SYSTEMS ANALYSIS OF TRANSCRIPTIONAL DATA PROVIDES INSIGHTS INTO MUSCLE’S BIOLOGICAL RESPONSE TO BOTULINUM TOXIN

Abstract

Introduction

In this study we provide global transcriptomic profiling and analysis of botulinum toxin A (BoNT-A)–treated muscle over a 1-year period.

Methods

Microarray analysis was performed on rat tibialis anterior muscles from 4 groups (n =4/group) at 1, 4, 12, and 52 weeks after BoNT-A injection compared with saline-injected rats at 12 weeks.

Results

Dramatic transcriptional adaptation occurred at 1 week with a paradoxical increase in expression of slow and immature isoforms, activation of genes in competing pathways of repair and atrophy, impaired mitochondrial biogenesis, and increased metal ion imbalance. Adaptations of the basal lamina and fibrillar extracellular matrix (ECM) occurred by 4 weeks. The muscle transcriptome returned to its unperturbed state 12 weeks after injection.

Conclusion

Acute transcriptional adaptations resemble denervated muscle with some subtle differences, but resolved more quickly compared with denervation. Overall, gene expression, across time, correlates with the generally accepted BoNT-A time course and suggests that the direct action of BoNT-A in skeletal muscle is relatively rapid.

Skeletal muscle contraction is controlled by impulses received from the central nervous system via the neuromuscular junction (NMJ). In cases where skeletal muscle function is impaired due to altered activity of nerve impulses, as in disorders such as cerebral palsy (CP), it can be advantageous to suppress muscle contraction by reducing NMJ activity. Signal reduction can be achieved by physically decoupling the muscle and nerve by selective dorsal rhizotomy,¹ or through the use of chemical agents such as neurotoxins.² One such neurotoxin in common clinical use is onabotulinum toxin A (BoNT-A), which has applications ranging from decreasing spasticity, tics, and tremors, to managing pain and controlling glandular secretions.³

BoNT-A is 1 of 7 serotypes produced by Clostridium botulinum that functions to reversibly paralyze muscle by affecting the NMJ. BoNT-A reduces pre-synaptic acetylcholine (ACh) release by specifically cleaving a SNARE protein, SNAP25, required for its exocytosis. BoNT-A–induced neuromuscular block causes physical and physiological changes to the NMJ and skeletal muscle fiber.^4,5 Experimental studies have shown that muscle reinnervation via neuronal sprouting begins immediately after injection, with control slowly reverting back to the parent terminal over time.^6,7 It has also been observed that, during this time, skeletal muscle is characterized by reduced fiber size, paresis, and atrophy,⁸ until it gradually regains functionality. When used as a therapeutic agent in disorders such as CP, BoNT-A is administered by intramuscular injection repeatedly over extended periods of time. Although clinical experience has demonstrated that the injection effects last 3–6 months,⁹ there is not yet a cohesive temporal picture or a clear understanding of underlying muscle functional and transcriptional regulation. Although the effects of BoNT-A treatments in skeletal muscle have been studied extensively experimentally,^4,7,10 to the best of our knowledge, only a single genomic study was published that focused on certain genes associated with BoNT-A action in skeletal muscle.¹¹ In contrast, we now report a complete systems analysis of the BoNT-A–treated skeletal muscle transcriptome over a period of 1 year, with the goal of understanding the underlying biological response to BoNT-A and the relationship between transcriptional and functional changes associated with its reversible paralysis. We analyze our results in the context of “physiological families” of skeletal muscle, as presented recently.^12,13 Thus, the primary goal for this study was 2-fold: (1) to create a documented model for global transcriptional changes that occur with neurotransmitter blockade using BoNT-A in skeletal muscle; and (2) to gain insights into the biological basis for adaptation and recovery of muscle after BoNT-A treatment.

METHODS

Animals

All procedures were performed with the approval of the institutional animal care and use committee of the University of California, San Diego. Mature male Harlan Sprague-Dawley rats (age 3 months, weight 399 ± 3.05 g) were given a single 100-μl injection in the tibialis anterior (TA) muscle with either saline or saline with 6 U/kg BoNT-A (Botox; Allergan, Irvine, California). At 1, 4, 12, and 52 weeks after injection, rats were euthanized by intracardiac pentobarbital sodium (0.5 ml of 390-mg/ml solution) injection. Maximum isometric contraction strength was measured for all rats before euthanasia, as described previously.¹⁴ After the animals were euthanized, bilateral TA muscles were excised, weighed, and snap-frozen in isopentane cooled with liquid nitrogen (−159°C). All samples were stored at −80°C for further analysis.

Hydroxyproline Assay

A modified version of the hydroxyproline assay¹⁵ was used to determine collagen content. Briefly, muscles were hydrolyzed at 110°C overnight in hydrochloric acid, then methyl red was added and samples were pH adjusted. Chloramine T and p-diaminobenzaldehyde were added sequentially to the samples, which were then incubated for 30 minutes at 60°C. A standard curve was determined, and samples were read at 550 nm and 558 nm.

RNA Preparation

Samples were prepared for 5 groups (n =4/group) that included tissue from TAs of BoNT-A–injected rats at 1, 4, 12, and 52 weeks after injection. Control tissue was obtained from the contralateral TA of saline-injected rats euthanized at 12 weeks. RNA was extracted with Trizol (Invitrogen, Carlsbad, California) and RNeasy (Qiagen, Valencia, California). Briefly, 30 mg of frozen tissue was mixed with 0.5 ml of Trizol and homogenized at 4°C in a bullet blender (Next Advance, Inc., Averill Park, New York). The homogenate was mixed with 100 μl of chloroform, and samples were incubated for 2 minutes at room temperature and spun at 4°C for 15 minutes. The aqueous portion was removed and mixed with equal amounts of 70% ethyl alcohol. The solution was then washed through an RNeasy spin column, incubated for 15 minutes with RNAse-free DNAse (Qiagen), washed 3 times, and eluted according to the manufacturer’s instructions. Absorbance was measured at 260 nm to determine RNA concentration, and the 260/280-nm absorbance ratio was calculated to determine RNA purity. RNA was reverse-transcribed into cDNA using a synthesis system (SuperScript First-Strand Synthesis System; Life Technologies, Grand Island, New York).

Microarray Data Collection

Microarrays (RG-230 2.0; Affymetrix, Santa Clara, California) were used for analysis of all samples. The Cancer Center Microarray Shared Resource (University of California, San Diego) provided RNA processing and quality control using the GS FLX System (Roche Diagnostic Corp., Basel, Switzerland).

Real-Time Quantitative Polymerase Chain Reaction

Real-time quantitative PCR (qPCR) was conducted to validate expression of 8 genes (Chrna1, Myl3, Sln, Myog, Aqp4, Runx1, Scd1, Atp1b4) with cDNA prepared from RNA samples used for microarray analysis. We also quantified the expression of muscle-specific tyrosine kinase (MuSK) receptor through qPCR as it was undetectable at any time-point on our gene chip. RNA was reverse-transcribed into cDNA using the synthesis system (SuperScript First-Strand). Samples were diluted 1:100, and qPCR was performed using KAPA SYBR FAST Master Mix (Kapa Biosystems, Woburn, Massachusetts) and the Eppendorf Mastercycler System (Eppendorf, Hamburg, Germany). Primers for Chrna1, Sln, Myl3, Myog, Aqp4, Runx1, and GAPDH were designed in Oligo 6.8 (Molecular Biology Insights, Cascade, Colorado, and Allele Biotechnology, San Diego, California), whereas those for Scd1, MuSK, and Atp1b4 were ordered pre-made from Integrated DNA Technologies (Coralville, Iowa). Primer sequences for these genes are listed in Table 1.

Table 1.

List of forward and reverse strand primer sequences that were utilized for validation of gene expression using qPCR.

Gene	Forward sequence	Reverse sequence
Chrna1	TACTTGAATCCTTTCGCGCT	CTTAACCGCTGAGCCATCTC
Sln	TGGTGTGCACTCAGAAGTCC	TGAGGAGCACAGTGATCAGG
Myl3	AATCCTACCCAGGCAGAGGT	CATATGTGCCCGTGTCTTTG
Myog	ACCAGGAGCCCCACTTCTAT	TTACACACCTTACACGCCCA
Aqp4	GCATGTGATCGACATTGACC	GTGAAACAAGAAACCCGCAT
Runx1	TAACCCTGCCTGGGTGTAAG	GGACTCGGATCTTCTGCAAG
GAPDH	AGACAGCCGCATCTTCTTGT	TGATGGCAACAATGTCCACT

A temperature gradient was used to determine the optimal reaction temperature for each primer based on the DNA melting temperature curve and single product production on an agarose gel. Samples were run in triplicate using the following protocol: samples were heated to 95°C for 2 minutes; then run through 40 cycles of heating at 95°C for 15 seconds, cooling to 55°C for 15 seconds, and heating for 20 seconds to the optimal primer temperature determined by the temperature gradient described above. The triplicate results of each gene from qPCR data were normalized with respect to the housekeeping gene, GAPDH. Fold change was computed in accordance with a previous publication.¹⁶

Microarray Data Preprocessing

Expression data were preprocessed using software available through R¹⁷ and Bioconductor.¹⁸ Gene chip robust multi-array average (GCRMA) was employed for normalizing expression using the “gcrma” function available through the GCRMA package.¹⁹ All raw. CEL files along with GCRMA-normalized data are available through Gene Expression Omnibus (GEO),²⁰ accession GSE52350. Outlier samples were those with average intersample correlation <2 standard deviations (SD) below the mean. A single array at 1 week (3.4 SD below mean) was removed. Annotation files for RG 230 2.0 (GPL 1355) were downloaded from GEO. Multiple probes were accounted for using the “collapseRows” function in R’s WGCNA library.²¹ All probes with missing ENTREZ gene identifiers were excluded from this study. Based on this processing, we obtained a final reduced data set containing log2-based normalized expression values of 13,751 genes across 19 samples.

Differential Analysis through Pairwise Comparisons

Pairwise comparison between every time-point BoNT-A injected vs. saline was performed using Cyber-T²² Bayes regularized analysis for 2-sample unpaired data, with a confidence interval of 8. This study utilized control tissue from saline-injected rats 12 weeks after injection for all pairwise comparisons, in contrast to using age-matched controls. (Rats from this time were considered adult animals, representative of rats from the other time-points in the study.) Previous studies showed that skeletal muscle glucose uptake²³ and muscle protein expression²⁴ changes little among rats until they reach >18 months of age. Because all our rats were within this age group, we considered it acceptable to perform pairwise comparison using a control from a single time-point. Fold change for each gene was computed as the difference in mean log-based expression between treated and control samples. Genes with a log2-based fold of >1 and a Benjamini–Hochberg (BH) P <0.05 were identified as being significantly differentially expressed, as presented in Table S1a (refer to Supplementary Material, available online).

Enrichment Analysis

DAVID²⁵ was used to identify enrichment of genes (categories: GO_BP_FAT and KEGG_PATHWAYS) as shown in Table S1b (online).

RESULTS

Differential Gene Expression over Time

Gene expression changed dramatically during the experimental time period. Table 2 summarizes the number of genes that were identified as being differentially regulated at each time-point. Consistent with previous studies, pairwise analysis revealed that muscle is transcriptionally hyperactive, with dramatic transcriptional changes at 1 week (compared with 4, 12, and 52 weeks). Visual analysis of differentially regulated genes suggests that the bulk of regulation occurs at 1 week, with a large fraction of genes (1718 of 1989) being exclusively and significantly regulated at this time (Fig. 1). As expected, the genes regulated at 1 week cover a wide spectrum of functions, such as stabilizing the NMJ, sarcomeric contraction, and muscle metabolism. Of the 113 genes regulated exclusively at 4 weeks, most were associated with extracellular matrix (ECM) and collagen fibril organization (Table S1b, online). No genes were regulated significantly across the entire course of the study.

Table 2.

Summary of differentially regulated genes identified at each time (with respect to saline injected muscle, BH <0.05).

Time (in weeks)	Differentially expressed	Upregulated	Downregulated
1	1989	1183	806
4	372	303	69
12	32	19	13
52	32	19	13

FIGURE 1.

A 4-way Venn diagram depicting the distribution of differentially expressed gene across all pairwise comparisons and counts of overlapping genes between multiple pairwise comparisons. The count within each shaded area represents the number of differentially expressed genes identified in common between time-points. For example, 231 genes were differentially expressed at both 1 and 4 weeks.

Systems Analysis of Differential Expression in Skeletal Muscle

In contrast to using traditional ontology enrichment to analyze transcriptional regulation, we systematically categorized and analyzed differentially expressed genes in the novel framework of “physiological networks” specifically identified in skeletal muscle from 2 previous studies.^12,13 Wang et al.¹³ identified families of genes based on 4 major functions occurring in skeletal muscle, namely mechanical, metabolic, excitation–contraction coupling, and signaling, whereas Langfelder and Horvath²¹ characterized the physiology of the muscle into 8 distinct “networks” required for its functioning. Taken together, these models have identified gene networks that are crucial for normal skeletal muscle function and homeostasis.

Utilizing these models to guide our analyses, we derived a systems view of the regulation underlying skeletal muscle after BoNT-A treatment. Based on our data set, transcriptional activity of muscle can be grouped into 7 networks: (1) neuromuscular junction; (2) the excitation–contraction coupling (ECC) system; (3) muscle contraction; (4) energy metabolism and mitochondrial biogenesis; (5) ECM; (6) oxidative stress; and (7) muscle atrophy and recovery (Fig. 2 and Table S2, available online). Each of these networks can be considered in their physiological context in light of the genes measured.

FIGURE 2.

Overview of transcriptional changes occurring in adult skeletal muscle after BoNT-A treatment. Functional changes represented in blue boxes show associated gene expression until 4 weeks after injection. Genes in yellow boxes exhibit upregulation, and genes in blue boxes exhibit downregulation. Genes identified with an asterisk represent immature muscle isoforms. Signaling pathways are indicated with red text. Dotted lines indicate association, and arrows indicate a cause–effect relationship.

Neuromuscular Junction

Expression changes at the NMJ are presented graphically in Figure 3. Consistent with previous experimental studies, BoNT-A injection leads to rapid disruption and repair of the NMJ. Genes encoding postsynaptic proteins were detected, including the adult nAChr subunits Chrna1, Chrnd, and Chrne, as well as the developmental subunit Chrng, which is usually only expressed in humans prior to week 33 of gestation. The coreceptor for Agrin, Lrp4, and Emb and linker protein Rapsn were all upregulated. Chrna1 and Emb were upregulated until 4 weeks. Two immature isoforms of Na²⁺ and K⁺ channels, Scn5a and Kcnn3, were upregulated significantly at 1 week. Genes selectively involved with the synaptic basal lamina, including Lama5, Col4a5, and Nid2, were upregulated only at 4 weeks.

FIGURE 3.

Expression of genes involved in the neuromuscular junction over time. Each box has 4 partitions representing the time-points at which samples were obtained. Each box represents the fold changes observed and is colored according to the legend. Genes that are not represented in the microarray or identified as differentially expressed are presented in gray. Solid lines indicate either an interaction among genes or an association between genes.

Excitation–Contraction Coupling

Genes involved in ECC and maintenance of calcium homeostasis, such as ion pumps and ion channels, were differentially regulated, especially at 1 week (Fig. 4), along with Cacnb1 (an L-type voltage-gated Ca²⁺ channel) and Fkbp1a (an Ryr1 binding protein). Sustained and significant upregulation of sarcoplasmic Ca²⁺ handler sarcolipin (Sln) occurred up to 12 weeks. Jph1 and genes required to modulate cytosolic Ca²⁺ levels, including Pde4d, Calm3, and Camk2a, were downregulated at 1 week.

FIGURE 4.

Expression of genes involved in excitation–contraction coupling and muscle contraction after BoNT-A injection. Cellular localization as well as expression levels are depicted as described in Figure 3.

Aqp4 was the most strongly downregulated gene, solely at 1 week, along with several other K⁺ ion channels, such as Kcnc1, Kcnab1, and Kcnj11, and ion pumps, such as Atp1b1 and Atp1b.

Muscle Contraction and Activation

Muscle contraction requires coordinated effort between the contracting sarcomeres and cytoskeletal framework. There is a general downregulation of genes associated with activating sarcomere contraction in fast fibers, particularly 1 week after injection (Fig. 4). These include tropomyosin (Tpm3, Tpm2), troponins (Tnnc1, Tnnt1, Tnni1), and tropomodulin (Tmod1), and genes that encode proteins associated with the sarcomeric contractile apparatus, such as myosin light chains (Myl2, Myl3), Myl kinases (Mylk2), Mybpc2, and myosin heavy chains (Myh2, Myh7). There was a significant downregulation of the M-line structural proteins, including myomesins (Myom1 and Myom2), Z-disk–associated proteins such as Actn3, Myot, myozenins (Myoz1, Myoz2), and Ldb3. Cytoskeletal proteins Ank1, Sgc, and LARGE were downregulated, whereas cytoskeletal proteins required to increase sarcolemmal stability were upregulated (Csrp3, Dysf, Dtna, Flnc, Lmna). We also observed strong upregulation of a muscle-specific calcium-handling protein, Ankrd1, up to 4 weeks. Upregulation of certain cardiac isoforms, such as Actc1, Myl6b, and Tnnt2, along with immature isoforms normally absent from adult muscle (Myh3, Myh8), was also observed. Consistent with this observation was the appearance of developmental myosin isoforms in 14% of the 1-week and 1-month injected muscles, but not in control muscle.

Energy Metabolism and Mitochondrial Biogenesis

Genes involved in energy metabolism, specifically mitochondrial energy production from glycolysis and β-oxidation, were downregulated significantly 1 week after injection (Fig. 5). The glucose transporter, Glut4/SLc2a4, and glycolysis intermediates and enzymes, Pgm5, Gys, Pygm, and Pfkfb1, were downregulated at 1 week. Other enzymes involved in generation of pyruvates acting in the cytosol, such as Gpi, Pfkm, Pgam2, Eno3, Pkm2, Pdk4, Ldha, and Ldhb, were downregulated at 1 week. Enzymes involved in each step of the TCA cycle, including Cs, Aco2, Idh2, Idha, Idhb, Idhg, Dlst, Dld, Suclg1, Sdha, Sdhb, Sdhc, Sdhd, Fh1, Mdh1, and Mdh2, were downregulated. AMP deaminase (Ampd3), required for replenishing TCA cycle intermediates, was upregulated strongly.

FIGURE 5.

Mitochondrial metabolism in skeletal muscle and a depiction of the expressed fold changes of genes involved. Cellular localization as well as expression levels are depicted as described in Figure 3.

Genes associated with β oxidation and lipid metabolism were all downregulated. These included: fatty acid transporters Cd36 and Fabp3; Lipin-1 [required to break down triacylglycerol to free fatty acids (FFA)]; ATP-dependent enzymes required to convert FFA into long, medium, and short acyl-CoA esters (Acadvl, Acsl6, Acss1) and their transporters (Cpt2); Hadh; and Echs1, Echdc1, and Echdc2. Prolonged upregulation of stearoyl–coenzyme A desaturase 1 (Scd1) was observed for most of the study. Genes of the immediate adenosine triphosphate (ATP) replenishment system of muscle, Ckmt2 and Ak1, were downregulated. Major energy/ATP availability sensor, AMPK α/β/γ (Prkaa1, Prkag3), was upregulated. Targets of AMPK, the peroxisome proliferator–activated receptor-γ (PPAR) cofactors, Ppargc1a and Ppargc1b, were downregulated.

Downregulation of solute carriers necessary for metabolism, such as several members of the mitochondrial phosphate transporter family (Slc25–Slc25a23 in particular) and members of the monocarboxylate transporter subfamily (Slc16–Slc16a3 in particular) was observed.

Changes to Extracellular Matrix

There was a general upregulation of ECM genes, particularly at 4 weeks after injection. Genes encoding proteins of the basal lamina, such as Fbn1, and its collagens (Col4a1, Col4a2, Col8a1), were upregulated (Fig. 6). Genes associated with the fibrillar ECM, including Col1a1, Col1a2, Co114a1, Col3a1, Col5a1, Col5a2, Col5a3, and Col6a3, and other ECM-associated genes, such as Lum, Ctgf, Bgn, and Postn, were upregulated. Enzymes Lox and Loxl1, involved in collagen cross-linking, were also upregulated. S100a4, a biomarker correlated with proliferation of fibroblasts, was upregulated through 12 weeks. Increased collagen at the protein level in samples from 4 weeks was also detected using the hydroxyproline assay (Fig. S1, online).

FIGURE 6.

Diagram of fold changes of genes involved in the basal lamina and fibrillar ECM. Cellular localization as well as expression levels are depicted as described in Figure 3.

Oxidative Stress Response

The most striking change in expression of genes involved in oxidative stress was the global activation of chemoprotective and antioxidant genes, especially at 1 week, which involved the isoforms of glutathione S-transferase (Gst, specifically Gstm1 and Gstt2), Gpx3, Hmox, Nqo1, Aldh3a2, Txn1, and metallothioneins (Mt1a, Mt2a). The mitochondrial reactive oxygen species (ROS) scavenger, Sod2, was downregulated at 1 week.

Muscle Atrophy and Recovery

After BoNT-A injection, muscle appears to activate conflicting cellular programs, showing simultaneous signs of breakdown and repair. Upregulation of myogenic regulatory factors (MRFs), Myod1, Myog, and Myf6, at 1 week, with a concomitant and drastic increase in 2 potent regulators of cell proliferation, Cdkn1a and Cdkn1c, was observed. These, in conjunction with activated Pcna, serve as markers of satellite cell activation in skeletal muscle. Signaling pathways active in BoNT-A–treated skeletal muscle are as follows (Fig. 7):

FIGURE 7.

Illustration of a representative set of active transcription factors and signaling pathways involved in atrophy and muscle recovering from BoNT-A injection and their fold changes across time. A detailed list of genes is provided in Table S2 (available online). Cellular localization as well as expression levels are depicted as described in Figure 3.

Transforming growth factor-beta (TGF-β) pathway

Several genes in the TGF-β pathway, including Tgfb2, Fst, Myc, Ltbp1, and early response factors downstream, Junb and Fos, were upregulated significantly up to 4 weeks. Small GTPases, RhoA and RhoC, downstream of the TGF-β pathway, were upregulated at 1 week with Mstn receptor Acvr2b, Acvr1, and a TF, Atf4, downregulated at 1 week. Interestingly, inhibitors of Tgfb1, such as Smad7 and Fkbp1a, were upregulated at 1 week.

Nuclear factor–kappaB (NF-κB) signaling

Several genes, including Traf2, Nfkb2, and Nfkbie, and ubiquitin ligases downstream of the NF-κB pathway, Atrogin1/Fbxo32 MuRF1/Trim63, and Casp3, were upregulated at 1 week. Positive activators of the NF-κB pathway, such as Ascc2 and Litaf, were also upregulated at 1 week.

MAPK signaling

Several members of the MAPK family were upregulated at 1 week, including Mapk1, Mapk3, Map3k1, Map3k14, Map4k4, and Mapk1ip1, as well as its downstream targets, such as Eif4e2, which initiates protein translation and activation of Myod1.

Insulin signaling

Although we found no significant regulation of Igf1 in our study, several genes of this pathway were upregulated at 1 week, including Igf1r, Igfbp5, and Shc, and downstream, Pik3r4 and Akt1. Regulation of Igfbp5 and Igf2 is observed at 4 weeks. Glut4/Slc2a4 and Irs1 were downregulated at 1 week.

ID signaling pathway

Inhibitor of DNA binding (ID) proteins Id1, Id2, Id3, and Id4 were upregulated up to 4 weeks. This pathway is believed to play a role in repairing muscle.

TP53/cell cycle control

Activation of genes that may play a role in satellite cell proliferation and activation of apoptosis, including Cdkn1a, Gadd45a, Pcna, and Myc, were upregulated at 1 week.

VEGF pathway

Genes involved in angiogenesis, including Vegfa, Vegfb, angiopoietins (Angpt1, Angpt2), Nos3, Rtn4, and Nrp1, were downregulated at 1 week.

Transcriptional regulation of factors required for proteolytic degradation, such as Ca²⁺-dependent calpains (Capn2, Capn3) and lysosomal cathepsins (Ctsl1), was seen. Runx1, a transcription factor that promotes anti-atrophic programs, was upregulated strongly up to 4 weeks.

Validation of Regulated Gene Expression Using qPCR

To validate the chip-based expression results, we performed qPCR on a subset of 8 relevant genes. Some have been highlighted previously as being active in skeletal muscle during atrophy, specifically after BoNT-A treatment, including Myog, Chrna1, Sln, and Myl3. Genes that were shown more recently to be active in atrophy/muscle recovery from atrophy include Runx1.²⁶ Genes that were ranked highly in our analyses with a demonstrated role in skeletal muscle include Scd1 and Aqp4, and another with no known role in skeletal muscle, but highly ranked in our differential analysis, was Atp1b4. The average fold change of all genes was normalized with respect to GAPDH. Average fold change was computed with reference to the saline-injected samples. The trends seen in qPCR were similar to those computed through microarray differential analysis (Fig. 8). Quantitative correlation between relative gene expression levels from microarray data and qPCR (r²) ranged from 0.74 to 0.99.

FIGURE 8.

Fold changes observed based on the qPCR assay compared with the microarray data. Each plot shows a comparison between the gene’s calculated average fold change (log2-based) with respect to control using qPCR and microarray analysis computed for each time-point.

Correlation of Gene Expression with Muscle Function Postinjection

Because measurement of isometric contraction was made before and after BoNT-A injection on the same set of rats used for gene expression analysis, we could study the correlation over the time course of our experiment. We correlated isometric contraction strength postinjection with genes identified as significantly altered for the 1- and 4-week time periods (Table S1b, online). At 1 week, 72 genes were positively correlated and 37 genes were negatively correlated (P <0.05) with contraction strength. These same 109 genes showed the opposite weak correlation (albeit non-significantly) with isometric force prior to injection (Fig. 9, left), suggesting that expression levels may be functionally significant. Positively correlated genes were enriched for skeletal muscle contraction and include genes such as Chrna1, Myl2/3, Tnni1, Tnnc1, Lama5, Scn5a, Myoz2, and Tpm3 (Table S1b, online). Because a functional contractile apparatus is required for muscle contraction, it is not surprising that increasing expression of related genes may improve contraction strength after BoNT-A injection. At 4 weeks, we observed 15 genes correlating positively (P <0.05) and 37 correlating negatively (P <0.05; Fig. 9, right) with isometric force after injection. Negatively correlated genes were overrepresented for angiogenesis, cell death, and ECM (such as Lox and Col1a1) (Table S1b, online). Although it has been shown previously that fibrosis and ECM remodeling may lead to abnormal muscle function,²⁷ the link between angiogenesis and muscle force is less clear. These data may be indicative of still-injured muscle undergoing continued repair and regrowth at 4 weeks.

FIGURE 9.

Heat map showing the correlation between differentially expressed genes with isometric contractile strength at (a) 1 week (left) and (b) 4 weeks after injection (right). Green represents negative correlation and red represents positive correlation.

DISCUSSION

This study is a high-throughput analysis of global expression changes in BoNT-A–treated mammalian skeletal muscle over a period of 1 year (Fig. 2). Albeit with some differences, the transcriptional regulation observed in chemodenervated muscle after BoNT-A is similar to that seen in denervation models²⁸ and neuromuscular diseases,^29,30 with suppression of metabolism and muscle contraction, activation of atrophic pathways, and increased oxidative stress. We discuss the observed regulation in what follows, defining the “early” response, which is more complex, and the “late” response to neurotoxin.

Early Response to Botulinum Toxin Injection

Alterations in ECC and Sarcomeric Contraction

Reduced availability of ACh due to BoNT-A–impaired exocytosis causes increased expression of specific nAChRs, Chrnd, Chrne, embryonic Chrng, and especially Chrna1.⁵ As previously reported, overexpression of Chrng, ordinarily undetectable in adult skeletal muscle, is a compensatory mechanism to create greater current flow, as it is localized along the entire length of the fiber.^31,32 Although our microarray did not capture MuSK, a crucial protein for nAChR clustering after BoNT-A¹⁰ (upregulation confirmed through qPCR; see Fig. S2, online), Lrp4 (a coreceptor for Agrin³³) and Rapsn (required for clustering nAChRs on the postsynaptic membrane³⁴) were upregulated significantly. The observed upregulation of Lrp4 and Emb may serve as a “retrograde signal” to stabilize the NMJ, facilitating formation of terminal sprouts and induction of nAChRs.^35,36

Imbalance in ion flux after BoNT-A was reflected in the regulation of adult isoforms of K⁺ gating channels such as Kcnc1 and Kcnab1, and inwardly rectifying Kcnj11. Interestingly, strong upregulation of 2 non-adult voltage-gated Na²⁺/K⁺ ion channels, Scn5a (a cardiac isoform recently implicated in the occurrence of fibrillation potentials in denervated muscle fibers³⁷) and Kcnn3 (a K⁺ channel implicated in fibrillation and hyperpolarization of denervated muscles^38,39), suggests hyperexcitability of BoNT-A–injected muscle, analogous to denervated models. Reduced electrical activity also implies absence of an active need to maintain the Na⁺/K⁺ gradient, reducing the utility of certain Na⁺/K⁺ ion pumps, such as Atp1b1 and Atp1b2.⁴⁰

The transmission of neural excitation past the NMJ terminates at a specialized set of voltage sensors within the muscle T-system called dihydropyridine receptors (DHPRs), which are coupled mechanically to ryanodine receptors (RYRs) in the sarcoplasmic reticulum (SR). Downregulation of docking protein Jph1 (which holds the T-system spatially close to the SR) suggests instability in the structural and spatial association between the SR and T-tubules at 1 week. Upregulation of Cacnb1 of the DHPR Fkbp1a⁴¹ (essential for minimizing Ca²⁺ leakage) in conjunction with the protracted and sustained upregulation of Sln (which inhibits the uptake of Ca²⁺ back into the SR) suggests increased availability of cytoplasmic Ca²⁺, in contrast to denervated muscle.⁴²

Genes that affect free Ca²⁺ dynamics, such as Calm3, calcineurin (Ppp3cb/Ppp3ca), and Camk2a, were clearly regulated. Aqp4, a water channel expressed at the sarcolemma of fast-twitch skeletal muscle, which has an expression that is altered in dystrophy and atrophy, was the single most strongly downregulated gene. Although the exact physiological role of Aqp4 in skeletal muscle has yet to be defined, recent research on Aqp4^−/− mice suggests it has a role in regulating the osmotic balance of muscle, affecting Ca²⁺ handling.⁴³ Taken together, these data suggest a lack of calcium homeostasis and impaired Ca²⁺ handling in BoNT-A–treated muscle, especially at 1 week after injection. These results are consistent with muscle responding to decreased neural activity.

The transmitted action potential in normal skeletal muscle is ultimately converted to mechanical contraction through physical coupling of several proteins within the muscle.¹³ As expected, we observed suppression of several adult sarcomeric proteins of fast muscle. Downregulation of Myoz1, combined with upregulation of Csrp3,²⁸ suggests a shift in fiber composition at 1 week after BoNT-A treatment. This “mixed” state of expression beyond 1 week is further compounded by upregulation of genes expressing cardiac immature fiber isoforms, such as Actc1, Myl6b, and Tnnt2, with the largest increases occurring in the expression of Myh3 and Myh8.

A disrupted state of the sarcolemma at 1 week is evidenced by downregulation of Ank1⁴⁴ (necessary to maintain integrity of network SR) along with upregulation of several other cytoskeletal proteins, γFilamin, Sgc, Dmd, Dtna, and Dysf. Overall, the observed activation of several mixed muscle isoforms points to activation of programs not seen in adult skeletal muscle, reinforcing the general idea that muscle injected with BoNT-A reverts to a more “immature” state in order to recover contractile function.

Reduced Metabolism and Impaired Mitochondrial Biogenesis

With BoNT-A–induced paralysis, there is reduced requirement for ATP consumption. Akin to denervation models,²⁸ we observed suppression of most genes involved in energy metabolism and production (via both oxidation of lipids and glycolysis), specifically at 1 week, which resolve by 4 weeks. As described in the Results section and in Figure 5, there is clear downregulation of enzymes involved in energy production via glycolysis, except for hexokinase (Hk6).

It has been reported previously in atrophy with preferential loss of fast muscle fibers that there is dramatic upregulation of Ampd3⁴⁵ (replenishes TCA-cycle intermediate substrates). Dramatic and prolonged upregulation of Ampd3 in our study not only suggests an impaired TCA cycle, but further supports the idea of a shift in fiber composition. Another observation of BoNT-A–treated muscle is prolonged upregulation of Scd1 (also validated using qPCR). Deficiency in Scd1, a rate-limiting enzyme that catalyzes the synthesis of monounsaturated fatty acid, has been correlated with increased oxygen consumption and subsequent β-oxidation in skeletal muscle.^46,47 Conversely, overexpression studies have shown decreased fatty acid oxidation, increased TAG synthesis, monosaturation of muscle fatty acids, and impaired glucose uptake and insulin signaling pathway.⁴⁸ The fact that it is upregulated until 12 weeks after injection leads us to speculate that Scd1 may play a significant role in reduced energy production (via β oxidation) in addition to playing a role in reducing glucose uptake after BoNT-A injection.

Suppression of Ckmt2 [outer mitochondrial membrane enzyme required for generating ATP from phosphocreatine and adenosine diphosphate (ADP)] and Ak1 (cytoplasmic enzyme that catalyzes generation of ATP from ADP) also points to reduced ATP turnover in injected muscle.

Upregulation of major energy/ATP availability sensors, AMPK α/β/γ (Prkaa1, Prkag3), point to reduced availability of ATP at 1 week. However, the targets of AMPK, the PPAR cofactors^49,50 Ppargc1a and Ppargc1b, known biomarkers of mitochondrial biogenesis in skeletal muscle, were downregulated,⁵¹ suggesting possible stress-induced impairment of mitochondrial biogenesis.

Increased Oxidative Stress and Metal Ion Imbalance

Denervation and immobilization studies have demonstrated repeatedly that 1 of the causes of atrophy is increased accumulation of ROS and trace metals in skeletal muscle.^52–54 In that same vein, we observed transcriptional activation of several oxidative stress markers implicated in atrophy, such as the metallothioneins (Mt1a, Mt2a).^45,55 Most striking, however, was the increase in the various isoforms of glutathione S-transferase, including Gstm1 and Gstt2, as a compensatory response to increased production of ROS or oxidative stress.⁵⁶ Interestingly, however, the mitochondrial ROS scavenger superoxide dismutase 2 (Sod2) was downregulated and may have been confounded by mitochondrial dysfunction.

Imbalance of metal ion concentration has been reported previously in studies of immobilization and disuse.⁵⁷ Zinc ion homeostasis has been linked closely to a redox state of cells in various tissues.⁵⁸ We observed significant upregulation of zinc SLCs (Slc30a2, Slc30a4), which are suggested to confer a cytoprotective effort by preventing cells from free Zn ion toxicity.⁵⁹ Although the exact physiological role of Zn ion toxicity in chemodenervated muscle is not understood fully, we hypothesize that the observed upregulation of these transporters in conjunction with increased expression of metallothioneins suggests a metal ion imbalance that may contribute to BoNT-A–induced atrophy of muscle.

Competing Pathways Contributing to Concomitant Atrophy and Recovery of Skeletal Muscle

Atrophy and consequent muscle loss in skeletal muscle can occur through activation of the NF-κB signaling pathway (Traf2, Nfkb2, and Nfkbie, and positive activators Ascc2 and Litaf) along with activation of the TGF-β pathway.⁶⁰ Loss of muscle mass has been attributed to accelerated proteolytic degradation of the contractile apparatus through deployment of factors such as Capn2, Ctsl1, and Casp3,⁶¹ and eventual degradation of the fragmented actin–myosin complexes through the ubiquitin-proteasomal system. Similar to denervation studies, but in contrast to neuromuscular diseases,^29,30 activation of atrophy markers, Atrogin1/MAFbx (Fbxo32) and Trim63 (MuRF1), 2 muscle-specific ubiquitin ligases downstream of the NF-βB pathway, is observed after BoNT-A treatment.^11,62

Several studies have demonstrated the role of TGF-β signaling in atrophying skeletal muscle and the powerful role of TGF-β family growth factors such as myostatin (Mstn) in regulating muscle size.⁶³ Although differential regulation of Mstn was not observed in our study, follistatin (Fst),⁶⁴ a myostatin inhibitor, was upregulated significantly. This, along with the repression of Acvr2b, a transmembrane activin receptor of Mstn, points to inhibition of the pro-atrophic action of Mstn in injected muscle. Upregulation of early-response genes downstream of Tgfb1, including Junb and Fos along with small GTPases RhoA and RhoC and its inhibitors (Smad7 and Fkbp1a), further emphasizes the conflicting signaling of muscle treated with BoNT-A. In contrast to Na²⁺/H⁺, K⁺-ATPases, such as Atp1b1, Atp1b2, and Atp1b4, have been shown to localize to intracellular stores, predominantly the inner myonuclear membrane, in perinatal skeletal muscle of placental mammals and to regulate TGF-β signaling in skeletal muscle. Although no direct evidence of its localization patterns exists, we hypothesize that prolonged upregulation of Atp1b4 (validated via qPCR) may be contributing to similar functions in BoNT-A–treated muscle.

Insulin-like growth factors (IGFs) and their role in upregulation of nAChRs, muscle growth,^69,70 and their metabolic effects, have been studied extensively. The observed regulation of IGF binding proteins, such as Igfbp5 (inhibits action of Igf1 by sequestering it to the ECM⁶⁷ and suppresses nerve sprouting⁷), is consistent with previous studies of BoNT-A treatment.¹¹ Although we found no significant regulation of Igf1, upregulation of its receptor, Igf1r, may compensate for the decreased availability of Igf1. Activation of Igf1r results in phosphorylation of insulin receptor substrates (Irs1) and regulation of several downstream players such as Akt1, Pik3 (Pik3r4), and the energy/ATP availability sensor AMPK α/β/γ. The observed spike in Igf2 at 4 weeks correlates with studies showing a preferential spike in Igf2 nearly 20 days after denervation/nerve injury.⁶⁵

Although the exact role of myogenic regulatory factors⁶⁸ (MRFs) in differentiated post-mitotic skeletal muscle is not understood fully, the observed upregulation of MyoD, Myog, and Mrf4/Myf6 may reflect satellite cell activation.⁶⁹ These factors may be necessary for activating transcriptional programs required for recovery of muscle function, such as Ankrd1. Concomitant with activation of MRFs, there is upregulation of several cell-cycle control genes known to play a role in satellite cell proliferation, such as Tp53, Pcna, Myc, and cyclin-dependent kinase inhibitors Cdkn1a (suggested to confer a protective, anti-apoptotic effect⁷⁰) and Cdkn1c. Gadd45a, a marker for atrophy also involved in cell-cycle control has been identified repeatedly as being overexpressed in models of denervation/chemodenervation.^10,11,28 Recent reports have indicated that the pro-atrophy transcription factor Atf4 may induce expression of Gadd45a in muscles subject to 3 distinct skeletal muscle stresses of fasting, immobilization, and denervation.⁷¹ Its expression was shown to be necessary but not sufficient for expression of Gadd45a. However, we observed a conflicting program of regulation in our data with Atf4 being downregulated, suggesting alternate roles for Atf4 and regulation of Gadd45a in BoNT-A–treated muscle at 1 week.

Further regulatory conflicts occur through upregulation at 1 week of 4 inhibitors of DNA binding genes (Id1, Id2, Id3, and Id4), which have been shown to inhibit muscle growth and differentiation.⁷² Reduced contractile activity leads to a reduction in signaling that promotes muscle growth but inhibits complete fiber death (autophagy) triggered through pathways such as Pik3/Akt⁷³ and activation of runt transcription factor (Runx1). Runx1 has been shown to sustain muscle²⁶ under atrophic conditions by inducing expression of genes required for muscle growth and function (Myh2, Scn5a, Rrad, Myh3, and Chrng) and repressing atrophy genes (Gadd45a and Aqp4).

Later Response to Botulinum Toxin Injection

Importantly, we found that, by 4 weeks, the transcriptional events leading to muscle atrophy and weakness were essentially completed. Although the functional properties of muscle are highly impaired at this time-point, the transcriptional response is essentially complete and is in the process of recovering. This is clearly seen by the fact that, of the 1989 genes regulated after injection, only 231 were actually still changing after 4 weeks.

Recovering NMJ, Sustained Oxidative Stress, and Lack of Ca²⁺ Homeostasis

Although the muscle slows down transcriptionally, the expression of certain pathways is still significant at 4 weeks. In contrast to other genes of the NMJ, Emb and Chrna1 were still upregulated up to 12 weeks, with significant remodeling of the synaptic basal lamina (Lama5, Col4a5, and Nid2 were all upregulated) at 4 weeks. Nid2 is involved in synapse maintenance⁷⁴ and is associated selectively with the synaptic basal lamina at the NMJ. Taken together, these findings suggest reinnervation and a continuing effort by muscle toward stabilization of the NMJ at 4 weeks.

Upregulation of the SR calcium-sequestering protein Casq2 suggests an effort by muscle to maintain Ca²⁺ within the SR, possibly counteracting continued overactivation of Sln. Interestingly, we found strong upregulation of a calmodulin-like protein called Calml3 beginning at 4 weeks. Although its exact function is not yet determined in skeletal muscle, it is known to compete with cal-modulin in other tissues, further suggesting alterations in the Ca²⁺ handling properties of muscle.⁷⁵ Upregulation of certain transcription regulators of atrophy and growth, such as Myod1, Id1, Id3, Runx1, Gadd45a, Cdkn1a, and Cdkn1c, were observed until 4 weeks.

ECM Remodeling and Fibrosis

The most pronounced effect at 4 weeks was active remodeling of the ECM and possibly even fibrosis. ECM production is regulated in part by activation of several targets of TGF-β, including Ctgf and Ltbp1. Upregulation of early growth response (Egr-1), a zinc-finger transcription factor known to regulate collagen expression (particularly Col1a2) in response to TGF-β⁷⁶ and act downstream of multiple pro-fibrotic agents to regulate transcription, was observed. The persistent activation of these genes in conjunction with dramatic upregulation of ECM genes beyond 1 week (see Results and Fig. S1, online) leads us to propose activation of similar fibrotic programs by 4 weeks after BoNT-A injection, resulting in fibrosis of injected tissue. This also emphasizes the possibility of a multifaceted role of Igfbp5 after BoNT-A injection.^77–79

Significant regulation of Scd1, Sln, Cdkn1a, Cdkn1c, and S100a4 beyond 4 weeks suggested incomplete recovery of muscle, even 4 weeks after treatment. Importantly, we observed no biologically significant changes, at least transcriptionally, in skeletal muscle treated with BoNT-A at 52 weeks after injection.

In conclusion, in this analysis we have presented a global assessment of changes occurring in BoNT-A–treated muscle over a period of 1 year. By utilizing previously described physiological networks of muscle, we provided a systems-level analysis that categorically assesses expression changes after BoNT-A treatment. Dramatic transcriptional regulation in several of these networks was evident at week 1, leading to derangement of the ECM and fibrillar components by week 4. The shift toward expression of slow and immature isoforms emulating “immature” muscle possibly aids in muscle recovery. It should be noted that, although this is a transcriptional expression study, the data have clinical relevance. Specifically, they indicate that, at the molecular level, the effects of BoNT-A are relatively rapid, because most transcripts returned to control levels within 4 weeks. This is consistent with the use of the term “reversible chemodenervation,” often used with reference to the action of BoNT-A. It is also of interest to note that, in spite of the relatively fast transcriptional response, the structural and functional response lags somewhat. This is probably a function of the length of time required for a neuromuscular unit to recover from a period of denervation-induced atrophy and fibrosis Transcriptional regulation associated with atrophy and fibrosis suggests the possibility of transient extracellular effects after BoNT-A injection. Although no long-term transcriptional abnormalities were observed, further studies are necessary to determine optimal intervals for BoNT-A treatment from both a biological and physiological point of view.

Abbreviations

ACh

acetylcholine

ADP

adenosine triphosphate

ATP

adenosine triphosphate

BH

Benjamini–Hochberg

BoNT-A

botulinum neurotoxin A

CP

cerebral palsy

DHPR

dihydropyridine receptors

ECC

excitation–contraction coupling

ECM

extracellular matrix

GCRMA

gene chip robust multi-array average

GEO

Gene Expression Omnibus

IGF

insulin-like growth factor

MRF

myogenic regulatory factor

MuSK

muscle-specific tyrosine kinase

nAChR

nicotinic acetylcholine receptor

NF-kappaB

nuclear factor–κB

NMJ

neuromuscular junction

qPCR

real-time quantitative polymerase chain reaction

ROS

reactive oxygen species

RYR

ryanodine receptor

SLC

solute carrier

SR

sarcoplasmic reticulum

TA

tibialis anterior