Closed circulatory systems underlie the function of vertebrate organs, but in long bones their structure is unclear although they constitute the exit route for bone marrow (BM) leukocytes. To understand neutrophil migration from BM, we studied the vascular system of murine long bones. Here, in a mouse model, we show that hundreds of capillaries originate in BM, traverse cortical bone perpendicularly along the shaft and connect to the periosteal circulation. Structures similar to these trans-cortical vessels (TCVs) also exist in human limb bones. TCVs express arterial or venous markers and transport neutrophils. Furthermore, over 80% of arterial and 59% of venous blood passes through TCVs. Genetic and drug-mediated modulation of osteoclast count and activity leads to substantial changes in TCV numbers. In a murine model of chronic arthritic bone inflammation, new TCVs develop within weeks. Our data indicate that TCVs are a central component of the closed circulatory system in long bones and may represent an important route for immune cell export from BM.
The function of any vertebrate organ is dependent on effective blood circulation. Arterial blood rich in oxygen and nutrients enters the organ, typically through large supplying vessels, and is then distributed via arterioles of gradually decreasing diameter down to the finest capillaries that allow diffusion of gases and nutrients to all cells in the organ1,2,3. Capillaries are the transit zones from the arterial to the venous system. The latter system collects blood from small-diameter venules and transfers it into veins of ever-increasing calibre, finally leaving the organ in a deoxygenated and nutrient-depleted condition by exiting vessels4.
Bones are responsible for the provision of structural support for the entire body, and for the attachment of tendons and muscles that allow movement5. To achieve this task, bones maintain a hard shell consisting of a composite material. This is based on a protein-rich structural framework, mainly of type I collagen6, that is embedded with nanoglobules of the inorganic mineral hydroxyapatite7,8,9. On its exterior surface, bone is covered with the highly vascularized periosteum which is connected to the general body circulation10.
Long bones have a large internal cavity lined by the inner bone surface, the endosteum11, and filled with BM, a highly vascularized tissue12,13. BM contains haematopoietic stem cells (HSC14,15) and requires extensive blood supply to mediate the transport of oxygen16, nutrients and signalling molecules. Mature immune cells17,18 and erythrocytes, both formed from dividing HSC, as well as platelets derived from BM-based megakaryocytes, must be able to migrate rapidly from the BM and reach the general circulation19. To achieve this, a very effective communication between the BM vascular system and external circulation must exist, a fact that has been exploited in emergency medicine for many years. Originally developed for battlefield administration of fluids and analgesics20, the use of direct intra-osseous infusion is now widely utilized in emergency medicine when peripheral venous access is difficult21. Thereby, intra-osseous infusions show pharmacokinetics that are indistinguishable from those of intravenous injections22. Also, intratibial injections in mice are rapidly distributed systemically23.
When measuring the migration of BM neutrophil granulocytes, the most abundant innate immune cell in humans and mice24, we found that just minutes after the intravenous application of granulocyte colony-stimulating factor (G-CSF), a haematopoietic cytokine that can mobilize movement of neutrophils and HSC into the general circulation25, neutrophils began to migrate rapidly. They then entered BM blood vessels, apparently flowing towards the endosteal surface17,26. A few minutes later those neutrophils were found in the general circulation. Our observations also indicated that there must be very effective blood exchange between the external circulation and the microvasculature at the endosteal surface. However, the physical basis of this rapid transport system in long bones remains largely elusive. A recent study has demonstrated bone canals in the calvaria that contain blood vessels and allow the transit of neutrophils. The presence of similar canals in the tibia was also reported, but the extent to which such canals contribute to overall blood flow in the bone remained unclear and a differentiation between venous and arterial flow was not performed27.
In general, the hard outer shell of bones creates a barrier for blood vessels entering from the outside, as each vessel requires a transit canal through the composite material of the bone shaft. Current models of blood flow in murine bones describe arterial inflow from one or several nutrient arteries that enter the bone either at the epiphyses or midway along the shaft5,28 and link to capillaries within the BM. Via transitional type H vessels, arterial capillaries connect to the venous tree29,30 that empties into a large central sinus31. The exit sites of the central sinus are not well described in the mouse, while in the guinea pig there appears to be just one exit32. In summary, existing models describe only a very limited set of arterial entries and venous exits in the long bones of mice for an otherwise extremely dense BM vascular bed. It is difficult to comprehend the highly effective and rapid blood exchange between BM and the general circulation based simply on these few vascular connections.
Given the limited knowledge about the way in which blood enters and leaves long bones, we studied the distribution of blood vessels in murine long bones using a variety of imaging techniques that have either only recently become available or have not yet been utilized in this field, specifically light-sheet fluorescence microscopy (LSFM33) and X-ray microscopy (XRM)34. Here we show that murine long bones are supplied by approximately 16 nutrient arteries and a central sinus with two exit sites. In addition to these known structures, we discovered hundreds of capillaries along the entire bone shaft that cross the cortical bone perpendicularly and form a direct connection between the endosteal and periosteal circulations. These TCVs can be either arterioles or venules and effectively transport blood and thereby also neutrophils. Strikingly, >80% of arterial and 59% of venous blood flow in long bones travels via TCVs, while nutrient arteries and the central sinus play only minor roles in total volume flow. Diseases that affect bone physiology lead to substantial changes in TCV numbers. As a result, osteoclasts were found at branching points within existing TCVs and their genetic or drug-mediated modulation potently influenced total TCV numbers. We also found evidence for a similar direct connection between BM and peripheral circulation in human bones. Thus, in long bones, TCVs form the mainstay of blood circulation and constitute the missing link in the search for a fully functional closed circulatory system that is able to explain the well-known phenomena of bone haemodynamics.
Murine bones contain hundreds of TCVs
Whole-mount preparations of murine femurs, prepared without perfusion following sacrifice, showed a conspicuous pattern of blood-filled puncta along the entire diaphysis (Fig. 1a,b). Field emission scanning electron microscopy (FESEM) identified a matching pattern of indentations on the bone surface (Fig. 1c) appearing as holes of diameter 10–20 µm (Fig. 1d,e). Three-dimensional (3D) XRM of tibiae demonstrated that such holes were indeed canals crossing the entire compact bone and exiting in dints along the endosteal surface (Fig. 1f–h, Supplementary Fig. 1; Supplementary Videos 1 and 2). Using simpleCLEAR33, a method that enables the optical clearing of bones containing intact BM (Supplementary Fig. 2), we generated transparent bones from non-perfused mice. In these we detected hundreds of blood-filled vessels crossing the entire bone shaft between BM and the bone surface (Fig. 1i,j). Next, we used LSFM35 to reconstruct the 3D structure of the entire murine tibial blood vessel system that had been injected with anti-CD31-AlexaFluor 647 prior to sacrifice. Blood vessels running within the periosteum and on the bone surface, including a nutrient artery, were tracked to the BM cavity where they arborized into the sinusoidal network (Fig. 1k,l; Supplementary Videos 3 and 4). Higher magnification of midline sections through the marrow cavity demonstrated a tight network of vessels that was particularly dense at the endosteal surface (Fig. 1l). Importantly, individual vessels were observed originating at the endosteum, crossing the entire diaphyseal cortical bone (CB) and connecting to the periosteal circulation on the exterior (Fig. 1l; Supplementary Video 5). Although such structures are numerous in murine long bones, we found no thorough characterization of these structures in the recent literature5,27,31,36—only the suggestion of comparable structures in much larger guinea pig bones32. Thus, we coined the term trans-cortical (blood) vessels (TCVs) to describe this vascular system.
TCVs can be either arterial or venous and directly connect the periosteum to BM
CD31/Sca-1 double staining revealed that TCVs showed either arterial or venous features and formed continuous networks with vessels within the BM (Fig. 2a–d; Supplementary Videos 5 and 6). Next, we investigated how the large nutrient arteries of the bone connected to the complex intra-BM sinusoidal network. LSFM demonstrated several entry sites for nutrient arteries along the bone shaft, and two exit sites of the central collecting vein/sinus at the upper and lower ends of the tibia (Fig. 2a–d; Supplementary Fig. 3). This analysis also indicated a continuous network of nutrient arteries from the endosteal sinusoidal network to the venous system (Fig. 2c,d). Fine mapping of TCVs, arteries and veins using confocal/two-photon laser scanning microscopy (TPLSM) of cleared bones confirmed this connection, and also showed that arterial TCVs can feed directly into the venous circulation at the endosteal surface while a significant portion of the veins feed directly into bone-traversing TCVs (Supplementary Fig. 3; Supplementary Video 6). This loop would allow cells entering the BM via arterial TCVs to exit again immediately via venous TCVs, thereby facilitating the rapid release of cells from BM.
The average TCV diameter was 11 µm (Fig. 2e), and C57BL/6J tibiae contained more than 900 of these vessels (Fig. 2f). This diameter is considerably less than that recently reported for bone channels in mice27. Results for TCV diameter were identical as derived from different analyses, and always ~50% of the size of BM sinusoids (Fig. 2g–n). Furthermore, we identified TCVs in all types of murine long bones investigated, and also similar structures in the flat calvaria bone (Supplementary Fig. 4), thus confirming recently published results27.
TCVs vary in structure and molecular composition and are located in distinct zones of the bone shaft
A more detailed characterization showed four main TCV types with different degrees of straightness, position and orientation along the bone (Fig. 3a–d). Depending on their localization along the bone shaft, up to 60% of TCVs were arteries and 40% veins (Fig. 3e). This was further demonstrated by the expression of typical endothelial markers (CD31, CD34 and von Willebrand factor37) in all TCVs (Supplementary Fig. 9a), while arterial TCVs selectively expressed Sca-1 and α-SMA5,31 (Supplementary Fig. 9b). Venous TCVs expressed Ephrin B438, while we could not detect the lymphatic endothelial marker Lyve-139 in TCVs (Supplementary Fig. 9c,d). Pimonidazole staining showed hypoxic areas associated with venous TCVs, but not around arterial TCVs (Supplementary Fig. 5). Individual TCVs were rarely straight, but instead showed varying levels of directional change (Fig. 3f,g) that correlated with the thickness of the cortical bone traversed (Fig. 3h).
The majority of arterial and venous blood flow in long bones travels via TCVs
Given their localization and number, we hypothesized that TCVs may contribute substantially to blood flow into and out of long bones. This concept was supported by the accumulated cross-sectional area of vessels entering or leaving the bone, which was dominated by the contribution of TCVs (Fig. 3i,j). Indeed, intravital imaging demonstrated effective erythrocyte transport through both venous and arterial TCVs with transport speeds around twofold higher than that of type H vessels within BM30,40 (Fig. 3k,l; Supplementary Video 7). Quantification in tibiae revealed that the combined volume flow in all TCVs was by far the largest component of blood flow into and out of the bone (Fig. 3m), accounting for 83% of arterial and 59% of venous flow (Fig. 3n). Hence, blood flow through the long bones in mice is dominated by the contribution made by TCVs, while nutrient arteries and large exiting veins play only minor roles.
Neutrophil granulocytes mobilized by G-CSF exit the bone via TCVs
To examine whether TCVs are limited to erythrocyte transport or also allow the transit of leukocytes, we injected G-CSF into mice. G-CSF mobilizes neutrophil granulocytes from the BM within minutes17,18,24. We hypothesized that, in order to migrate from the BM so rapidly, neutrophils might be travelling through TCVs, as has been shown recently for the calvaria27. Intravital imaging of mice with EGFP+ neutrophils41 indeed showed multiple exiting cells following systemic injection of G-CSF, sometimes even travelling against the direction of blood flow (Fig. 3o,p; Supplementary Video 8). This process was confirmed in an animal model with neutrophil-specific tdTomato expression26.
Osteoclasts are essential for TCV formation
Trans-cortical vessels require the presence of narrow canals in cortical bone (Fig. 1f–h) that are then lined by endothelial cells. From human anatomy, it is known that the basic multicellular unit (BMU) of osteons constantly generates holes along the long axis of bones42. At the tip of the BMU in the cutting cone are osteoclasts that can directly dissolve bone matrix and generate a canal in the calcified matrix42,43. Osteoclasts are also thought to generate much thinner canals perpendicular to osteons (that is, across compact bone) that might form the basis for Volkmann’s canals in the human skeleton44. Hence, we speculated that osteoclasts might also be involved in the generation of TCVs. In murine bone sections with tdTomato expression driven by an ectopic BAC-CX3CR1 promotor construct, osteoclasts were prominent in their expression of high levels of tdTomato and tartrate-resistant acid phosphatase (Supplementary Fig. 6a). These cells were found in large numbers on the endosteal surface of long bones and sporadically within TCVs, but rarely on the periosteal side (Fig. 4a–c, Supplementary Fig. 6b; Supplementary Video 9). Osteoclasts were distinguishable from normal macrophages as multinucleated cells with a specific set of surface markers and an extremely high expression of tdTomato (Supplementary Fig. 6b–e). Using high-resolution confocal microscopy of bone sections, we found individual osteoclasts mostly in the middle of existing TCVs, where they also appeared to contact canaliculi projected by osteocytes (Fig. 4c,d; Supplementary Video 15). Indeed, canaliculi were observed to commence at the inner surface of TCVs (Fig. 4e,f). Osteoclasts often showed a resorption lacuna at the contact area with the TCV wall (Fig. 4g,h; Supplementary Video 10), suggesting ongoing TCV remodelling or arborization. Thus, osteoclasts are strategically positioned for involvement in TCV formation.
We therefore investigated an animal model of human TNFα overexpression characterized by osteoclast hyperactivity and osteoporosis45. Animals from this model were frail and had very thin cortical bones (Fig. 4i). Although the absolute number of TCVs was less than in wild-type littermates (Fig. 4j), they had more TCVs per bone volume which was reversed only following zoledronate-induced bone thickening (Fig. 4k,l). Interestingly, the blockade of osteoclast function by zoledronate44 led to a highly significant reduction in TCV numbers in this model after 28 days. Since wild-type littermates also showed a reduction in TCV numbers following administration of zoledronate (Fig. 4j), we conclude that osteoclast-mediated bone resorption is important in the maintenance of TCVs44. The role of osteoclasts in TCV generation was further demonstrated in the bones of 3-week-old mice with genetically induced osteoclastic hyperactivity46 (Ctnnb1ex2fl/fl;Col10a1-Cre, Supplementary Fig. 7). In contrast, mice with genetically induced osteo blast hyperactivity (Ctnnb1ex3fl/+;Col10a1-Cre, Supplementary Fig. 7) showed a mild yet insignificant tendency toward lower TCV numbers.
De novo formation of TCVs in chronic, but not acute, arthritis
Next, we queried whether chronic inflammatory bone destruction, where sustained increased recruitment of immune cells is an underlying pathophysiological phenomenon, affects the TCV system. We used the model of glucose-6-phosphate isomerase (G6PI)-induced arthritis, in which arthritis is induced in wild-type mice by a single immunization with G6PI47 in complete Freund's adjuvant (CFA48). The acute and self-limiting course of G6PI-induced arthritis can be switched to chronic-destructive arthritis by transient depletion of regulatory T cells (Treg) prior to immunization49,50. Clinical signs of arthritis developed in a highly synchronized manner and were first detectable at day 8, reaching maximum levels at day 14 after immunization47. Thereafter, chronic arthritis persisted for >90 days resulting in joint destruction, whereas acute arthritis had resolved by day 6049. In chronic arthritis we noted a strong endothelial activation of TCVs that was characterized by significantly increased expression of ICAM-1 and VCAM-1 in TCV endothelial cells (Fig. 5a–d). Furthermore, while TCV numbers were significantly increased in chronic arthritis at day 62, we observed no such changes in acute arthritis (Fig. 5e,f). Interestingly, at day 62 the number of TCVs in certain Treg-depleted mice injected with PBS/CFA alone was also higher than at day 14 (Fig. 5e), suggesting that the intense and protracted myelopoiesis induced by CFA48 also induces an increase in TCVs. Confirming and extending our earlier observations51,52, we also observed bone neo-formation in chronic arthritis (Fig. 5a,b, white arrows). The newly formed bone also contained TCVs (not shown). Collectively, these data show that TCVs can develop de novo as a result of chronic arthritis. This observation raised the question as to whether systemic inflammation in general influences the number of TCVs. In a model of chronic gut inflammation53 (Fig. 5g), however, we detected no increased numbers of TCVs, suggesting that only inflammatory processes affecting bone modulate the number of TCVs in mice. To confirm that assumption, we analysed the effects on TCVs of lethal irradiation followed by bone marrow transplantation26. Interestingly, irradiated and transplanted mice showed a highly significant decrease in TCVs in cortical bone compared to untreated controls (Fig. 5h). An additional factor that determined TCV numbers was age. Aged mice showed a highly significant reduction in both TCV numbers and bone thickness compared to younger animals, while the overall shape of the vessels remained unchanged (Supplementary Fig. 8). In summary, these experiments indicate that several processes affecting bone homoeostasis influence the remodelling and arborization of the TCV system in long bones.
Evidence of direct trans-cortical blood transport in human bone
We next questioned whether TCVs are a unique feature of rodents or whether they can also be found in human bone. Intra-operative images of bone surfaces from human patients showed characteristic punctate bleeding points along the shaft of various long bones (Fig. 6a–c) that are considered to indicate viable bone structure54. Next, we reconstructed the intratibial blood flow of a human volunteer using high-resolution 7 tesla (T) ultra-high-field magnetic resonance imaging. This approach demonstrated the presence of vessels highly reminiscent of a nutrient artery, and a central sinus that branched into fine sinuses within the tibia (Fig. 6d–h; Supplementary Video 11). In some sections, small canal-like structures were seen entering the cortical bone (Fig. 6g,h). Interestingly, endoscopic images from human femoral necks showed very fine blood vessels emanating directly from the bone shaft, and these were clearly percolated (Supplementary Video 12). Hence, we finally investigated cleared samples of human femoral necks sourced from orthopaedic surgeries by LSFM. In this case we detected vessels (Fig. 6i,j) that were structurally similar to TCVs (Supplementary Video 14), yet were much thicker in diameter than their murine counterparts (Fig. 6k). Taken together, these data suggest that human long bones, at least in some areas, also possess a system of TCVs that directly connects the vascular system of the BM to the peripheral circulation through cortical bone. This is highly reminiscent of similar structures found in the human skull27.
Our results identify evolutionarily conserved blood vessels in mice that are essential in deriving a closed circulatory system in long bones. These vessels are responsible for the majority of blood flow into and out of bones, and mediate the recruitment of immune cells from the BM to the circulation. These structures may have escaped attention because an array of advanced imaging approaches is required to identify and characterize them accurately. Previous analyses using X-ray microtomography detected canal systems in murine bones, but were unable to establish these as percolated blood vessels55,56. Studies in guinea pigs32 concluded only that there is a system of canals through the bone but that their precise role—for example, in terms of flow direction or the ability to transport immune cells—could not be identified at the time due to lack of suitable methods. Comparable studies in mice have only recently been performed, and these detected bone channels that are similar to TCVs27. The latter study is hence an independent confirmation of our data, which we further extend by providing a more advanced molecular characterization of TCV endothelia as well as a comprehensive assessment of TCV function and their role in total bone blood supply.
Both the number of TCVs and their predominance in the total blood flow of long bones are remarkable, but have not been taken into account in recent measurements of intra-osseous blood flow in mice40. Such a system might provide an explanation for the rapid biodistribution of intra-osseous injections.
We can demonstrate hypoxic areas around venous TCVs indicating that the dense TCV network influences the oxygenation of BM. This might have consequences for haematopoiesis, where it has been shown that the level of oxygen is critical for the functioning of HSC and that low oxygen tension might be required to maintain HSC in a resting state57. Our data show an inhomogeneous distribution of arterial and venous TCVs along the bone shaft: the upper half of the tibia is enriched in arterial TCVs, while in the lower half arterial and venous TCVs are equally distributed. We also found a limited number of nutrient arteries at the metaphysis. It is likely that this distribution leads to differences in oxygen tension and thus HSC biology in specific areas of BM.
Other potential functions of TCVs include nutrient and oxygen supply for intra-osseous osteocytes58, and contact formation between osteocytes and blood-borne osteoclasts. We have observed the close proximity of osteocytes to TCVs and even direct physical contact between osteoclasts and osteocyte canaliculi, which confirms recent ultra-structural studies59. It has been speculated that fluid movement in canaliculi is initiated by bone strain and guides bone remodelling processes60. Our observation of TCV-associated osteoclasts penetrating the walls of TCVs rich in canaliculi would fit within such a concept. Notably, osteocytes are also the main source of RANKL, the key effector molecule for osteoclast differentiation and activity61,62,63. The close physical association between TCV-associated osteoclasts and osteocytes may allow guided osteoclast-mediated bone resorption, as RANKL signalling is most effective via direct cell–cell contact64. These observations suggest a more direct interaction between osteocytes and osteoclasts than previously expected, which is orchestrated by TCVs. It is known that osteocytes produce a broad range of factors with vasoactive function: nitric oxide, for example, can regulate vasodilatation, endothelial cell proliferation and migration65,66,67. Prostaglandin E2, which is known to be produced by osteocytes, has regulatory effects on endothelial permeability and angiogenesis63,68,69. Thus, it is conceivable that osteocytes affect not only bone metabolism, but might also play a role in regulating cortical bone vascularization.
One model of chronic arthritis showed that new TCVs can form rapidly, which may facilitate the continuous efflux of neutrophils and other leukocytes from BM into joints, as has been shown in stroke-related neutrophil recruitment27. Each TCV requires a full canal through cortical bone, ~10–15 µm in width and >100 µm in length. Observations in human rheumatoid arthritis have also suggested an increase in cortical micro-canals70, indicating that changes in the cortical vascular system are highly dynamic. In humans, a BMU consisting of osteoclasts, osteoblasts and reversal cells/osteoprogenitors forms a bone-cutting cone to generate new Haversian canals43. However, the diameter of such canals is too great for comparison to murine TCVs and hence human-type BMU cannot be responsible for TCV generation in mice. Instead, we observed mostly single osteoclasts at the centre of existing TCVs that appear to be essential for TCV formation, reorganization and branching. Such branching of TCVs is reminiscent of the human Haversian system, which also changes via lateral and dichotomous bifurcation of existing canals during bone remodelling71. In accordance with this, we can show that blocking of osteoclast function limits TCV formation and remodelling in mice. Therefore not only the anti-resorptive function of bisphosphonates72, but also their potential effects on bone vascularization, should be taken into consideration in future studies on arthritis and osteoporosis.
Our investigations on models of genetically, age- and inflammation-mediated bone diseases suggest that the TCV system is intimately associated with bone turnover, and therefore may play an important role in various bone diseases. In regard to ageing, for instance, we found substantial TCV loss suggesting that TCVs are not indefinitely stable. Notably, human osteocyte numbers decline with age, which has been suggested to impair bone stability due to insufficient repair of micro-cracks62. Hence, loss of TCVs and the resulting decline in vascularization of bone could be an attractive concept, explaining the decline in osteocyte numbers during ageing. Future studies should therefore aim to identify the factors that maintain TCVs. Such considerations are also important for appropriate fracture healing, which may require extensive TCV remodelling73. Based on the localization of osteoclasts in the few observable incomplete TCVs, our current data support a concept that argues for generation from within the bone marrow out, at least to a large extent. Thereby osteoclast function plays an important role in TCV generation, yet the endothelial-specific genes involved in TCV generation and the exact developmental timing of their generation are important issues that need to be further clarified in future studies.
Another condition that induces a sharp decline in TCVs is the irradiation of bone during stem cell transplantation. Although it has been stated that irradiation promotes bone loss and the development of insufficiency fractures, the underlying mechanisms that affect bone quality remained largely undefined74,75,76,77. Irradiation-induced decline in TCV numbers could thus represent a hitherto unrecognized mechanism that explains radiotherapy-induced bone loss.
In summary, these data provide a new concept of bone and BM physiology by showing the existence of a conserved vasculature that not only permits the rapid efflux of leukocytes into the circulation for host defences, but also controls bone homoeostasis and function. Since key bone pathologies are associated with alterations in the TCV system, entirely new research possibilities that further characterize the role of TCVs in skeletal biology and disease can be envisioned.
All animal experiments were performed in accordance with German guidelines and laws, were approved by local animal ethic committees and were conducted according to the guidelines of the Federation of European Laboratory Animal Science Associations. For all experiments, female mice were used with the exception of the DBA/1 DEREG strain. Here both sexes were used. The age of all mice was between 7 and 12 weeks unless stated otherwise.
C57BL/6JOlaHsd, C57BL/6Rj, LysM-EGFP, CatchupIVM-red, Ctnnb1ex2fl/fl;Col10a1-Cre+ and Ctnnb1ex3fl/+;Col10a1-Cre+ mice were bred and housed under specific pathogen-free conditions at the animal facility of the University Duisburg-Essen. LysM-EGFP, CatchupIVM-red, Ctnnb1ex2fl/fl;Col10a1-Cre+ and Ctnnb1ex3fl/+;Col10a1-Cre+ mice were described previously26,41,46. Ctnnb1ex2fl/fl;Col10a1-Cre+ mice were generated by crossing Ctnnb1tm2.1Kem78 and Tg(Col10a1-cre)1427Vdm mice79, while Ctnnb1ex3fl/+;Col10a1-Cre+ mice were generated by crossing Ctnnb1tm1Mmt80 and Tg(Col10a1-cre)1427Vdm mice. Gpr15gfp/+ Foxp3ires-mrfp mice were described previously81, and were bred and housed together with C57BL/6JRj mice under specific pathogen-free conditions in the Laboratory Animal Facility of University Hospital Essen. Reportable experiments involving C57BL/6JOlaHsd, C57BL/6JRj, LysM-EGFP and CatchupIVM-red mice were approved by Landesamt für Natur, Umwelt und Verbraucherschutz (LANUV) of North-Rhine Westphalia, registration numbers 84-02.04.2013.A328, 84-02.04.2013.A129 and 81-02.04.2017.A456.
The DBA/1-DEREG mice were generated by speed congenic back-cross of DEREG mice82 onto the DBA/1J strain, and were housed in the specific pathogen-free facility of University Hospital Jena. All experiments involving DBA/1-DEREG mice were conducted following approval by Thüringer Landesamt für Verbraucherschutz, Bad Langensalza, Germany, registration number 02-079/14.
To generate CX3CR1-cre;tdTomato mice, STOCK Tg(Cx3cr1-cre)MW126Gsat/Mmucd mice (identification number 036395-UCD) were obtained from the Mutant Mouse Regional Resource Center, a NIH-funded strain repository, and were donated to the MMRRC by the NINDS-funded GENSAT BAC transgenic project. They were crossed with B6;129S6-Gt(ROSA)26Sortm9(CAG-tdTomato)Hze/J mice83, resulting in CX3CR1-cre;tdTomato mice. CX3Cr1-Cre;tdTomato mice were bred and housed together with hTNFtg mice45 at the animal facilities of the University of Erlangen, under specific pathogen-free conditions. Experiments involving hTNFtg mice were approved by the Veterinary Office of the Government of Lower Franconia, registration number 54-2532.1-26/12. All mouse strains used are listed in further detail in the reporting summary document and in Supplementary Table 1.
Cryo-sectioning of murine long bones
Murine long bones were fixed in 4% PFA/PBS for 4 h at room temperature after perfusion with EDTA/PBS and 4% PFA/PBS, embedded in OCT compound (Sakura Finetek GmbH) and snap-frozen in liquid nitrogen. A Thermo Fisher cryostat and Cryofilm Type 2 C(9) (Section-Lab Co) were used for sectioning of bone samples.
Histological TRAP staining
After removal, all soft tissue bone samples from CX3CR1-cre;tdTomato mice were fixed in 4% PFA/BS overnight and subsequently incubated for 7 days in decalcification buffer (14% EDTA, 25% ammoniac) under agitation. After completion of decalcification, the remaining bone tissue was saturated with 30% sucrose overnight and then cryo-embedded. Cryo-sections of 7 µm were obtained, washed with distilled water to remove the cryo-embedding matrix and incubated for 5–15 min with TRAP staining solution (Sigma-Aldrich, Acid Phosphatase, Leukocyte (TRAP) Kit, no. 387 A) according to the manufacturer's instructions until proper purple staining of osteoclasts was complete. Nuclei of the same samples were stained with DAPI, 0.2 µg ml–1 (Sigma-Aldrich, no. D9542) for a further 10 min. After washing with distilled water, samples were mounted with fluorescence mounting medium (DAKO, no. S3023) for preservation. Imaging was performed with a Keyence Fluorescence Microscope BZ-X700, where bright-field was used for TRAP staining and fluorescence light was used for DAPI staining (Ex 360/40, DM 400, BA 460/50) and tdTomato signal (Ex 545/25, DM 565, BA 605/70).
Histological immunofluorescence staining
For staining cryo-sections of murine long bones, samples were blocked and permeabilized with 1% BSA and 1% Tween20 in PBS for 1 h at room temperature. Blood vessels were stained, with the antibodies listed in Supplementary Table 2, for 4 h at room temperature. Primary antibodies were washed off with PBS. ICAM-1, VCAM-1 and NG2 were counterstained with chicken anti-rat AF647 antibody, Ephrin B4 with donkey anti-goat AF647 antibody and CD34 with Streptavidin AF488for 4 h at room temperature, and washed three times with PBS. Bone sections were DAPI-stained by embedding with DAPI Fluoromount-G (cat. no. 0100-20, Southern Biotech). Detailed information on all antibodies used is listed in Supplementary Table 2.
To clarify oxygen transport via TCVs, pimonidazole staining (Hydroxprobe Red 549 Kit, Hydroxyprobe, cat. no. HP7-200Kit) indicating hypoxia was performed.
Female 7–12-week-old C57BL/6J mice received 120 mg kg–1 pimonidazole hydrochloride in PBS by intravenous injection and were sacrificed 2 h later. Histological cryo-sections and immunofluorescence staining of long bones were processed as described above. Detection of hypoxia was performed using a kit including mouse Dyligh549 anti-pimonidazole antibody (1:100 for 4 h at room temperature).
Whole-mount staining and optical clearing of human bone tissue
An adult human femoral head and neck was obtained from a patient undergoing total hip arthroplasty for osteoarthritis. The patient gave informed consent prior to surgery, and the institutional ethics committee of University Hospital Erlangen approved the study. The femoral neck was fixed in 4% PFA/PBS for 24 h at 4–8 °C. Tissue samples were blocked and permeabilized with 1% BSA and 1% Tween20 for 7 days under slight shaking at 4–8 °C. For staining of endothelium we used an anti-human CD31-AF594 antibody (Biolegend, cat. no. 303126, 1:200), and for arterial staining an AF647-labelled anti-human alpha-smooth muscle actin antibody (Novus Biologicals, cat. no. NBP2-34522AF647, 1:200). For tissue staining, samples were incubated for 7 days with slight shaking at 4–8 °C. Samples were then washed twice with 1% Tween 20/PBS for 24 h and cleared with an adjusted simpleCLEAR protocol. According to sample size, bone tissues were each dehydrated with 50, 70 and twofold 100% ethanol for 24 h in gently shaken 50 ml tubes at 4–8 °C, and finally cleared with ethyl cinnamate (Sigma-Aldrich, cat. no. 112372-100 G) at room temperature for 24 h.
Further information on patient recruitment and software versions used for data collection and processing can be found in the reporting summary document and Supplementary table 4.
Induction and assessment of arthritis
Recombinant human G6PI was prepared as previously described47. DEREG mice were immunized on day 0 with a subcutaneous injection of 400 μg recombinant human G6PI emulsified 1:1 (vol/vol) with CFA (Sigma-Aldrich), cat. no. F5881-10ML), with PBS/CFA alone or with PBS without CFA.
Mice were examined for signs of arthritis at least three times per week, and disease severity was recorded for each mouse. The score comprises the number of swollen toes, each assigned 0.5 points, as well as the level of swelling and redness in each of the metatarsal/metacarpal regions and the carpal–metacarpal/tarsal–metatarsal joints. Swelling and redness were determined using the following scoring system: 0, normal; 1, mild redness and swelling; 2, moderate swelling; 3, severe swelling with oedema. The maximum score for each mouse was 33.
To deplete regulatory Tregs, DEREG mice were each treated with 0.5 mg diphtheria toxin (Calbiochem, cat. no. 322326-1MG) intraperitoneally on days −2, −1, +4 and +5 relative to immunization with G6PI, as previously described50,82.
At 14, 28 or (66) 67 days after immunization, animals were narcotized with 1.5–2.0% isoflurane (Forene 100%, AbbVie Deutschland GmbH & Co.) and injected intravenous with 10 µg CD31-AF647, in a total volume of 150 µl PBS with an insulin syringe into the retro-bulbar plexus. Twenty minutes after antibody injection, the mice were killed by CO2 inhalation and perfused with 15 ml cold 5 mM EDTA/PBS, followed by perfusion with 15 ml cold 4% PFA/PBS. Subsequently the legs were fixed in 4% PFA/PBS overnight at 4–8 °C. After fixation the bones were prepared for optical clearing as described above.
Colitis-associated colonic cancer induction by azoxymethane/dextran sulfate sodium
For induction of colonic cancer, female 7–12-week-old Gpr15gfp/+ Foxp3ires-mrfp mice were injected intraperitoneal with 12.5 mg kg–1 body mass of the pro-carcinogen azoxymethane (AOM, Sigma-Aldrich, cat. no. A5486-25MG). In weeks 1, 4 and 7 following AOM administration, mice received drinking water supplemented with 2% dextran sulfate sodium salt (DSS, MP Biomedicals, cat. no. 0216011010) for 5 days. Mice were sacrificed at week 12, and bone samples were prepared for LSFM as described above.
Zolendronic acid treatment
To inhibit osteoclast activity, female 7–12-week-old hTNFtg and C57BL/6 WT littermates were treated with 100 µg kg–1 body mass zolendronic acid (4 mg 100 ml–1, medac GmBH) in 100 µl PBS once weekly for four weeks. Control mice received pure PBS once per week for four weeks. All treatments were administered by intraperitoneal injection. Mice were examined for signs of arthritis at least three times per week, and body weight and grip strength were recorded for each mouse.
Five weeks after starting zolendronic acid treatment, the animals were narcotized with 1.5–2.0% isoflurane (Forene 100%, AbbVie Deutschland GmbH & Co.) and injected intravenous with 10 µg CD31-AF647, in a total volume of 150 µl PBS with an insulin syringe into the retro-bulbar plexus. Twenty minutes after antibody injection, mice were killed by CO2 inhalation and perfused with 15 ml cold 5 mM EDTA/PBS, followed by a perfusion with 15 ml cold 4% PFA/PBS. Subsequently the legs were fixed in 4% PFA/PBS overnight at 4–8 °C. After fixation, the bones were prepared for optical clearing as described above.
Lethal irradiation and reconstitution of mice with donor bone marrow
Bone marrow from four female 7–12-week-old C57BL/6JRj donor mice was flushed out of the tibia and femur with sterile PBS. The marrow was re-suspended in sterile PBS to break up any clumps and passed through a 70 µm strainer to remove large fragments. When ready for injection, cells were centrifuged for 10 min at 1,200 r.p.m. and re-suspended in PBS to give a final concentration of 34 million cells ml–1.
The night before irradiation, recipients were denied food then irradiated with 9.5 Gy (950 Rad) before being housed again with access to food and antibiotic-supplemented water (1:100 ciprofloxacin 200). Six hours after irradiation, mice were injected intravenous with 5 million cells in 150 µl sterile PBS. They were maintained on antibiotic water for 2 weeks before being sacrificed 4 weeks after irradiation, when bone marrow and bone samples were prepared for LSFM as described above.
Images and videos of exposed human long bones
Demonstrating the clinical impact of our hypothesis, cortical bleeding was documented by images and videos of patients undergoing orthopaedic procedures on diaphyseal long bones (fibula/tibia/femur) and femoral neck. Individuals were selected randomized and incidentally by the orthopaedic surgeon. Here, informed consent was given following the institutional guidelines (Orthopaedic and Trauma Department, University of Duisburg-Essen).
Further information on patient recruitment is listed in the reporting summary document.
LSFM of optically cleared samples
For LSFM imaging of simpleCLEAR optically cleared samples, we used an LaVision BioTec Ultramicroscope (LaVision BioTec) with an Olympus MVX10 zoom microscope body (Olympus), a LaVision BioTec Laser Module, an Andor Neo sCMOS Camera with a pixel size of 6.5 µm, and detection optics with an optical magnification range 1.263–12.63 and a numerical aperture (NA) of 0.5. Because a non-specific autofluorescence signal is useful for visualizing general tissue morphology, a 488 nm optically pumped semiconductor laser (OPSL) was used for generation of autofluorescent signals. For CD31-AF594 excitation, we used a 561 nm OPSL and, for CD31-AF647, Sca-1-AF647 and SMA-AF647 excitation, a 647 nm diode laser. Emitted wavelengths were detected with specific detection filters: 525/50 nm for autofluorescence, 620/60 nm for CD31-AF594 and 680/30 nm for CD31-AF647, Sca-1-AF647 and SMA-AF647. The optical zoom factor for measurements ranged from 1.26 to 12.60, and the light-sheet thickness ranged from 5 to 10 µm.
Further information on software versions used for data collection and processing is listed in the reporting summary document and Supplementary table 4.
X-ray microtomography imaging
For X-ray microtomography imaging of hTNFtg tibiae, simpleCLEAR optically cleared samples were rehydrated by incubation in 1% Tween20 in 70% ethyl alcohol, followed by 1% Tween in 50% ethyl alcohol and two further treatments with pure PBS. All incubations were performed at room temperature with gentle shaking in 5 ml Eppendorf tubes.
Micro-CT imaging was performed using the cone-beam desktop micro-computer tomograph µCT 40 (SCANCO Medical AG). Settings were optimized for calcified tissue visualization at 45 peak kilovoltage (kVp), with a current of 177 µA and 240 ms integration time for 500 projections per 180° and, furthermore, 8.0 µm was set as isotropic voxel size for optimal resolution. For the segmentation of 3D volumes, respective greyscale thresholds were determined using the operating system Open VMS from SCANCO Medical. The entire tibial cortex only was chosen as volume of interest for bone volume analysis.
Further information on software versions used for data collection and processing is listed in the reporting summary document and Supplementary Table 4.
Single- and two-photon laser scanning microscopy of cleared organs
For high-magnification imaging of ethyl cinnamate-cleared bones, a Leica TCS SP8 fully automated epifluorescence confocal microscope (Leica Microsystems) with Acousto-Optical Tunable Filter (AOTF) and Acousto-Optical Beam Splitter (AOBS) scanoptics, HyD detection, two-photon and compact OPO on a DM6000 CFS frame was used. Imaging of ethyl cinnamate-cleared tibiae and fibulae was performed with a ×25 HCX IRAPO L water-immersion objective with a NA of 0.95.
Since optical clearing is reversible, the cleared samples were embedded in ethyl cinnamate-filled microscopy chambers, which were sealed with a cover slip. Fluorescence signals were generated via sequential scans, exciting Sca-1-AF647 via single-photon excitation using an HeNE laser at 633 nm and detecting in confocal mode with an internal HyD at 660–720 nm. The second confocal mode sequence included a DPSS single-photon laser at 561 nm for excitation of CD31-AF594, and an internal PMT detector at 600–640 nm. The third sequence was performed with a Titan-Sapphire laser tuned to 960 nm for SHG detection at 460/50 nm detected with an external photomultiplier tube (PMT NDD1).
For histological CLSM data, a Leica SP5 II confocal microscope (Leica Microsystems) with AOTF and AOBS, and HyD detection on a DMI6000 CS frame, was used. Imaging of coverslip-embedded samples was performed using an HCX PL APO ×100 oil objective with a NA of 1.44. Fluorescence signals were generated via sequential scans, exciting tdTomato or AF555 using a DPSS laser at 561 nm and detecting with an HyD tuned to 600–650 nm. The second sequence for visualizing AF488 signals comprised an Argon laser at 488 nm for excitation and an HyD detector tuned to 500–550 nm. As a third sequence, a 633 nm Helium–Neon laser for Alexa Fluor 647 excitation and an HyD tuned to 650–700 nm for detection were used. The fourth sequence included a 405 nm laser Diode for DAPI excitation and HyD detection at 470–520 nm.
Further information on software versions used for data collection and processing is listed in the reporting summary document and Supplementary table 4.
Mice were prepared for intravital TPLSM as previously described4. TPLSM was performed with a Leica system as described above. EGFP+ cells of female 7–12-week-old LysM-EGFP mice and tdTomato+ cells of CatchupIVM-red mice were excited at 960 nm, at which point bone tissue additionally emits a SHG signal at 480 nm. Fluorescent cells were detected with specific filters at either 525/50 nm (EGFP) or 585/50 nm (tdTomato), and SHG was detected via a 460/50 nm filter.
Blood flow was visualized by injecting (intravenous) either 1.5 mg ml–1 rhodamine dextran (Sigma-Aldrich, cat. no. R9379–100MG) or 1 µM Qtracker 655 Vascular Labels (Thermo Fisher, cat. no. Q21021MP) in a total volume of 100 µl PBS. Fluorescence was excited at 960 nm and detected with either a 585/40 nm (rhodamine dextran) or a 650/50 nm (Qtracker 655) filter. Imaging was performed in both resonant and non-resonant detection mode. Scan speed was adjusted individually for different vessel types, from 600 Hz to 12 kHz.
Neutrophils were activated by injecting (intravenous) 100 µg kg–1 body weight human recombinant granulocyte-stimulating factor (Neupogen, Amgen GmbH) in a total volume of 100 µl PBS. The raw data were reconstructed and analysed using Imaris software (Bitplane) and ImageJ.
Further information on software versions used for data collection and processing is listed in the reporting summary document and Supplementary table 4.
X-ray microscopy of murine long bones
Female 7–12-week-old C57BL/6J mice were painlessly killed via cervical dislocation, and the hind legs were prepared. Surrounding muscle tissue was circumspectly removed from the entire leg. The remaining tissue was digested in a collagenase solution consisting of 1 mg ml–1 collagenase Type IV and 10 mM HEPES in HBSS, with gentle shaking for 12 h at 37 °C. After incubation, the remaining tissues were dissected into their constituent parts. The separated tibiae were collected and incubated again in collagenase solution for 12 h at 37 °C.
The X-ray microscope is a sample-rotating system providing a suitably long working distance and adjustable energy range for obtaining 3D density data from a sample, whereas conventional 2D radiographs capture the X-ray density of a sample from hundreds of different angles. The data are reconstructed into a 3D data set with each voxel containing a value of X-ray density of that location in space. This method can capture complex internal geometries where there is sufficient contrast between material densities34. Recent advances in this field allow voxel sizes below the micrometre range84.
Scanning of the tibia was performed using an isotropic voxel size of 1.7 μm with the ×4 objective on a Zeiss Versa 520 (Carl Zeiss). A high signal–noise ratio was achieved by collecting 1,885 projections per rotation with a projection exposure time of 8 s (40 kVp voltage, 3 W power, 360° angle range). Generated data were reconstructed and analysed using Imaris (Bitplane).
Field emission scanning electron microscopy of murine long bones
Female 7–12-week-old C57BL/6J mice were painlessly killed via cervical dislocation and the hind legs were prepared. The muscles were circumspectly removed from bones (femurs and tibiae), the latter then being fixed in a solution of 2% glutaraldehyde, 3% formaldehyde, 0.01 M calcium chloride, 0.01 M magnesium chloride and 0.09 M saccharose for 12 h at room temperature. Fixed samples were washed twice for 10 min with TE buffer (20 mM TRIS, 1 mM EDTA, pH 6.9) and dehydrated with a graded series of acetone (10, 30, 50, 70, 90%) for 30 min for each step on ice. Samples in the 100% acetone step were allowed to reach room temperature before a further change of 100% acetone. Samples were then subjected to critical-point drying with liquid CO2 (CPD 30, Bal-Tec). Dried samples were fixed onto aluminium stubs with plastic conductive carbon cement (PLANOCARBON, Plano, cat. no. N650) and covered with gold film by sputter-coating (SCD 500, Bal-Tec) before examination with a field emission scanning electron microscope, Zeiss Merlin (Zeiss), using an Everhart Thornley HESE2 detector and an in-lens SE detector (at a ratio of 25:75) with an acceleration voltage of 5 kV.
Magnetic resonance imaging of a human shank
Approval from the local institutional ethics committee of the medical faculty of the University of Duisburg-Essen was gained prior to this study. After signing informed consent, the right lower leg of a healthy 47-year-old male subject was imaged on a 7-Tesla research whole-body magnetic resonance system (Magnetom 7 T, Siemens Healthcare). The leg was placed feet-first and supine within an in-house-developed eight-channel radiofrequency transmit/receive head coil85. An additional seven-channel loop receive-only radiofrequency array was placed on top of the tibia and fixed with a vacuum pillow and Velcro strips86. Transmitter adjustment was performed with a vendor-provided B1 mapping sequence based on a spin-echo and a stimulated echo87. For high-resolution imaging, T1-weighted, fat-saturated pulse sequences were acquired in 2D (fast low-angle shot, FLASH) and 3D (volume-interpolated breath-hold examination, VIBE). Additionally, a time-of-flight sequence was used to distinguish the direction of blood flow88. Time-of-flight MR angiography images are expected to show blood flowing in the caudocranial direction hyper-intensely when the 80 mm (width) saturation band is placed cranially (11 mm gap to excitation slab), while flow in the craniocaudal direction will be hyper-intense when the saturation band is placed caudally. Further imaging details are summarized in Supplementary Table 5.
Image evaluation was performed on the magnetic resonance console Syngo VB17 (Siemens Healthcare), and then data export was reconstructed using ImageJ and Imaris (Bitplane). All MR sequences were acquired in transverse orientation with phase-encoding direction anterior–posterior. A parallel imaging factor of 2 was applied for each sequence89.
Further information on patient recruitment and on software versions used for data collection and processing is listed in the reporting summary document and Supplementary Table 4.
Blood vessel diameter measurement
Arteries and veins were identified by their specific antibody staining (veins: CD31+/Sca-1–, arteries: CD31+/Sca-1+) in Imaris. To measure the diameter of the vessel types identified, 5 µm optical sections were generated with the 'slice' tool and diameters were measured via the 'measurement point' tool in Imaris. This analysis was done with LSFM data of entire tibiae, intravital TPLSM data and histological bone sections. In the case of intravital TPLSM, only TCVs and sinusoids were measured as these were identifiable by their characteristic morphology and location within the bone.
The quantification of total vessel numbers in entire tibiae was based on 100 µm optical sections of LSFM data generated by the slice tool in Imaris. These sections were exported as tiff. files and imported into ImageJ. The vessels were quantified by manual counting via the 'cell counter' tool in ImageJ. In this process we also included vessel orientation and distribution, taking into account the anterior and posterior aspects as well as the upper and lower half of the tibia. The results were confirmed by individual quantification by four independent persons.
TCV type analysis
Quantification of arterial and venous TCVs, plus the quantification of intracortical loops and direct, bifurcated and complex TCVs, was based on manual counting of vessels in 100 µm optical sections via the cell counter tool of ImageJ as described above.
To analyse the straightness of dTCVs, the measurement point tool in Imaris was used, where every shift in direction was set as a new measurement point. The corresponding thickness of the compact bone was analysed with the 'this tool. Here the measured distance was defined as the position of the particular dTCV from the endosteum to the periosteum perpendicular to the bone shaft. The results were confirmed by individual quantification by four independent persons.
For co-localization analysis of histological sections, 3D CLSM stacks were used. Data files were de-convolved using Huygens Professional software and imported into ImageJ. Freehand regions of interest (ROIs) were defined around TCVs and analysed for co-localization with Coloc 2 Plugin.
Estimation of cell velocity from TPLSM images
Erythrocyte velocity was estimated from scanning microscopy images as follows. We extracted ROIs from scanning microscopy images of blood vessels. ROIs contained dark traces left by individual, unlabelled erythrocytes moving with fluorescently labelled blood flow in a horizontal direction (parallel to scanning direction). Typically, several parallel straight linear traces, either diagonally or vertically, could be identified per ROI. The horizontal length, Δxcell, of each trace was interpreted as the distance covered by an erythrocyte while being scanned over time Δt. Thus the velocity of the erythrocyte could be estimated as
with ℓpix the length of a pixel; Δtrow the time needed by the microscope to scan a complete single row of pixels; Δncol and Δnrow the number of pixel columns and rows, respectively, spanned by the trace; and the slope β = Δnrow ⁄ Δncol of the trace. Thus, we estimated erythrocyte velocity from the measurement parameters ℓpix and Δtrow, and the slopes β of the traces extracted from the ROIs.
For each ROI, the slope β was estimated from cross-correlations
of grey-level intensity in pixel columns c and c + Δc at pixel row lags Δr. One or more traces induce a maximum of R(Δc, Δr) for Δr ∝ βΔc. Moving through a series of Δc allows this maximum of cross-correlation to wander through a corresponding series of Δr values. The slope β was then computed by a linear least-square fit to the series of maximum positions (Δc, Δr).
Total blood flow can be calculated from the estimated cell velocity and total cross-sectional area of vessel types defined. Based on total vessel numbers per tibia and vessel diameters defined, the total cross-sectional area (Atotal) of the vessel types identified was calculated as follows:
Here, r2 is the squared radius of the vessel type and n is the total number of vessel types.
As blood flow of the central sinus was not measureable via intravital TPLSM, blood egression was calculated based on NA influx, arterial TCV influx and venous TCV-based blood effusion.
Further information on research design is available in the Nature Research Reporting Summary linked to this article.
A Julia package implementing the computational procedure for the estimation of blood vessel speed is freely available as a source code at https://github.com/DanielHoffmann32/CellSpeedEstimation.jl.
For normal quantile plots, evaluated data were ranked. Rank-based z-scores were calculated based on the mean (µ), standard deviation (x) and sample number (σ):
Finally the z-scores were converted to predicted data values by calculating
and fit into the normal quantile plot.
For calculation of statistical significance, GraphPad Prism 7 was used. Data are presented as mean ± s.e.m. and were analysed using two-sided Student’s t-test, two-sided Mann–Whitney U-test or Kruskal–Wallis H-test, with Dunn’s multiple comparisons test as a post hoc procedure. P < 0.05 was considered significant.
The data that support the findings of this study are available from the corresponding author upon request
Faury, G. Function-structure relationship of elastic arteries in evolution: from microfibrils to elastin and elastic fibres. Pathol. Biol. (Paris) 49, 310–325 (2001).
Bettex, D. A., Pretre, R. & Chassot, P. G. Is our heart a well-designed pump? The heart along animal evolution. Eur. Heart J. 35, 2322–2332 (2014).
Pittman, R. N. Oxygen transport in the microcirculation and its regulation. Microcirculation 20, 117–137 (2013).
Aitsebaomo, J., Portbury, A. L., Schisler, J. C. & Patterson, C. Brothers and sisters: molecular insights into arterial-venous heterogeneity. Circ. Res. 103, 929–939 (2008).
Sivaraj, K. K. & Adams, R. H. Blood vessel formation and function in bone. Development 143, 2706–2715 (2016).
Blumer, M. J., Longato, S. & Fritsch, H. Structure, formation and role of cartilage canals in the developing bone. Ann. Anat. 190, 305–315 (2008).
Sommerfeldt, D. W. & Rubin, C. T. Biology of bone and how it orchestrates the form and function of the skeleton. Eur. Spine J. 10(Suppl 2), S86–S95 (2001).
Zimmermann, E. A., Busse, B. & Ritchie, R. O. The fracture mechanics of human bone: influence of disease and treatment. Bone Rep. 4, 743 (2015).
Nair, A. K., Gautieri, A., Chang, S. W. & Buehler, M. J. Molecular mechanics of mineralized collagen fibrils in bone. Nat. Commun. 4, 1724 (2013).
Simpson, A. H. The blood supply of the periosteum. J. Anat. 140(Pt 4), 697–704 (1985).
Xie, Y. et al. Detection of functional haematopoietic stem cell niche using real-time imaging. Nature 457, 97–101 (2009).
Roche, B. et al. Structure and quantification of microvascularisation within mouse long bones: what and how should we measure? Bone 50, 390–399 (2012).
Acar, M. et al. Deep imaging of bone marrow shows non-dividing stem cells are mainly perisinusoidal. Nature 526, 126–130 (2015).
Chen, J. Y. et al. Hoxb5 marks long-term haematopoietic stem cells and reveals a homogenous perivascular niche. Nature 530, 223–227 (2016).
Köhler, A. et al. Altered cellular dynamics and endosteal location of aged early hematopoietic progenitor cells revealed by time-lapse intravital imaging in long bones. Blood 114, 290–298 (2009).
Spencer, J. A. et al. Direct measurement of local oxygen concentration in the bone marrow of live animals. Nature 508, 269–273 (2014).
Köhler, A. et al. G-CSF mediated thrombopoietin release triggers neutrophil motility and mobilization from bone marrow via induction of Cxcr2 ligands. Blood 117, 4349–4357 (2011).
Devi, S. et al. Neutrophil mobilization via plerixafor-mediated CXCR4 inhibition arises from lung demargination and blockade of neutrophil homing to the bone marrow. J. Exp. Med. 210, 2321–2336 (2013).
Junt, T. et al. Dynamic visualization of thrombopoiesis within bone marrow. Science 317, 1767–1770 (2007).
Chatfield-Ball, C., Boyle, P., Autier, P., van Wees, S. H. & Sullivan, R. Lessons learned from the casualties of war: battlefield medicine and its implication for global trauma care. J. R. Soc. Med. 108, 93–100 (2015).
Clemency, B. et al. Intravenous vs. intraosseous access and return of spontaneous circulation during out of hospital cardiac arrest. Am. J. Emerg. Med. 35, 222–226 (2017).
Von Hoff, D. D., Kuhn, J. G., Burris, H. A. 3rd & Miller, L. J. Does intraosseous equal intravenous? A pharmacokinetic study. Am. J. Emerg. Med. 26, 31–38 (2008).
Morelli, D., Menard, S., Cazzaniga, S., Colnaghi, M. I. & Balsari, A. Intratibial injection of an anti-doxorubicin monoclonal antibody prevents drug-induced myelotoxicity in mice. Br. J. Cancer 75, 656–659 (1997).
Gunzer, M. Traps and hyper inflammation - new ways that neutrophils promote or hinder survival. Br. J. Haematol. 164, 188–199 (2014).
Panopoulos, A. D. & Watowich, S. S. Granulocyte colony-stimulating factor: molecular mechanisms of action during steady state and ‘emergency’ hematopoiesis. Cytokine 42, 277–288 (2008).
Hasenberg, A. et al. Catchup: a mouse model for imaging-based tracking and modulation of neutrophil granulocytes. Nat. Methods 12, 445–452 (2015).
Herisson, F. et al. Direct vascular channels connect skull bone marrow and the brain surface enabling myeloid cell migration. Nat. Neurosci. 21, 1209–1217 (2018).
Morrison, S. J. & Scadden, D. T. The bone marrow niche for haematopoietic stem cells. Nature 505, 327–334 (2014).
Itkin, T. et al. Distinct bone marrow blood vessels differentially regulate haematopoiesis. Nature 532, 323–328 (2016).
Kusumbe, A. P., Ramasamy, S. K. & Adams, R. H. Coupling of angiogenesis and osteogenesis by a specific vessel subtype in bone. Nature 507, 323–328 (2014).
Ramasamy, S. K. et al. Regulation of hematopoiesis and osteogenesis by blood vessel-derivedsignals. Annu. Rev. Cell. Dev. Biol. 32, 649–675 (2016).
De Bruyn, P. P., Breen, P. C. & Thomas, T. B. The microcirculation of the bone marrow. Anat. Rec. 168, 55–68 (1970).
Klingberg, A. et al. Fully automated evaluation of total glomerular number and capillary tuft size in nephritic kidneys using lightsheet microscopy. J. Am. Soc. Nephrol. 28, 452–459 (2017).
Hanke, R., Fuchs, T. & Uhlmann, N. X-ray based methods for non-destructive testing and material characterization. Nucl. Instrum. Meth. A 591, 14–18 (2008).
Stelzer, E. H. Light-sheet fluorescence microscopy for quantitative biology. Nat. Methods 12, 23–26 (2014).
Schneider, P., Voide, R., Stampanoni, M., Donahue, L. R. & Muller, R. The importance of the intracortical canal network for murine bone mechanics. Bone 53, 120–128 (2013).
Muller, A. M. et al. Expression of the endothelial markers PECAM-1, vWf, and CD34 in vivo and in vitro. Exp. Mol. Pathol. 72, 221–229 (2002).
Chi, J. T. et al. Endothelial cell diversity revealed by global expression profiling. Proc. Natl Acad. Sci. USA 100, 10623–10628 (2003).
Williams, S. P. et al. Genome-wide functional analysis reveals central signaling regulators of lymphatic endothelial cell migration and remodeling. Sci Signal. 10, eaal2987 (2017).
Ramasamy, S. K. et al. Blood flow controls bone vascular function and osteogenesis. Nat. Commun. 7, 13601 (2016).
Faust, N., Varas, F., Kelly, L. M., Heck, S. & Graf, T. Insertion of enhanced green fluorescent protein into the lysozyme gene creates mice with green fluorescent granulocytes and macrophages. Blood 96, 719–726 (2000).
Parfitt, A. M. Osteonal and hemi-osteonal remodeling: the spatial and temporal framework for signal traffic in adult human bone. J. Cell. Biochem. 55, 273–286 (1994).
Lassen, N. E. et al. Coupling of bone resorption and formation in real time: new knowledge gained from human Haversian BMUs. J. Bone Miner. Res. 32, 1395–1405 (2017).
Zebaze, R. M. et al. Differing effects of denosumab and alendronate on cortical and trabecular bone. Bone 59, 173–179 (2014).
Keffer, J. et al. Transgenic mice expressing human tumour necrosis factor: a predictive genetic model of arthritis. EMBO J. 10, 4025–4031 (1991).
Houben, A. et al. beta-catenin activity in late hypertrophic chondrocytes locally orchestrates osteoblastogenesis and osteoclastogenesis. Development 143, 3826–3838 (2016).
Schubert, D., Maier, B., Morawietz, L., Krenn, V. & Kamradt, T. Immunization with glucose-6-phosphate isomerase induces T cell-dependent peripheral polyarthritis in genetically unaltered mice. J. Immunol. 172, 4503–4509 (2004).
Billiau, A. & Matthys, P. Modes of action of Freund’s adjuvants in experimental models of autoimmune diseases. J. Leukoc. Biol. 70, 849–860 (2001).
Frey, O. et al. Regulatory T cells control the transition from acute into chronic inflammation in glucose-6-phosphate isomerase-induced arthritis. Ann. Rheum. Dis. 69, 1511–1518 (2010).
Win, S. J., Kühl, A. A., Sparwasser, T., Hünig, T. & Kamradt, T. In vivo activation of Treg cells with a CD28 superagonist prevents and ameliorates chronic destructive arthritis in mice. Eur. J. Immunol. 46, 1193–1202 (2016).
Wehmeyer, C. et al. Sclerostin inhibition promotes TNF-dependent inflammatory joint destruction. Sci. Transl. Med. 8, 330ra335 (2016).
Irmler, I. M. et al. 18 F-Fluoride positron emission tomography/computed tomography for noninvasive in vivo quantification of pathophysiological bone metabolism in experimental murine arthritis. Arthritis Res. Ther. 16, R155 (2014).
Pastille, E. et al. Transient ablation of regulatory T cells improves antitumor immunity in colitis-associated colon cancer. Cancer Res. 74, 4258–4269 (2014).
Gunal, I., Ozcelik, A., Gokturk, E., Ada, S. & Demirtas, M. Correlation of magnetic resonance imaging and intraoperative punctate bleeding to assess the vascularity of scaphoid nonunion. Arch. Orthop. Trauma. Surg. 119, 285–287 (1999).
Voide, R. et al. The importance of murine cortical bone microstructure for microcrack initiation and propagation. Bone 49, 1186–1193 (2011).
Schneider, P. et al. Ultrastructural properties in cortical bone vary greatly in two inbred strains of mice as assessed by synchrotron light based micro- and nano-CT. J. Bone Miner. Res. 22, 1557–1570 (2007).
Mercier, F. E., Ragu, C. & Scadden, D. T. The bone marrow at the crossroads of blood and immunity. Nat. Rev. Immunol. 12, 49–60 (2011).
Lai, X. et al. The dependences of osteocyte network on bone compartment, age, and disease. Bone Res. 3, 15009 (2015).
Hasegawa, T. et al. Three-dimensional ultrastructure of osteocytes assessed by focused ion beam-scanning electron microscopy (FIB-SEM). Histochem. Cell. Biol. 149, 423–432 (2018).
Smit, T. H., Burger, E. H. & Huyghe, J. M. A case for strain-induced fluid flow as a regulator of BMU-coupling and osteonal alignment. J. Bone Miner. Res. 17, 2021–2029 (2002).
Delgado-Calle, J. et al. MMP14 is a novel target of PTH signaling in osteocytes that controls resorption by regulating soluble RANKL production. FASEB J. 32, 2878–2890 (2018).
Bellido, T. Osteocyte-driven bone remodeling. Calcif. Tissue Int. 94, 25–34 (2014).
Schaffler, M. B., Cheung, W. Y., Majeska, R. & Kennedy, O. Osteocytes: master orchestrators of bone. Calcif. Tissue Int. 94, 5–24 (2014).
Honma, M. et al. RANKL subcellular trafficking and regulatory mechanisms in osteocytes. J. Bone Miner. Res. 28, 1936–1949 (2013).
Chen, K., Pittman, R. N. & Popel, A. S. Nitric oxide in the vasculature: where does it come from and where does it go? A quantitative perspective. Antioxid. Redox Signal. 10, 1185–1198 (2008).
Huang, N. F., Fleissner, F., Sun, J. & Cooke, J. P. Role of nitric oxide signaling in endothelial differentiation of embryonic stem cells. Stem. Cells Dev. 19, 1617–1626 (2010).
Morbidelli, L., Donnini, S. & Ziche, M. Role of nitric oxide in the modulation of angiogenesis. Curr. Pharm. Des. 9, 521–530 (2003).
Birukova, A. A. et al. Prostaglandins PGE2 and PGI2 promote endothelial barrier. Exp. Cell Res. 313, 2504–2520 (2007).
Pai, R. et al. PGE(2) stimulates VEGF expression in endothelial cells via ERK2/JNK1 signaling pathways. Biochem. Biophys. Res. Commun. 286, 923–928 (2001).
Werner, D. et al. Early changes of the cortical micro-channel system in the bare area of the joints of patients with rheumatoid arthritis. Arthritis Rheumatol. 69, 1580–1587 (2017).
Maggiano, I. S. et al. Three-dimensional reconstruction of Haversian systems in human cortical bone using synchrotron radiation-based micro-CT: morphology and quantification of branching and transverse connections across age. J. Anat. 228, 719–732 (2016).
Suen, P. K. & Qin, L. Sclerostin, an emerging therapeutic target for treating osteoporosis and osteoporotic fracture: a general review. J. Orthop. Transl. 4, 1–13 (2016).
Tomlinson, R. E. & Silva, M. J. Skeletal blood flow in bone repair and maintenance. Bone Res. 1, 311–322 (2013).
Green, D. E. & Rubin, C. T. Consequences of irradiation on bone and marrow phenotypes, and its relation to disruption of hematopoietic precursors. Bone 63, 87–94 (2014).
Wright, L. E. et al. Single-limb irradiation induces local and systemic bone loss in a murine model. J. Bone Miner. Res. 30, 1268–1279 (2015).
Yang, B. et al. Effect of radiation on the expression of osteoclast marker genes in RAW264.7 cells. Mol. Med. Rep. 5, 955–958 (2012).
Zhang, J. et al. Differences in responses to X-ray exposure between osteoclast and osteoblast cells. J. Radiat. Res. 58, 791–802 (2017).
Brault, V. et al. Inactivation of the β-catenin gene by Wnt1-Cre-mediated deletion results in dramatic brain malformation and failure of craniofacial development. Development 128, 1253–1264 (2001).
Gebhard, S. et al. Specific expression of Cre recombinase in hypertrophic cartilage under the control of a BAC-Col10a1 promoter. Matrix Biol.: J. Int. Soc. Matrix Biol. 27, 693–699 (2008).
Harada, N. et al. Intestinal polyposis in mice with a dominant stable mutation of the beta-catenin gene. EMBO J. V. 18, 5931–5942 (1999).
Wan, Y. Y. & Flavell, R. A. Identifying Foxp3-expressing suppressor T cells with a bicistronic reporter. Proc. Natl Acad. Sci. USA 102, 5126–5131 (2005).
Lahl, K. et al. Selective depletion of Foxp3+regulatory T cells induces a scurfy-like disease. J. Exp. Med. 204, 57–63 (2007).
Madisen, L. et al. A robust and high-throughput Cre reporting and characterization system for the whole mouse brain. Nat. Neurosci. 13, 133–140 (2010).
Shearing, P. R. & Brandon, N. P. X-ray nano computerised tomography of SOFC electrodes using a focused ion beam sample-preparation technique. J. Eur. Ceram. Soc. 30, 1809–1814 (2010).
Kraff, O. et al. An eight-channel phased array RF coil for spine MR imaging at 7 T. Invest. Radiol. 44, 734–740 (2009).
Rietsch, S. H. G. et al. An 8-channel transceiver 7-channel receive RF coil setup for high SNR ultrahigh-field MRI of the shoulder at 7T. Med. Phys. 44, 6195–6208 (2017).
Jiru, F. & Klose, U. Fast 3D radiofrequency field mapping using echo-planar imaging. Magn. Reson. Med. 56, 1375–1379 (2006).
Johst, S. et al. Time-of-flight magnetic resonance angiography at 7 T using venous saturation pulses with reduced flip angles. Invest. Radiol. 47, 445–450 (2012).
Griswold, M. A. et al. Generalized autocalibrating partially parallel acquisitions (GRAPPA). Magn. Reson. Med. 47, 1202–1210 (2002).
We thank the IMaging Center ESsen (IMCES: https://imces.uk-essen.de) Light Microscopy Unit (LMU), the IMCES Electron Microscopy Unit (EMU) and the Optical Imaging Centre Erlangen (OICE: http://www.oice.uni-erlangen.de) for support with imaging. In addition, we wish to thank R. Burgemeister (Carl Zeiss Microscopy) for support through the Zeiss labs@location program and M. Löffler (DCN, TU Dresden) for his help with X-ray microscopy. J. Kamradt is acknowledged for critical reading of the manuscript. This work was supported by funds from the German Research Foundation (SPP1480 Immunobone ) to M.G., G.S., T.K., A.I.G., A.V., G.K. and M.H.; FZT 111 (Center for Regenerative Therapies Dresden, Cluster of Excellence) to A.I.G.; the Collaborative Research Centre (CRC) 1181 to G.K., M.H. and G.S.; the German Ministry of Education and Research (BMBF NeuroImpa 01EC1403A) to T.K.; and the European Union (EU HEALTH-2013-INNOVATION-1, MATHIAS) to M.G.. The work of G.S. was also supported by the Innovative Medicine Initiative (IMI)-funded project RTCure and the European Research Council (ERC) Synergy grant NanoScope.
The authors declare no competing interests.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Figures 1–9 and Supplementary Tables 1–5
X-ray microscopy (XRM) imaging of a C57BL/6J tibia. X-ray microscopy allows visualization of the compact bone structures of a C57BL/6J tibia, including multiple pores on the bone surface. Optical sectioning of the 3D reconstruction identifies bone pores as canals, traversing the entire compact bone. Zooming into the 3D reconstruction allows the differentiation of bone canals from osteocyte lacunae in the compact bone. Experiments were performed three times individually, with similar results
XRM imaging of C57BL/6J bone canals. Multiple pores can be detected on the surface of a tibia via XRM. While one dominant pore is observed in the posterior metaphyseal bone, numerous smaller pores can be observed all over the bone surface. These pores form canals traversing the entire compact bone. This allows a connection between the bone surface and bone marrow lacuna, shown via an optical journeythrough one exemplary canal. Experiments were performed three times individually, with similar results
LSFM of a simpleClear-treated C57BL/6J tibia. LSFM of a simpleClear optically cleared tibia allows visualization of the entire bone. While there are multiple blood vessels (CD31, red) detected on the bone surface (autofluorescence, grey) forming the periosteum, one dominant vessel is found at the posterior diaphysis. This vessel is entering the bone shaft and ramifies in the bone marrow. Additionally, a central vessel canal can be observed in the centre of the bone marrow, as well as multiple trans-cortical vessels connecting the bone marrow with the bone surface by traversing the compact bone. Experiments were performed six times individually, with similar results
LSFM visualization of a nutrient artery and nutrient sinus. Double staining of CD31 and SCA-1 allows the identification of veins (CD31+SCA-1–, blue) and arteries (CD31+SCA-1+, red) in a simpleClear-treated C57BL/6J tibia. The dominant vessel at the posterior diaphysis shown in Supplementary Video 3 is thereby identified as a nutrient artery, which ramifies in the marrow. The central sinus is exiting the bone shaft (autofluorescence, grey) at the anterior diaphysis. Experiments were performed 15 times individually, with similar results
The periosteal vessel network of a C57BL/6J tibia. High-magnification TPLSM of a simpleClear-treated tibia allows visualization of the arterial (CD31+SCA-1+, red) and venous (CD31+SCA-1–, blue) vessel network in the tissue surrounding the compact bone (SHG, grey). According to their structural orientation, the muscle vascularization can be distinguished from the periosteum, forming a dense vessel network along the bone surface. Experiments were performed five times individually, with similar results
Visualization of arterial and venous connections in a C57BL/6J fibula. High-magnification TPLSM of a fibula showing venous (CD31+SCA-1–, blue) and arterial (CD31+SCA-1+) vessels within the simpleClear-treated bone (SHG, grey). Multiple venous and arterial TCVs are connecting the periosteum to the marrow vascularization within the fibula. In the marrow, the venous central sinus is connected to the sinusoidal network. Additionally, arteries are running along the bone shaft and connected to the sinusoids. Experiments were performed five times individually, with similar results
Blood flow in different cortical vessels. The blood flow in different types of cortical vessel is visualized by intra-vital TPLSM of LysM-EGFP tibiae (SHG, grey). TCVs, nutrient arteries (NAs) and the bone-exiting sinus differ not only in diameter but also in speed of blood flow (rhodamine dextran, red) and cell transport (EGFP, green) (n = 25 TCV, 7 NA and 5 central sinus scans)
Intra-vital imaging of G-CSF mobilization in a LysM-EGFP mouse. Intra-vital imaging of the tibial surface (SHG, grey) shows blood flow (rhodamine dextran, red), but only rarely cell transport (EGFP, green), through TCVs under untreated conditions. About 20 min after application of G-CSF, an increase in cell transport by the bloodstream, as well as active cell migration against the direction of blood flow, can be observed in TCVs and a nutrient artery. Experiments were performed five times individually, with similar results
Location of osteoclasts at the endosteum and in trans-cortical canals. LSFM imaging of a simpleClear-treated CX3CR1-cre;tdTomato tibia shows high numbers of osteoclasts (red) located along the endosteum of the diaphysis. Furthermore, osteoclasts can be found in the vascularized (CD31, turquoise) trans-cortical canals (TCCs) preferentially located in the centre of the compact bone (autofluorescence, grey). Widening of TCCs at these locations may indicate bone remodelling and formation of new TCVs emanating from existing TCCs. Experiments were performed eight times individually, with similar results
Osteoclasts remodelling trans-cortical canals. The generation of SHG signals (grey) via TPLSM imaging enables the visualization of compact bone tissue. The widened TCC area associated with osteoclast location (CX3CR1-cre;tdTomato, red) lacks SHG signals, suggesting the formation of a resorption lacuna where the compact bone tissue is dissolved by the adjacent osteoclast. Experiments were performed three times individually, with similar results
7T MRI imaging of a human shank. 3D reconstruction of 7T MRT data allows visualization of a human shank and identification of specific structures including muscle tissue (brown), compact bone (grey), arteries (red) and veins (blue). Optical clipping of the tibia shows a nutrient artery entering the bone shaft and a central sinus running parallel in the bone cavity. Experiments were performed twice individually, with similar results
Blood flow egression from human bone. Surgical exposure of the human femoral neck shows blood egression from multiple pores on the bone surface. Experiments were performed twice individually, with similar results
Whole-mount-stained and simpleCLEAR-cleared human femoral neck. Whole-mount staining of a human femoral neck with CD31 (turquoise) and α-smooth muscle actin (α-SMA, red) permits visualization of arterial and venous vessels in the human tissue sample. A large artery entering the compact bone (autofluorescence, grey) from the periosteum and a small artery traversing trabeculae in the bone marrow can be observed. Experiments were performed three times individually, with similar results
TCVs in the human femoral neck. Staining of veins (CD31+SCA-1–, turquoise) and arteries (CD31+SCA-1+, red) in the human femoral neck enables visualization not only of the Haversian vascular system in compact bone (autofluorescence, grey), but also of the presence of dTCVs directly connecting the bone marrow with the periosteal vessel network. Experiments were performed four times individually, with similar results
Osteoclast–osteocyte interaction in TCVs. 3D rendering of a confocal scanned histological bone section shows an osteocyte in the compact bone, identified by its characteristic dendritic-like morphology (phalloidin, green, DAPI blue). The osteocyte dendrites are connected to an osteoclast (Cx3cr1-cre;tdTomato, red) located in a TCV (phalloidin, green). Experiments were performed six times individually, with similar results
About this article
Cite this article
Grüneboom, A., Hawwari, I., Weidner, D. et al. A network of trans-cortical capillaries as mainstay for blood circulation in long bones. Nat Metab 1, 236–250 (2019). https://doi.org/10.1038/s42255-018-0016-5
Historical Biology (2020)
Impaired bone microarchitecture in distal interphalangeal joints in patients with primary hypertrophic osteoarthropathy assessed by high-resolution peripheral quantitative computed tomography
Osteoporosis International (2020)
The transcortical vessel is replacement of cortical capillary or a separate identity in diaphyseal vascularity
Anatomy & Cell Biology (2020)
Cytometry Part A (2020)
Annals of the New York Academy of Sciences (2020)