspirate CSF from the cerebellomedullary cistern using a stereotactic frame. The CSF obtained (100–150 μl from each animal) was immediately put on dry-ice and stored at −80 °C. The CSF samples were diluted 1:4 and then analyzed for presence of rat sCD21 with sandwich ELISA technique according to the manufacturer’s instructions (USCN Life, Wuhan, China, cat.nr E0750r). All tissues were stored at −80 °C.
DNA preparation and genotyping
Genomic DNA was extracted from rat tail tips according to methods previously described [
27]. Polymorphic microsatellite markers were selected from the Rat Genome Database (
http://rgd.mcw.edu) and the Ensembl database (
www.ensembl.org). The F2 intercross was genotyped with 113 microsatellite markers evenly distributed across the genome, with an average distance of 20 cM based on previous knowledge of optimum spacing [
28]. The successful genotyping rate was 95.3 %. Details of the F2 genotyping have been described previously [
22]. The G12 intercross was densely genotyped in the region of the
Cr2 gene and the peak marker from the F2 intercross, using four microsatellite markers (D13Rat192, D13Rat159, D13Rat141, and D13Rat49), with an average marker distance of 4 cM. The
Cr2 gene is located at the end of RNO13 and all markers were located up stream of
Cr2, which is why four markers were considered to give sufficient genomic resolution.
RNA and cDNA preparation
RNA and cDNA was prepared using standard methodology as previously described [
22].
All steps were performed under RNAse-free conditions. The tissue samples consisted of the injured L3 ventral cord quadrant from VRA animals in the kinetic study and the G12 cohort, while the corresponding L4 ventral cord quadrant was used in the F2 cohort. From the SNT animals, both mice and rats, both the ipsi- and contralateral L4 segments were used.
RT-PCR
Real-time PCR was conducted using a three-step PCR protocol using IQ5 and CFX384 software (Bio-Rad, Hercules CA). All primers were designed with Beacon Designer 5.0 software (Bio-Rad) and tested for specificity by running the amplified product on gels with silver staining. Two house-keeping genes were used to normalize the levels of mRNA expression of the studied transcripts; hypoxanthine guanine phosphoribosyl transferase (Hprt) and glyceraldehyde 3-phosphate dehydrogenase (Gapdh). Normalized expression levels were calculated with the IQ5 or CFX 384 software. We chose C1qb as a marker for C1q expression since it was previously shown to correlate with nerve cell loss [
21]. All primer sequences are presented in Tables
1 and
2. The
Cr2 gene is disrupted but not completely depleted in the
Cr2
−/−
which explains the small expression of the gene seen even in the
Cr2
−/−
.
Table 1
Sequences for rat RT-PCR primers
Gapdh | TCAACTACATGGTCTACATGTTCCAG | TCCCATTCTCAGCCTTGACTG |
Hprt | CTCATGGACTGATTATGGACAGGAC | GCAGGTCAGCAAAGAACTTATAGCC |
C1qb | TCATAGAACACGAGGATTCCATACA | GACCCAGTACAGCTGCTTTGG |
C3 | GGGAGCCCCATGTACTCCAT | GGGACGTCACCCTGAGCAT |
Gfap | AAGCACGAGGCTAATGACTATCG | AAGGACTCGTTCGTGCCG |
Mrf-1 | GGAGGCCTTCAAGACGAAGTAC | AGCATTCGCTTCAAGGACATAATA |
CD11b | ATCCGTAAAGTAGTGAGAGAAC | TCTGCCTCAGGAATGACATC |
Cr2 (CD21) | GGCTACCTTATGGCTGGAGAG | AGAGTCACAGTAGTCCCAAACC |
Cr1 (CD35) | GGCTTGAGACCGCTGTGAGG | TGGATTCATCAGTTGGATTTATAGGTTTGG |
CD19 | CTGTTGAGGACTGGTGGATGGATAG | CTCGCTGTCTGGCTCTTCATAGG |
Table 2
Sequences for mouse RT-PCR primers
Gapdh | TCAACTACATGGTCTACATGTTCCAG | TCCCATTCTCAGCCTTGACTG |
Hprt | CTCATGGACTGATTATGGACAGGAC | GCAGGTCAGCAAAGAACTTATAGCC |
C1qb | AGAGCAAGAGGAGGTTGTTCAC | GCAGGAAGATGGTGTTGGATAGG |
C3 | GCTGCTGTCTTCAATCACTTCATC | GCCTCTTGCCTCTTCTCTATGC |
CD11b | CCCAGAGGCTCTCAGAGAATGTC | CTTCATCTTCTGAAAGTCAATGTT |
Cr2 (CD21) | AACCTGGCTATTTGCTCACTGG | ACTTTCCTGGATGTTCACACTGG |
Gfap | GGTAAGATGACTGAGCGGATGG | TCGTGGTAAAGACTGTGGAGATG |
Mrf-1 | GGAGGCCTTCAAGACGAAGTAC | AGCATTCGGTTCAAGGACATAATA |
Microarray hybridization
RNA from 144 L4 ventral cord quadrants from F2 animals were used for microarray hybridization using Affymetrix Rat gene 1.0 ST Array chips (Affymetrix, Santa Clara, CA). See Additional file
1 for details. The microarray data is available in MIAME-compliant (minimal information about a microarray experiment) format at the ArrayExpress Database (
http://www.ebi.ac.uk/arrayexpress) under accession code E-MTAB-303.
eQTL analysis and gene expression network construction
An expression quantitative trait locus (eQTL) is a gene region that regulates gene expression rather than a functional trait or disease. eQTL mapping is performed by combining expression data from global gene expression profiling with a complete linkage map, derived from whole-genome mapping using for instance microsatellite or SNP markers dispersed throughout the genome. This enables the identification of large numbers of expression QTLs and the study of co-regulated transcripts at various positions of the genome. eQTLs are classified into cis- or trans-acting according to the distance between the locations of genetic marker and the affected transcript. For a cis-acting QTL, the expression is regulated from the same position as the gene is located in the genome.
The current eQTL mapping has been described previously [
22]. In brief, microarray gene expression data were normalized using the RMA algorithm [
29], implemented in the Bioconductor package “oligo.” Raw expression intensities were background corrected, quantile normalized, log2 transformed, and summarized on the level of transcript clusters. Transcript annotation was taken from the biocondutor package “ragene10sttranscriptcluster.db.” eQTLs were mapped for all transcript clusters using the QTL reaper software [
30] against the 113 genomic markers, and 10
6 permutations were performed in order to assess genome-wide significance of eQTLs;
p < 0.01 at genome-wide level was considered significant. The eQTLs were classified into
cis- or
trans-acting according to the distance between the locations of genetic marker and the affected transcript. If the distance was smaller than 20 Mb, we assumed
cis- and otherwise
trans-regulation. The identified Cr2 cluster (D13Rat49) was analyzed for enrichment of specific pathways and expression patterns using the Biocondutor package GOstats [
31]. To enable identification of strongly connected hub genes in the eQTL gene expression network, we applied a graphical Gaussian model (GGM) and for each cluster of
trans-regulated transcripts, we constructed gene expression networks as previously described reporting significant edges with FDR < 0.1 [
32].
Primary astrocyte and microglia cultures
Primary astrocytes and microglia were isolated from adult PVG brains of 10-week-old rats. A detailed protocol can be found in Additional file
1. The astrocytes and microglia were left unstimulated (only DMEM/F12 complete medium, supplemented with 10 % heat-inactivated FCS, penicillin-streptomycin 100 units/ml, 100 μg/ml) or stimulated with recombinant rat TNF-α (R&D Systems, Minneapolis, MN) at the concentration 20 ng/ml for 24 h after which the cells were lysed for RNA extraction and subsequent RT-PCR expressional analysis.
Immunohistochemistry and quantification of synaptophysin immunoreactivity
Rat and mouse spinal cord sections were serially cut (14 μm) on a cryostat (Leica Microsystems, Wetzlar, Germany) at the level of the L4 segment. Detailed protocols can be found in the Additional file
1. For rat antisera directed against synaptophysin (rabbit anti-rat 1:200, Invitrogen, Carlsbad, CA), GFAP (mouse anti-rat, 1:400, Sigma), Iba1 (rabbit anti-rat, 1:200, Wako, Richmond, VA), or C3 (mouse anti-rat, 1:100 Abbiotec, San Diego, CA) was used together with appropriate flourophore-conjugated secondary antibody (Cy3 donkey anti-rabbit 1:500, Jackson Immuno Research, West Grove, PA; Alexa Fluor 488 donkey anti-rabbit, 1:150 and Alexa Fluor 594, goat anti-mouse, 1:300, both from Invitrogen). For mouse, the following antisera were used: GFAP (mouse anti-rat, 1:400, Sigma), CD68 (rat anti-mouse, 1:100, Abcam, Cambridge, UK), and CD21 (rabbit anti-mouse 1:200, Abcam) with the following secondary antibodies; Alexa 488 donkey anti-rat 1:150 and Alexa 568, donkey anti-rabbit 1:300 (Invitrogen).
Sections were examined in a Zeiss LSM 5 Pascal confocal laser scanning microscope (Carl Zeiss GmbH, Göttingen, Germany) or a Leica DM RBE microscope system (Leica). Semiquantitative measurements of immunoreactivity were carried out in ImageJ (NIH, Bethesda, MD) on confocal images. The immunoreactivity in the ventral horn of the spinal cord was compared to the corresponding contralateral side in the same spinal cord section. The images were captured in the optical plane with the maximal immunoreactivity, and all settings for compared images were identical. At least four spinal cord sections from each animal were measured, and the mean ipsilateral/contralateral signal ratio for each animal was used for statistical analysis.
Statistical analysis
The software program R 2.6.0 was used to carry out statistical analyses and create all graphs depicting eQTL localization using the package qtl1.14-2. LOD > 3.3 corresponds to
p < 0.0001 [
33]. One-way ANOVA calculated with GraphPad Prism 5.0 (San Diego, CA) were carried out on RT-PCR and protein data; results are represented as mean ± SEM. Unpaired
t test was used to assess differences in immunoreactivity (GraphPad Prism 5.0). Correlations between genes in the co-expression network in the F2 intercross were calculated using Pearson’s algorithm assuming equal distribution and visualized graphically using linear regression plots, also in GraphPad Prism 5.0. In general,
p < 0.05 was considered statistically significant, except for in the microarray analysis where
p < 0.01 was used.
Study approval
All experiments in this study were approved by the ethical committee for animal experimentation (Stockholms Norra Djurförsöksetiska Nämnd), Stockholm, Sweden, under ethical permits N42/06, 225/08, N32/09, N343/10, and N122/11.
Discussion
Using an unbiased genetic approach, we find here a strong naturally occurring cis-acting genetic influence on the local expression of Cr2 in the spinal cord following nerve root injury. Our experimental data support the notion that CR2, previously mostly studied in context of B cell biology, is an integral part of the inherent CNS response to injury and that it may modulate synaptic plasticity by regulating upstream complement activity. Hypothetically, this could be done by secreted/shed soluble CR2.
Mapping of the genetic effect was carried out in a standardized nerve injury model, VRA, where motor axons are severed at the interface between the peripheral and central nervous systems, resulting in substantial loss of injured nerve cells. For technical reasons, a more peripheral nerve injury model, SNT, was used in mice. SNT also induces inflammatory activation of microglia and astrocytes in the vicinity of the lesioned motor neurons, albeit with more limited loss of axotomized cells than VRA [
40,
49,
50]. The so-called axon reaction has been a subject of study for more than a century, initially focusing on the characteristic morphological changes occurring in injured cells [
51]. Subsequent research has shown alterations in electrical activity at the level of the cell body believed to be the caused by elimination of afferent synaptic input to lesioned neurons [
52]. After acute mechanical injuries, the elimination of synaptic input has been assumed to be beneficial, perhaps by limiting excitotoxic stress [
50,
53]. However, more recent studies suggest that reduced loss of synapses after SNT or VRA is associated with improved functional outcome [
41,
54]. Likely, the molecular systems regulating plasticity of nerve terminals have to be finely tuned as excessive loss of synaptic connectivity may be an important biological substrate for chronic neurodegenerative diseases, which has been demonstrated to occur as an early disease-related phenomenon [
55,
56]. An increasing body of evidence suggests that the complement system plays an important role for synaptic plasticity. Thus, upstream complement proteins such as C1q and C3 mediate synaptic remodeling during development [
57,
58]. They are also implicated in normal aging, since a dramatic increase in C1q occurs in the brain of old mice and humans [
59], as well as in neurodegenerative diseases such as AD and Parkinson’s disease, and also MS [
6,
60]. The notion that dysregulation of the complement system occurs upstream rather than downstream of an inherent neurodegenerative process is supported by the observation that genetic variability in both
Clusterin and
CR1 is associated to risk of late-onset AD [
61,
62]. Although the functional consequences of complement activation in chronic neurodegenerative disease need to be defined more in detail, it seems plausible that excessive activation may exaggerate a neurodegenerative process.
The recent genetic association of
CR1 to AD [
62] led us to examine any possible naturally occurring strain differences in expression of the four most studied complement receptors, Cr1-4, after nerve injury. Cr2-4 could all be shown to be regulated from distinct genetic regions, with the most striking finding for Cr2, which was subject to a very strong monogenic
cis-acting regulation. In contrast, Cr1 was not regulated by injury in the rat.
CR2 has been studied mostly in context of B cell immunology, where it is expressed on mature B cells and forms a complex with CD19 and CD81. This tri-molecular formation functions as a co-receptor complex, where CR2 binds opsonized C3d and antigen-bound IgM resulting in an enhanced antigenic B cell response [
63]. However, it is unlikely that B cells cause the observed strain differences in Cr2 expression due to the low expression of its B cell binding partners. The role of CR2 in the CNS has received little attention, but expression of CR2 has been previously reported on activated astrocytes and has been demonstrated to regulate neurogenesis in the mouse [
17,
43].
Microglia are suggested to be involved in the emoval of synapses occurring after CNS injury, and complement is known to increase their phagocytic properties [
58,
64]. In the rat strains studied here, the microglia response was stronger in DA than PVG after both SNT and VRA. Interestingly, only CD11b (Cr3), but not Mrf-1, displayed higher levels in DA, suggesting the possible existence of microglia subsets with different phagocytic abilities. This may have functional implications as mice deficient for
Cr3 (CD11b/CD18) display reduced loss of synapses during development of the visual system [
58].
The similar expression pattern of Cr2 and Gfap, and the Cr2 expression in the astrocyte cultures suggest astrocytes as a source of locally expressed Cr2 in the CNS. The existence primarily of a soluble form could explain the difficulty to stain for CR2 in the CNS [
43]. We examined this by measuring sCR2 levels with Elisa in CSF, which verified the presence of soluble CR2 and that levels increased following injury. The increase in sCR2 may not only reflect increased transcription, but can be the result of increased shedding of the CR2 ectodomain, which corresponds to sCR2, due to oxidative stress triggered by VRA [
24,
65].
The results presented herein suggest that activated PVG astrocytes produce more CR2/sCR2, in turn of possible relevance for protecting the integrity of synaptic networks. This interpretation is supported by results obtained in
Cr2
−/−
mice, which display increased loss of synaptophysin-labeled synaptic terminals after SNT. Possibly, deficiency in CR2 also affects glial activation after nerve injury, with enhanced microglia and reduced astrocyte activation, as supported by quantification of the microglia immunolabeling signal in the axotomized sciatic motor pool. This is also supported by the constructed gene expression network, where Cr2 correlates positively with anti-inflammatory and negatively with pro-inflammatory transcripts, and by previous findings, where sCR2 was shown to modulate monocyte activity [
66]. CR2 binds to C3dg and iC3b, both breakdown products of C3 [
67]. Thus, increased sCR2 after injury may serve a regulatory role as a binder of deposited C3 fragments in the area of inflammation. However, it is also possible that CR2 could have a more complex role in altering the balance or activation steps of the C3 activation/cleavage cascade, with multiple active breakdown products. For instance, iC3b is the primary ligand for CR3 (CD11b/CD18) [
68], whereas C3b is not bound by CR3 [
69], but instead by CR1 [
70]. This suggests that increased levels of sCR2 in context of nerve injury leads to decreased iC3b generation, which in turn attenuates the process of synaptic loss, as the CR3 (CD11b/CD18) positive microglia have been shown to be involved in removal of synapses [
58]. It could also explain the increased microglia activation seen in the Cr2
−/− mice, as iC3b binding to CR3 can contribute to increased microglia activation [
44].
The complexity of complement activity is illustrated by a recent study, which may at first seem contradictory to ours, since
Cr2
−/−
mice demonstrated improved initial outcome after experimental TBI [
71]. However, this model differs considerably from the ones used here, since in TBI there is a dramatic loss of blood-brain barrier function leading to influx of immune cells and complement proteins from the systemic circulation. Furthermore, it is difficult to dissociate the effect of targeting both CR1 and CR2, since in mice, but not in humans and rats, the two proteins are coded in the same gene. It should also be underscored that the presence or expression of complement proteins is not synonymous with complement activation, as defined by the release of complement protein fragments with immune signaling properties and activation of the complement cascade [
12]. Lastly, it is also of importance to note the different roles of soluble as compared to membrane bound forms of complement receptors, where they in one form may be activating and in the other regulatory.
Acknowledgements
We would like to thank Professor Birgitta Heyman for kindly providing the Cr2
−/−
mice. This work was supported by the 6th Framework Program of the European Union, NeuroproMiSe, LSHM-CT-2005-018637, EURATools, LSHG-CT-2005019015, and the 7th Framework Program of the European Union, EURATrans, HEALTH-F4-2010-241504, by the Swedish Research Council, the Swedish Brain Foundation, the King Gustaf V:s 80-years foundation, Dr. Åke Olssons Stiftelse för utbildning, faculty grants from the Linneaus University, and the Swedish Association of Persons with Neurological Disabilities. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
RL and FP conceived the experiments and wrote the manuscript. AB, RL, and JZ performed the mouse experiments. RL and AB analysed the mouse data. RL, MS, SA, FAL, NA, KH, and MD performed the rat experiments. SA and CD performed the cell cultures. MH, MS, and RL performed the bioinformatics analysis. NH provided the facilities for the bioinformatics studies. BN and KNE provided reagents and edited the manuscript. SC provided the reagents and facilities for the mouse experiments. FP provided the facilities and funding. All authors read and approved the final manuscript.