The Role of Tumor Necrosis Factor-alpha (TNF-α) in Skeletal Muscle Regeneration

Materials and Methods

Animals

Animals were fed and housed according to guidelines of the National Health & Medical Research Council (NH&MRC) of Australia, and all procedures were approved by the Animal Experimentation Ethics Committee at the University of Western Australia. All mice, apart from the null strains, were obtained from the Animal Resource Centre (Murdoch University, Western Australia). C57/BL10ScSn mice (hereafter referred to as C57/BL) were used as controls for TNF-α null and TNF-α/LT-α null mice. The TNF-α knockout (Sloan-Kettering; New York, NY) (Marino et al. 1997) and TNF-α/LT-α (Hoffman-LaRoche; Basel, Switzerland) double-knockout mice were bred on a C57BL/129 background and were kindly donated by Hugh Reid (LaTrobe University; Victoria, Australia). Uninjured and experimentally injured adult C57/BL, SJL/J, BALB/c, TNF-α(-/-) null, and TNF-α(-/-)/LT-α(-/-) null mice were used for basic histology and immunohistochemistry. Uninjured dystrophic mdx mice were also used to model the endogenous muscle injury in DMD. Ten TNF-α(-/-) and 14 TNF-α(-/-)/LT-α(-/-) mice of a range of ages and sexes were used, as shown in Table 1. The relatively small number of these knockout mice was due to their very limited availability.

OPEN IN VIEWER

Table 1

Knockout mice: surgical procedure (crush/whole muscle graft), time of sampling after injury, sex, and body weight

	TNF-α(-/-) Mice		TNF-α(-/-)/LT-α(-/-) mice a
Sampled	Crush	Muscle graft	Crush	Muscle graft
Day 1	M 26 g; F 16 g	–	F 23 g	–
Day 3	M 25 g	F 17 g	F 20 g; M 23 g	M 26 g
Day 5	M 26 g	F 27 g	F 20 g	M 20 g
Day 7	M 26 g	M 25 g	M 25 g	F 23 g
Day 10	F 16 g	–	M 25 g	–
Day 14	–	M 26 g	–	F 22 g

^a

Four female double-knockout mice used to establish primary cultures of double-knockout muscle.

Results

Uninjured, Dystrophic, and Denervated Muscle

The histological appearance of uninjured skeletal muscle from BALB/c, C57/BL, and SJL/J mice was similar. Furthermore, no differences were noted among muscles from these mice aged 3 weeks, 10 weeks, 1 and 3 years of age (data not shown). Dystrophic mdx muscle showed variation in muscle fiber size, many mature muscle fibers with centrally placed nuclei (characteristic of regenerated mouse muscle), focal areas of necrosis with infiltrating phagocytic cells, and areas of regeneration (not shown), as has been well described elsewhere (McGeachie et al. 1993). Denervated muscle exhibited muscle fiber atrophy and central nuclei at 4 weeks, with no necrosis or inflammation, as previously described (McGeachie and Grounds 1989).

OPEN IN VIEWER

Figure 2

TNF-α (brown) staining in injured and regenerating C57/BL muscle. (A) Negative control, primary antibody omitted, transverse section through Day 3 crush-injured muscle. ^∗, central necrotic core. (B) Longitudinal section through crush-injured muscle at 3 days, showing intense staining in many cells. (C) TNF-α-stained control section of crush-injured TNF-α/LT-α null muscle, showing some nonspecific staining on the edges of muscle fibers but no staining in inflammatory cells. (D) Transverse section showing intense staining of whole EDL muscle graft and adjacent host TA muscle 3 days after transplant. (E) High-power view showing strong staining of a macrophage (m) and neutrophil (n) in Day 3 crush-injured muscle; inset, strong staining of neutrophils in Day 1 crush-injured muscle. (F) Intensely TNF-α-positive mast cells in a Day 5 graft. (G) Strong staining is seen in persisting necrotic myofibers and variable TNF-α staining in myotubes (arrows). (H) A muscle spindle in the host TA muscle of a Day 5 graft, showing TNF-α-positive staining around intrafusal muscle fibers (IF) in what appears to be the inner capsule membrane, and also in the outer spindle capsule (C). Bars = 0.1 mm.

Antibodies to TNF-α showed little staining of uninjured muscle (not shown). No differences were visible in the staining among strains or among muscle at 3 weeks, 10 weeks, 1 and 3 years of age (not shown). TNF-α staining was not seen in mast cells of uninjured muscle, although mast cells were clearly present, as demonstrated by Toluidine blue staining (Nahirney et al. 1997) (not shown). TNF-α staining was increased in regions of necrosis and inflammation (see Figure 2), and strong TNF-α staining was seen in mast cell granules in mdx muscles (similar to staining in Figure 2F).

At 1 week after denervation, TNF-α staining was similar to uninjured muscle, but at 4 weeks there was increased TNF-α staining in atrophic muscle fibers and in muscle spindles (muscle spindle staining as in Figure 2H). Muscle spindles were not stained in uninjured muscle (not shown). TNF-α staining in mast cells did not change as a result of denervation.

TNF-α Staining of Muscles Regenerating After Crush Injury or Grafting

No differences were observed between the intensity of staining with TNF-α antibodies in regenerating muscles (crushed or grafted) of BALB/c, C57/BL, and SJL/J mice. Necrotic myofibers stained strongly TNF-α-positive and the sarcoplasm of the uninjured portion of crush-injured muscle also stained TNF-α-positive (Figure 2B). Sections of single- and double-knockout muscle stained with TNF-α antibody (as a negative control) showed some staining of damaged myofibers but no staining of inflammatory cells (Figure 2C), demonstrating that although some nonspecific binding occurs, the increased TNF-α staining seen in injured TNF-α (+/+) muscles is real. The host TA muscle underlying the EDL autografts also showed strong TNF-α staining across all time points examined (Figure 2D). However, sham-operated TA muscle was TNF-α-negative, indicating that the presence of the graft is necessary to increase TNF-α expression in the host muscle (not shown).

OPEN IN VIEWER

Figure 3

TNF-α staining of cultured muscle. (A and inset) Negative control, primary antibody omitted; C57 culture showing some nonspecific staining. (B) TNF-α staining of TNF-α/LT-α double-knockout muscle, showing weak staining in adherent cells. (C) TNF-α-stained SJL/J culture, showing strongly positive myotubes. (D) TNF-α-stained C57 culture, showing staining of adherent cells and intense staining of rounded cells. Bars = 0.1 mm.

TNF-α antibodies stained all neutrophils in crush-injured muscle at Day 1 (Figure 2E, inset), and the staining intensity and number of positive neutrophils decreased after this time. Macrophages in crush-injured and grafted muscles stained strongly for TNF-α (Figure 2E). The small numbers of lymphocytes present in injured muscle and endothelial cells in small blood vessels were also TNF-α-positive.

Lymphocytes were identified as cells with a small rim of cytoplasm surrounding a dense, non-lobulated nucleus. Mast cells in injured muscle were strongly TNF-α-positive, with staining localized to mast cell granules (Figure 2F). TNF-α-positive muscle spindles were present in crush-injured muscles and, after muscle transplantation, in both the EDL graft and host TA muscle (Figure 2H). Both the inner and outer spindle capsule membranes stained TNF-α-positive [see Maier (1997) for spindle morphology], and the intrafusal muscle fibers within the spindles were variably TNF-α-positive.

Myoblasts were difficult to identify in sections of crush-injured muscle. Crescent-shaped cells, presumed to be myoblasts (activated satellite cells), were seen within the contour of the basement membrane of necrotic myofibers in grafts. These “cuffing” myoblasts were TNF-α-negative to moderately TNF-α-positive. Myotubes in crush-injured and grafted muscles ranged from TNF-α-negative to moderately TNF-α-positive, and the latter were usually larger more mature myotubes. In general, TNF-α staining was confined to the cytoplasm of myotubes (Figure 2G). The regenerated zone of the muscle usually had a much lower intensity of TNF-α staining than freshly injured and necrotic muscle.

Tissue Culture Studies

To specifically address the production of TNF-α in myoblasts and myotubes, immunostaining was also carried out on primary cultures of skeletal muscle. Muscle cultures from BALB/c, C57/BL, and SJL/J mice showed intense cytoplasmic staining with TNF-α antibodies in mononucleated cells and myotubes, often present in mononuclear cells in a granular distribution (Figures 3C and 3D). Negative controls, in which the primary antibody was omitted, showed some scattered nonspecific staining (Figure 3A). In addition, muscle cultures from TNF-α(-/-)/LT-α(-/-) double-knockout mice served as negative controls. The presence of some TNF-α staining in adherent mononuclear cells confirmed some nonspecific staining in myoblasts (Figure 3B).

OPEN IN VIEWER

Figure 4

Hematoxylin and eosin staining of single- and double-knockout muscle. (A) Longitudinal section through crush-injured TNF-α/LT-α double-knockout muscle at Day 3 after injury, showing leukocyte infiltration and necrotic muscle tissue (^∗). (B) Whole muscle graft, showing the central necrotic core (^∗) and regeneration in the graft periphery at Day 7. (C) Neutrophil infiltration of a necrotic muscle fiber, Day 1 crush. (D) Macrophages within a necrotic muscle fiber, Day 3 crush. (E) Areas containing inflammatory cells and myotubes in host TA muscle (arrows). (F) Leukocyte infiltration through muscle (arrowhead). Bars = 0.1 mm.

Discussion

Although immunostaining showed strong upregulation of TNF-α in many cells in response to skeletal muscle damage, the absence of TNF-α and its homologue LT-α had no apparent effect on the processes of regeneration and new muscle formation in TNF-α(–) and TNF-α(-/-)/LT-α(-/-) mice. An important finding was a high incidence of “bystander” damage to adjacent host muscles in response to overlying muscle grafts in the single- and double-knockout host mice. These observations are discussed in detail below.

Immunostaining of Normal and Regenerating Muscle

The lack of TNF-α expression in normal uninjured skeletal muscle indicates that, unlike in other tissues (Hunt et al. 1992), TNF-α is not involved in muscle homeostasis and is expressed only in situations of injury and inflammation. The absence of TNF-α staining in uninjured skeletal muscle is highlighted by the lack of TNF-α staining of host muscles in sham-operated mice, indicating that significant muscle disturbance or the presence of necrotic muscle in the overlying graft is required for detectable TNF-α expression in skeletal muscle.

After experimental injury, myofibers in both the directly injured and “uninjured” regions of muscle expressed TNF-α. This was seen in muscle adjacent to the crush injury and also in the “uninjured” host TA muscle underlying muscle grafts. Expression of TNF-α by damaged myofibers after acute injury has not been described previously. This observation is compatible with results showing TNF-α staining in dystrophic mdx myofibers and in muscle fibers of DMD patients (Tews and Goebel 1996), and in inflammatory myopathies (De Bleecker et al. 1999). The greater intensity and wider distribution of staining observed in the present study may be due to the relatively high antibody concentrations used (1:400 present study; 1:500 DeBleecker; 1:1000 Tews) on paraffin (rather than frozen) sections, and the acute nature of the experimental injuries. Necrotic muscle sarcoplasm stained particularly strongly for TNF-α, supporting the reported role of TNF-α in the induction of muscle proteolysis (Goodman 1991; Zamir et al. 1992; Cassatella 1995; Argiles et al. 1998), although this may be an indirect effect mediated by glucocorticoids (Ji et al. 1998). Muscle atrophy has also been implicated as a trigger for cytokine expression by muscle fibers (Tews and Goebel 1996), and expression of TNF-α was also seen in the present study in atrophic muscle fibers at 4 weeks after denervation. That staining of myotubes in vivo was due to endogenous production of TNF-α rather than to exogenous TNF-α protein binding to the cell surface is supported by the strong TNF-α staining of myotubes in tissue culture.

TNF-α staining of infiltrating inflammatory cells in injured muscle confirmed reports of TNF-α production by activated leukocytes (Vilcek and Lee 1991; Korner and Sedgwick 1996). Macrophages have previously been described as the major systemic source of TNF-α (Beutler and Cerami 1988), and this is supported by the strong TNF-α staining observed in these cells. Neutrophils are also known to produce TNF-α (Schmauder-Chock et al. 1994), and were correspondingly shown in this study to be TNF-α-positive. The decrease in the proportion of TNF-α-positive neutrophils across the time points in crush-injured muscle indicates that peak neutrophil activation occurs at Day 1 and subsequently decreases, in agreement with other studies in regenerating muscle (Grounds and Davies 1996). An alternative explanation is that different subsets of leukocytes may arrive in the muscle at different times, and these may possess different abilities to produce TNF-α. The accuracy of TNF-α staining in this study was further reinforced by the strong positive TNF-α staining seen in mast cell granules. It is well documented that TNF-α is stored within cytoplasmic granules (Malaviya et al. 1996) and is synthesized and released when mast cells are appropriately stimulated (Gordon and Galli 1991; Wedemeyer et al. 2000). We also observed that relatively few lymphocytes are present (compared with macrophages or neutrophils) in crush-injured and grafted muscles. Although lymphocytes are known to produce TNF-α (Ware et al. 1996; Tateyama et al. 1997), these cells were not present in sufficient numbers to be a major source of TNF-α in the muscle injury models described here.

Pathways of inflammatory cell recruitment, activation, and cell death differ among different types of muscle injury (Tidball 1995; Goebels et al. 1998), and this is reflected in the differences between the TNF-α staining observed in mdx muscle and in experimentally injured muscle. The relatively low TNF-α staining intensity observed in all cells except mast cells in mdx muscle indicates that the inflammatory cell activation and migration of specific subsets of cells (Tidball 1995; Spencer et al. 1997) in the two types of injury (i.e., acute experimental vs chronic) result in different levels of TNF-α production.

Muscle Spindles

TNF-α staining was observed in muscle spindles in cross-sections of injured and denervated muscle but was not seen in normal uninjured muscle. Muscle spindles are mechanoreceptors specialized to monitor changes in muscle length (Barker and Banks 1986; Prochazka and Gorassini 1998) and contain a variable number of small-diameter intrafusal muscle fibers with their attendant sensory and motor innervation, all located within a multilayered, fusiform outer capsule (Barker et al. 1974). Satellite cells are known to be present within the inner capsule of the muscle spindle (Bird 1978), and the intrafusal muscle fibers of muscle spindles are also known to be capable of regeneration in the same manner as extrafusal muscle fibers (Maier 1997). Muscle spindles show selective expression of neurotrophin-3 (Copray and Brouwer 1994), adhesion molecules (Maier and Mayne 1995), and high Pax7 (Rodger et al. 1999), but this is the first report demonstrating the presence of TNF-α. Staining of intrafusal myofibers of muscle spindles indicates that a response of both extrafusal (i.e., “normal”) and intrafusal (i.e., spindle) muscle fibers to injury is the production of TNF-α. The presence of TNF-α in the capsule cells of muscle spindles suggests that TNF-α may also play a role in regeneration of the associated neuromuscular junctions (Maier and Mayne 1995; Maier 1997).

Regeneration in TNF-α Knockout Mice

The first part of this study showed that TNF-α is highly expressed in injured and regenerating skeletal muscle. However, the absence of TNF-α had no effect on the timing and extent of inflammation and new muscle formation in regenerating skeletal muscles of null mice. No reduced macrophage activity was apparent in the TNF-α(-/-) mice, in agreement with results by Marino et al. (1997). Furthermore, in support of our observations, Marino reports that the absence of TNF-α (in null mice) has no effect on liver regeneration, based on assessment of both cell proliferation and liver weights (Marino, personal communication). These observations contrast with impaired liver regeneration reported in situations in which TNF-α is still present but its activity is blocked by antibodies (Ackerman et al. 1992; Yamada et al. 1997) or in mice deficient in the receptors for TNF-α (TNF-RI and/or TNF-RII) (Yamada et al. 1997). Knight et al. (2000) found that mice deficient in TNF-RI but not TNF-RII had impaired oval cell proliferation and reduced liver tumorigenesis when fed a carcinogenic diet. The presence of normal or high levels of TNF-α protein (in both of these situations) presumably does not stimulate compensatory upregulation of other cytokines or host cellular responses (Durbin et al. 2000). In contrast, it appears that changes in cytokine profiles (see below) probably do occur in mice completely lacking TNF-α protein as in the present study. Our in vivo studies in single- and double-knockout mice strongly indicate that any role played by TNF-α in skeletal muscle regeneration can be performed efficiently and effectively by the upregulation of other cytokines (when TNF-α is not present).

The important distinction between the markedly different outcomes (seen in both skeletal muscle and liver) when TNF-α is present but its activity is blocked either by antibodies to TNF-α itself or by blocked or absent receptors, compared with situations in which TNF-α protein is absent (and the probable consequence of an altered host cytokine response), has critical implications for the design of therapeutic strategies to modulate cytokine levels.

Cytokine Interactions

TNF-α exists within a complex network of interactions with a multitude of other cytokines. In TNF-α null mice, interleukin-12 (IL-12) expression is greatly increased (Hodge-Dufour et al. 1998). Because IL-12 and TNF-α are co-stimulators for IFN-γ, one of the essential cytokines for regulation of the inflammatory response (Storkus et al. 1998), the upregulation of IL-12 in the absence of TNF-α, could act in a compensatory manner to induce and maintain appropriate IFN-γ levels. IFN-γ induces macrophage activation, an important event in muscle regeneration (De Maeyer and De Maeyer-Guignard 1998). It also induces IL-12 (Hodge-Dufour et al. 1998) and can induce or suppress IL-1 synthesis (Schindler and Dinarello 1990). IL-1, like TNF-α, is a primary mediator of the inflammatory response (Schindler and Dinarello 1990; Fiers 1991) and is a strong candidate for potential upregulation in the absence of TNF-α. IL-1 is also likely to be important in muscle regeneration because of its roles in angiogenesis, fibroblast proliferation, and chemotaxis (Schindler and Dinarello 1990; Heckmann et al. 1993; Tidball 1995) and, like TNF-α, its direct effects on proliferation and fusion of myoblasts (Ji et al. 1998).

IL-6 expression increases in regenerating and diseased skeletal muscle (Kurek et al. 1996, 1998) and IL-6-induced protein degradation and muscle wasting may be important in DMD (Tsujinaka et al. 1998). IL-6 also stimulates myoblast proliferation (Austin and Burgess 1991) and is a candidate for upregulation in the absence of TNF-α. Another cytokine that may be of importance in muscle regeneration in the absence of TNF-α is transforming growth factor-beta (TGF-β) (Derynck and Choy 1998). In particular, TGF-β is a modulator of myoblast differentiation and fusion and in remodeling of the extracellular matrix (Murakami et al. 1999). TGF-β also attenuates the inflammatory response and inhibits the activity and/or production of a number of cytokines, including TNF-α, LT-α, and IL-1 (Derynck and Choy 1998). Like TNF-α, TGF-β is also elevated in skeletal muscle disorders (Murakami et al. 1999). However, because the muscles of TGF-β(–) mice appear normal (McLennan et al. 2000), this indicates that other genes are capable of substituting for the missing TGF-β in vivo.

The cytokines TNF-α, IL-1, IL-6, and IFN-γ can stimulate the production of glucocorticoid hormones by the pituitary and, in turn, glucocorticoids inhibit the production of these cytokines and thus exert an immunomodulatory effect (for review see Brattsand and Linden 1996). Cytokine regulation by glucocorticoids concerns not only the ligands but also signal transduction by the specific receptors.

Oligomerization of receptors for TNF-α has been reported to occur in the absence of ligand binding (Engelmann et al. 1990), thus inducing receptor signaling characteristic of TNF-α (or LT-α or LT-α2β1) without the specific ligand-receptor interaction. It is not known to what extent this might occur during injury. Such potential receptor signaling in the absence of TNF-α or LT-α ligand binding in the knockout mice must therefore also be considered as a possible contributing factor to the normal regeneration seen in null mice.

Studies in double-knockout mice indicate that, like TNF-α, LT-α has no essential role in muscle regeneration, and therefore LT-α₁β₂ and LT-α₂β₁ also have nonessential roles in the process. The similarity between regeneration in the single- and double-knockout mice indicates either that the cytokines upregulated in the absence of TNF-α are also sufficient to compensate for the lack of LT-α or that a different cytokine expression profile exists in the single-knockout mice that has essentially the same overall function.

Observations of thin, weak skin, and soft muscle with many capillaries in the TNF-α single- and double-knockout mice can be accounted for by the roles that TNF-α and other cytokines play in skin, angiogenesis and connective tissue. For example, keratinocyte proliferation is decreased and differentiation increased by TNF-α (Vassalli 1992), and TNF-α induces expression of cell adhesion molecules, such as ICAM on the surface of keratinocytes (Matsuura et al. 1998). Therefore, the anomalies observed in the skin of TNF-α knockout mice could be due to deregulation of keratinocyte proliferation, differentiation, and adhesion. TNF-α also stimulates fibroblast proliferation in culture (Vilcek et al. 1986), so the lack of TNF-α might result in decreased fibroblast proliferation and thus reduced connective tissue density. Because matrix metalloproteinases (MMPs) are involved in matrix remodeling during wound healing, angiogenesis, and tumor metastasis (for review see Forget et al. 1999) and are increased in inflammatory states such as chronic wounds (Wysocki et al. 1999), and because MMP2 and pro-MMP2 production by cultured human skin and isolated dermal fibroblasts is activated by TNF-α (Han et al. 2000), this potentially introduces further changes to the regulation of the skin and connective tissue in these mice.

Stress, Glucocorticoids, and TNF-α

The hyperactive behavior and highly stressed appearance of the single-knockout mice may be accounted for by the complex cytokine interactions in vivo, and the strong interaction of TNF-α with pituitary secretion of glucocorticoid hormones and the negative feedback interaction existing between them (as previously discussed). Upregulation of IL-1 may further exacerbate this situation because IL-1 is an important player in a number of the physical manifestations of stress (Nguyen et al. 1998) and administration of IL-1 produces stress-related changes (Shintani et al. 1995).

“Bystander Effect” Damage to the Muscle of Knockout Mice

The unexpected presence of myotubes throughout the host muscle after graft implantation is not a feature of normal muscle. Although small areas of myotube formation are sometimes seen on the adjacent host muscle, this is characteristically restricted to the surface of the host TA muscle lying immediately beneath the graft and is considered to be the result of direct physical damage from the surgery. The presence of myotubes throughout the majority of host muscle sections beneath the graft in both single- and double-knockout mice, and their absence from both local and distant muscle at Day 1, strongly supports the idea that such myotubes are not normally present in TNF-α null mice but are a consequence of the presence of the overlying graft. Bystander host muscle damage has also been observed in mouse soleus muscles after implantation of MHC-matched myoblasts (Wernig and Irintchev 1995), and this was attributed to locally released cytokines such as TNF-α.

The conspicuous number of leukocytes seen between host myofibers in knockout mice suggests that potential alterations in the absolute numbers or subtypes of inflammatory cells present in the double-knockout muscle could result from loss of the normal Type 1 inflammatory response with an increased or abnormal Type 2 inflammatory response (Durbin et al. 2000). Alternatively (as discussed above), an altered cytokine profile within the knockout mice may increase the toxicity of these inflammatory cells, resulting in bystander damage to the muscle fibers in their immediate vicinity. Because cytokine levels and inflammatory pathways are linked by regulatory mechanisms, a combination of these two possibilities may be responsible. Further investigation is under way in our laboratory into the cause of the observed bystander effect in the knockout mice. This bystander effect in knockout mice causing muscle damage and subsequent myotube formation has potentially important clinical implications with respect to various muscle diseases (such as myositis). A recent study of the pathology of X-linked muscular dystrophy in the mdx mouse in the absence of TNF-α revealed a complicated picture, with increased pathology of diaphragm muscle at 4 weeks of age but no effect on diaphragm muscle and decreased pathology of quadriceps muscles at 8 weeks in TNF-α(-/-)/mdx mice (Spencer et al. 2000). Although it was predicted that the absence of TNF-α would ameliorate the dystrophic process, this was not supported by the complex results. The compensatory role of other cytokines in the absence of TNF-α probably accounted for this. In addition, there may be a sustained bystander effect in the TNF-α(-/-)/ mdx mice. We propose that the alternative strategy of administering antibodies to block normal TNF-α activity (Ackerman et al. 1992; Illei and Lipsky 2000) might produce quite different results and reduce the pathology of the muscular dystrophy in the mdx mice (due to alterations to other cytokines as discussed above), and this approach has the major advantage of potential clinical application.