How does bones and muscles work together

The heart is made of special muscle called cardiac muscle. Joints connect bones. They provide stability to the skeleton, and allow movement. There are different types of joints. Joints in the arms and legs are synovial joints. The ends of the bones are covered with cartilage and separated by the joint cavity which is filled with a thick gel called synovial fluid. Synovial fluid helps to lubricate the cartilage and provides nourishment to it.

Ligaments stretch across the joint, connecting one bone to another and help to stabilise the joint so it can only move in certain directions. Joints in the spine and pelvis and the joints between the ribs and the sternum are cartilaginous joints — they provide more stability but not as much movement.

The bones are connected by cartilage in this type of joint. Fibrous joints allow no movement — just stability. They are held together by fibrous connective tissue. You have fibrous joints in your skull. They all have different forms of treatment. The best way to have healthy bones and prevent illness and injury to the bones is to eat a healthy diet that includes calcium-rich foods, limit soft drinks, caffeinated drinks and alcohol, be as active as you can, do weight bearing and high impact activities if you can, get enough sunshine and keep to a healthy weight.

Many conditions can affect the joints. Arthritis, which is characterised by joint pain and stiffness, is one of the most common. Different types of arthritis have different causes. Muscle injuries and disorders can cause weakness, pain or paralysis.

Sports injuries are a common way that muscles can be damaged. Conditions affecting the muscles include:. If you are having problems with any part of your musculoskeletal system, your doctor GP is a good place to start. Other healthcare professionals who are involved in diagnosing and treating musculoskeletal problems are physiotherapists and specialists such as rheumatologists or sports medicine physicians.

Learn more here about the development and quality assurance of healthdirect content. Exercise can help reduce the symptoms of arthritis and musculoskeletal conditions. Learn more about exercise, and where to find support. Call Read more on Musculoskeletal Australia website. Do you have painful feet due to arthritis, gout or bunions? Driving can be painful and exhausting if you have arthritis or a musculoskeletal condition, but there are things you can do to improve this. Find out more.

Does your child have juvenile arthritis? Do you have ankylosing spondylitis? Find out more about your musculoskeletal condition, how you can manage it, and where to find support.

Call MSK: Do you have Osteoporosis? Find out more about your musculoskeletal condition, how you can manage, and where to find support. Contact us today on Do you have Reactive Arthritis? Find out more about your musculoskeletal condition, how to manage, and where to find support. Call us today on Do you have bursitis? Employing this model, we calculated the network level effects of perturbing individual muscles.

Using this formalism, we are able to draw new parallels between this system and the primary motor cortex that controls it, and illustrate clinical connections between network structure and muscular injury.

PLoS Biol 16 1 : e This is an open access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability: All relevant data are within the paper and its Supporting Information files. The two musculoskeletal graphs used, as well as muscle community assignments, and data used to generate all figures can be found at DOI: The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist. The interconnected nature of the human body has long been the subject of both scientific inquiry and superstitious beliefs. From the ancient humors linking heart, liver, spleen, and brain with courage, calm, and hope [ 1 ] to the modern appreciation of the gut—brain connection [ 2 ], humans tend to search for interconnections between disparate parts of the body to explain complex phenomena.

Yet, a tension remains between this basic conceptualization of the human body and the reductionism implicit in modern science [ 3 ]. An understanding of the entire system is often relegated to a futuristic world, while individual experiments fine-tune our understanding of minute component parts. The human musculoskeletal system is no exception to this dichotomy. While medical practice focuses in hand, foot, or ankle, clinicians know that injuries to a single part of the musculoskeletal system necessarily impinge on the workings of other even remotely distant parts [ 4 ].

An injury to an ankle can alter gait patterns, leading to chronic back pain; an injury to a shoulder can alter posture, causing radiating neck discomfort. Understanding the fundamental relationships between focal structure and potential distant interactions requires a holistic approach. Here, we detail such an approach. Our conceptual framework is motivated by recent theoretical advances in network science [ 5 ], which is an emerging discipline built from an ordered amalgamation of mathematics specifically, graph theory [ 6 ] and physics specifically, statistical mechanics [ 7 ] , computer science, statistics [ 8 ], and systems engineering.

The approach simplifies complex systems by delineating their components and mapping the pattern of interactions between those components [ 9 ]. This representation appears particularly appropriate for the study of the human musculoskeletal system, which is composed of bones and the muscles that link them. In this study, we used this approach to assess the structure, function, and control of the musculoskeletal system. The use of network science to understand the musculoskeletal system has increased in recent years [ 10 ].

However, the framework has largely been employed to investigate the properties of local muscle or bone networks. For example, the local structure of the skull has been examined to investigate how bones can be categorized [ 11 ]. Additionally, studies of the topology of the musculoskeletal spine network have been conducted to evaluate stresses and strains across bones [ 12 ].

A few studies do exist that address the entire musculoskeletal system, although they do not use the mathematical tools that we employed here [ 13 , 14 ]. The current study differs from previous work in its assessment of the entire musculoskeletal system combined with the mathematical tools of network science. Within this broader context, we focused on the challenge of rehabilitation following injury to either skeletal muscle or cerebral cortex.

Direct injury to a muscle or associated tendon or ligament affects other muscles via compensatory mechanisms of the body [ 15 ]. Similarly, loss of use of a particular muscle or muscle group from direct cortical insult can result in compensatory use of alternate muscles [ 16 , 17 ].

How the interconnections of the musculoskeletal system are structured and how they function directly constrains how injury to a certain muscle will affect the musculoskeletal system as a whole.

Understanding these interconnections could provide much needed insight into which muscles are most at risk for secondary injury due to compensatory changes resulting from focal injury, thereby informing more comprehensive approaches to rehabilitation.

Additionally, an understanding of how the cortex maps onto not only single muscles but also groups of topologically close muscles could inform future empirical studies of the relationships between focal injuries including stroke to motor cortex and risk for secondary injury.

Using the Hosford Muscle tables [ 18 ], we constructed a musculoskeletal hypergraph by representing bones several of these are actually ligaments and tendons as nodes and muscles as hyperedges linking those nodes muscle origin and insertion points are listed in S9 Table. This hypergraph can also be interpreted as a bipartite network, with muscles as one group and bones as the second group Fig 1a.

This hypergraph representation of the body eliminates much of the complexity from the musculoskeletal system, encoding only which muscles attach to which bones. All analysis was applied to only one half left or right of the body, because each cerebral hemisphere controls only the contralateral side of the body.

Therefore, we further simplified our model by assuming left—right symmetry; in any figures in which both halves of the body are shown, the second half is present purely for visual intuition. The bone-centric graph A and muscle-centric graph B Fig 1b are simply the one-mode projections of C.

Then, the diagonal elements were set equal to zero, leaving us with a weighted adjacency matrix [ 5 ]. We obtained estimated anatomical locations for the center of mass of each muscle and bone by examining anatomy texts [ 19 ] and estimating x-, y-, and z-coordinates for mapping to a graphical representation of a human body Fig 1c.

To measure the potential functional role of each muscle in the network, we used a classical perturbative approach. To maximize simplicity and the potential for fundamental intuitions, we modeled the musculoskeletal system as a system of point masses bones and springs muscles. We stretched a muscle-spring and observed the impact of this perturbation on the locations of all other muscles. Physically, to perturb a muscle, we displaced all bones connected to that muscle by the same amount and in the same direction, stretching the muscle, and we held these bones fixed at their new location.

This process is also mathematically equivalent to simply altering the spring constant attributed to the particular muscle-spring. The system was then allowed to reach equilibrium. We fixed bones at the midline and around the periphery in space to prevent the system from drifting. Here, we have set all bones to have equal weight and all muscles to have equal spring constant, which is a simplification of the actual physical anatomy. For a discussion of how to account for additional physical properties, such as bone weight and muscle strength, and supplementary results using these properties, see S5 Text.

Moreover, sample trajectories that provide an intuition for the dynamics of our model have been included in the Supporting information S8 Fig. To measure the potential functional role of each muscle in the network, we stretched a muscle hyperedge and measured the impact of the perturbation on the rest of the network. Rather than perturbing the network in some arbitrary three-dimensional direction, we extended the scope of our simulation into a fourth dimension.

When perturbing a muscle, we displaced all of the nodes bones contained in that muscle hyperedge by a constant vector in the fourth dimension and held them with this displacement Fig 1d.

The perturbation then rippled through the network of springs in response. We sequentially stretched each muscle hyperedge and defined the impact score of this perturbation to be the total distance moved by all nodes in the musculoskeletal network from their original positions.

The displacement value is the summed displacement over all time points, from perturbation onset to an appropriate cutoff for equilibration time.

Here, we solved for the equilibrium of the system by allowing dynamics to equalize over a sufficient period of time. Note that the equilibrium can also be solved for using a steady-state, nondynamic approach; we chose to use dynamics in this instance to more broadly support future applications. We begin by constructing a null model that dictates the expected impact under a set of statistical assumptions. In the current study, we used several different null models with differing sets of assumptions, which we detail in later sections.

In this way, we calculated deviation from the expected value, in standard deviations similar to a z-score. Muscles can be naturally grouped according to the homunculus, a coarse one-dimensional representation of how the control areas of muscles group onto the motor cortex.

For a given homunculus group, we calculated the deviation ratio as the number of muscles with positive deviation divided by the total number of muscles in the group Table 2. The muscles on the left side have less impact than expected, given their hyperedge degree: their impacts are more than 1. The muscles on the right side have more impact than expected given their hyperedge degree: their impacts are more than 1.

This table shows the muscles that had the greatest positive and greatest negative difference in impact, relative to degree-matched controls. Categories on the left are composed entirely of muscles with less impact than expected, compared to degree-matched controls. Categories on the right are composed entirely of muscles with more impact than expected, compared to degree-matched controls. To understand both the function and control of the musculoskeletal system, we were interested in defining groups of densely interconnected muscles using a data-driven approach.

By maximizing Q, we obtained a partition of nodes muscles into communities such that nodes within the same community were more densely interconnected than expected in a network null model Fig 1b , right. Here, we also used a resolution parameter to tune the size and number of communities detected such that the number of communities detected matched the number of groups within the homunculus, for straightforward comparison.

We began by redefining the original muscle-centric matrix B following Jutla et al. The above method of community detection is nondeterministic [ 23 ]. That is, the same solution will not be reached on each individual run of the algorithm. Therefore, one must ensure that the community assignments used are a good representation of the network and not just a local maximum of the landscape. We therefore maximized the modularity quality function times, obtaining different community assignments.

From this set of solutions, we identified a robust representative consensus community structure [ 24 ]. S1 Fig illustrates how the detected communities change as a function of the resolution parameter for the muscle-centric network. We use rewired graphs as a null model against which to compare the empirical data. Specifically, we constructed a null hypergraph by rewiring muscles that are assigned the same category Table 3 , defined below uniformly at random.

In this way, muscles of the little finger will only be rewired within the little finger, and similarly for muscles in other categories. Importantly, this method also preserves the degree of each muscle as well as the degree distribution of the entire hypergraph. Categories were assigned to muscles such that the overall topology of the musculoskeletal system was grossly preserved, and changes were spatially localized. Specifically, we partitioned the muscles into communities of roughly size 3, such that each muscle was grouped with the two muscles that are most topologically related.

We then permuted only within these small groups. This is a data-driven way of altering connections only within very small groups of related muscles.

To partition muscles into communities, we took a greedy approach to modularity maximization, similar to prior work [ 25 ]. Furthermore, where K indicates the total number of communities. This term penalizes determining a set of communities that are highly unequal in size.

To conduct multidimensional scaling MDS on the muscle-centric network, the weighted muscle-centric adjacency matrix was simplified to a binary matrix all nonzero elements set equal to 1.

From this data, a distance matrix D was constructed, the elements D ij of which are equal to the length of the shortest path between muscles i and j, or are equal to 0 if no path exists.

To construct the binary matrix, a threshold of 0 was set, and all values above that threshold were converted to 1. However, to make analysis robust to this choice, we explored a range of threshold values to verify that results are invariant with respect to threshold.

The upper bound of the threshold range was established by determining the maximal value that would maintain a fully connected matrix; otherwise, the distance matrix D would have entries of infinite weight. In our case, this value was 0. Within this range of thresholds i. As a supplementary analysis, we also employed a method of constructing a distance matrix from a weighted adjacency matrix in order to preclude thresholding S5 Fig , and we again observed qualitatively consistent results.

We calculated the correlation between impact score and muscle injury recovery times. The recovery times and associated citations, listed in Table 4 , are average recovery times gathered from population studies. If the literature reported a range of different severity levels and associated recovery times for a particular injury, the least severe level was selected.

If the injury was reported for a group of muscles rather than a single muscle, the impact score deviation for that group was averaged together. Data points for muscle groups were weighted according to the number of muscles in that group for the purpose of the linear fit. We calculated the correlation between impact score deviation and the area of somatotopic representation devoted to a particular muscle group.

The areas of representation were collected from two separate sources [ 38 , 39 ]. The volumes and associated citations are listed in Table 5. In both studies, subjects were asked to articulate a joint repetitively, and the volumes of the areas of primary motor cortex that underwent the greatest change in BOLD signal were recorded.

We then calculated the correlation coefficient between cortical volumes and the mean impact of all muscles associated with that joint, as determined by the Hosford Muscle tables. To examine the structural interconnections of the human musculoskeletal system, we used a hypergraph approach. Drawing from recent advances in network science [ 5 ], we examined the musculoskeletal system as a network in which bones network nodes are connected to one another by muscles network hyperedges.

A hyperedge is an object that connects multiple nodes; muscles link multiple bones via origin and insertion points. The degree, k, of a hyperedge is equal to the number of nodes it connects; thus, the degree of a muscle is the number of bones it contacts. For instance, the trapezius is a high-degree hyperedge that links 25 bones throughout the shoulder blade and spine; conversely, the adductor pollicis is a low-degree hyperedge that links 7 bones in the hand Fig 2a and 2b.

High-degree hyperedges are most heavily concentrated at the core. Data available for e at DOI : The representation of the human musculoskeletal system as a hypergraph facilitates a quantitative assessment of its structure Fig 2c.

We observed that the distribution of hyperedge degree is heavy-tailed: most muscles link 2 bones, and a few muscles link many bones Fig 2d and 2e. To probe the functional role of muscles within the musculoskeletal network, we employed a simplified model of the musculoskeletal system and probed whether the model could generate useful clinical correlates. We implemented a physical model in which bones form the core scaffolding of the body, while muscles fasten this structure together.

Each node bone is represented as a mass, whose spatial location and movement are physically constrained by the hyperedges muscles to which it is connected. Next, we perturbed each of muscles in the body and calculated their impact score on the network see Materials and methods and Fig 1c and 1d. As a muscle is physically displaced, it causes a rippling displacement of other muscles throughout the network.

The impact score of a muscle is the mean displacement of all bones and indirectly, muscles resulting from its initial displacement. Muscles with a larger number of insertion and origin points have a greater impact on the musculoskeletal system when perturbed than muscles with few insertion and origin points [ 42 ].

In S11 Fig , we show that the network function as measured by the impact score was significantly correlated with the average shortest path length. While the network statistics are static in nature, their functional interpretation is provided by the perturbative simulations of system dynamics. Data available at DOI : To guide interpretation, it is critical to note that the impact score, while significantly correlated with muscle degree, is not perfectly predicted by it Fig 3a.

Instead, the local network structure surrounding a muscle also plays an important role in its functional impact and ability to recover. To better quantify the effect of this local network structure, we asked whether muscles existed that had significantly higher or significantly lower impact scores than expected in a null network. We defined a positive negative impact score deviation that measures the degree to which muscles are more less impactful than expected in a network null model see Materials and methods.

This calculation resulted in a metric that expresses the impact of a particular muscle, relative to muscles of identical hyperedge degree in the null model. In other words, this metric accounts for the complexity of a particular muscle Table 1. Is this mathematical model clinically relevant? Does the body respond differently to injuries to muscles with higher impact score than to muscles with lower impact score?

To answer this question, we assessed the potential relationship between muscle impact and recovery time following injury. Specifically, we gathered data on athletic sports injuries and the time between the initial injury and return to sport. We note that it is important to consider the fact that recovery might be slower in a person who is requiring maximal effort in a performance sport, compared to an individual who is seeking only to function in day-to-day life.

In order to generalize our findings to the entire population, we therefore also examined recovery time data collected from nonathletes, and we present these complementary results in the Supporting information S6 Text. Finally, to provide intuition regarding how focal injury can produce distant effects potentially slowing recovery, we calculated the impact of the ankle muscles and determined which other muscles were most impacted.

That is, for each individual ankle muscle, we calculated the impact on each of the remaining non-ankle muscles and then averaged this over all ankle muscles. Out of the non-ankle muscles, the single muscle that is most impacted by the perturbation of ankle muscles is the biceps femoris of the hip, and the second most impacted is the vastus lateralis of the knee.

Additionally, the muscle most impacted by perturbation to hip muscles is the soleus. What is the relationship between the functional impact of a muscle on the body and the neural architecture that affects control? Here, we interrogate the relationship between the musculoskeletal system and the primary motor cortex.

We examined the cerebral cortical representation map area devoted to muscles with low versus high impact by drawing on the anatomy of the motor strip represented in the motor homunculus [ 43 ] Fig 4a , a coarse one-dimensional representation of the body in the brain [ 44 ].

We observed that homunculus areas differentially control muscles with positive versus negative impact deviation scores Table 2. To probe this pattern more deeply, for each homunculus area, we calculated a deviation ratio as the percent of muscles that positively deviated from the expected impact score i. Data points are sized according to the number of muscles required for the particular movement. The plot is numbered as follows, corresponding to Table 5 : thumb 1 , index finger 2 , middle finger 3 , hand 4 , all fingers 5 , wrist 6 , elbow 7.

Activation volumes are defined as voxels that become activated defined by blood-oxygen-level-dependent signal during movement [ 38 , 39 ]. Again, impact deviation is a metric that accounts for the hyperedge degree of a particular muscle and is relative to the impact of muscles with identical hyperedge degree in the null model.

Thus, the impact deviation measures the local network topology beyond simply the immediate connections of the muscle in question.

As a final test of this relationship, we asked whether the neural control strategy embodied by the motor strip is optimally mapped to muscle groups. We constructed a muscle-centric graph by connecting two muscles if they touch on the same bone Fig 1c , left. We observed the presence of groups of muscles that were densely interconnected with one another, sharing common bones. We extracted these groups using a clustering technique designed for networks [ 45 , 46 ], which provides a data-driven partition of muscles into communities Fig 1b , right.

To compare the community structure present in the muscle network to the architecture of the neural control system, we considered each of the 22 categories in the motor homunculus [ 18 ] as a distinct neural community and compared these brain-based community assignments with the community assignments obtained from a data-driven partition of the muscle network. For example, the triceps brachii and the biceps brachii belong to the same homuncular category, and we found that they also belong to the same topological muscle network community.

Next, because the homunculus has a linear topological organization, we asked whether the order of communities within the homunculus Table 3 was similar to a data-driven ordering of the muscle groups in the body, as determined by MDS [ 48 ]. From the muscle-centric network Fig 1b , we derived a distance matrix that encodes the smallest number of bones that must be traversed to travel from one muscle to another. An MDS of this distance matrix revealed a one-dimensional linear coordinate for each muscle, such that topologically close muscles were close together and topologically distant muscles were far apart.

Our results from Fig 4d demonstrate a correspondence between the topology of the homunculus and a data-driven ordering of muscles obtained by considering the topological distances between them. This result could be interpreted in one of two ways: one reasonable hypothesis is that because most connections in the musculoskeletal network are short range, the finding is primarily driven by short-range connections. A second reasonable hypothesis is that while short-range connections are the most prevalent, long-range connections form important intramodular links that help determine the organization of the network.

To arbitrate between these two hypotheses, we considered two variations of our MDS experiment: one including only connections shorter than the mean connection length and the other including only connections longer than the mean connection length.

Notably, including both long and short connections leads to a stronger correlation with homuncular topology than considering either independently, suggesting a dependence on connections of all lengths. It would be interesting in the future to test the degree to which this network-to-network map is altered in individuals with motor deficits or changes following stroke.

By representing the complex interconnectivity of the musculoskeletal system as a network of bones represented by nodes and muscles represented by hyperedges , we gained valuable insight into the organization of the human body. The study of anatomical networks using similar methods is becoming more common in the fields of evolutionary and developmental biology [ 10 ]. However, the approach has generally been applied only to individual parts of the body—including the arm [ 49 ], the head [ 11 ], and the spine [ 12 ]—thereby offering insights into how that part of the organism evolved [ 50 , 51 ].

Moreover, even when full body musculature [ 13 ] and the neuromusculoskeletal [ 14 ] system more generally have been modeled, some quantitative claims can remain elusive, in large part due to the lack of a mathematical language in which to discuss the complexity of the interconnection patterns.

In this study, we offer an explicit and parsimonious representation of the complete musculoskeletal system as a graph of nodes and edges, and this representation allowed us to precisely characterize the network in its entirety. When modeling a system as a network, it is important to begin the ensuing investigation by characterizing a few key architectural properties. We observed that the degree distribution of the musculoskeletal system is significantly different from that expected in a null graph Fig 2e , displaying fewer high-degree nodes and an overabundance of low-degree nodes.

The discrepancy between real and null model graphs is consistent with the fact that the human musculoskeletal system develops in the context of physical and functional constraints that together drive its decidedly nonrandom architecture [ 53 ]. The degree distribution of this network displays a peak at approximately degree two, that is then followed by a relatively heavy tail of high-degree nodes. The latter feature is commonly observed in many types of real-world networks [ 54 ], whose hubs may be costly to develop, maintain, and use [ 55 , 56 ] but play critical roles in system robustness, enabling swift responses [ 55 ], buffering environmental variation [ 57 ], and facilitating survival and reproduction [ 58 ].

By contrast, there are only a few muscles that require a high degree to support highly complex movements, such as maintaining the alignment and angle of the spinal column by managing the movement of many bones simultaneously.

These expected findings provide important validation of the model as well as offer a useful visualization of the musculoskeletal system. The musculoskeletal network is characterized by a particularly interesting property that distinguishes it from several other real-world networks: the fact that it is embedded into three-dimensional space [ 59 ].

This property is not observed in semantic networks [ 60 ] or the World Wide Web [ 61 ], which encode relationships between words, concepts, or documents in some abstract and very likely non-euclidean geometry.

In contrast, the musculoskeletal system composes a volume, with nodes having specific coordinates and edges representing physically extended tissues. To better understand the physical nature of the musculoskeletal network, we examined the anatomical locations of muscles with varying degrees Fig 2c. Indeed, muscles at these locations must support not only flexion and extension but also abduction, adduction, and both internal and external rotation. It is important to note that significant variation exists within the musculoskeletal system across individuals, and not all anatomical atlases agree on the most representative set of insertion and origin points.

The results presented here reflect how the musculoskeletal system was presented in the text from which it was constructed [ 19 ] and therefore provide only one possible network representation of the musculoskeletal system. To assess the reliability of our results across reasonable variation of the musculoskeletal configuration, we created a second musculoskeletal network from an alternate atlas [ 64 ].

Using this second atlas, we observed consistent results, and we report these complementary analyses in S3 Text. It is also important to note that we mapped the first atlas [ 19 ] into a musculoskeletal graph composed of both bony and non-bony nodes.

This choice equates the structural roles of bones and certain tendons and ligaments, which is admittedly a simplification of the biology. One justification for this simplification is that non-bony structures frequently serve as critical attachment points of muscles i. The state-of-the-art prostheses have limited durability, wearing out quickly, particularly in young or active individuals. Current research is focused on the use of new materials, such as carbon fiber, that may make prostheses more durable.

Learning Objectives Summarize the structure and role of the musculoskeletal system. Key Points The skeleton, muscles, cartilage, tendons, ligaments, joints, and other connective tissues are all part of the musculoskeletal system, which work together to provide the body with support, protection, and movement. The muscles of the muscular system keep bones in place; they assist with movement by contracting and pulling on the bones. To allow motion, different bones are connected by joints which are connected to other bones and muscle fibers via connective tissues such as tendons and ligaments.

Malnutrition and arthritis are examples of disorders and diseases in the body that can severely impair the function of the musculoskeletal system. The Musculoskeletal System The musculoskeletal system provides support to the body and gives humans and many animal species the ability to move.