Direct transformation from quiescence to bone formation in the adult periosteum following a single brief period of bone loading

The concept of resorption preceding formation in a coupled response is well established as the normal sequence of remodeling in adult bone. So prevalent is this concept, however, that the idea of the direct activation of osteogenic modeling in normal adult bone is often ignored. This experiment documents the direct transformation of the normal, quiescent, adult periosteum to active bone formation. The osteogenic stimulus was provided by a single short period of dynamic loading. Periosteal activation and the production of new bone within 5 days of loading was unaccompanied by resorption or the presence of osteoclasts. We therefore conclude that an adult resting periosteum can become directly converted to formation as a physiologic response to an appropriate osteogenic stimulus without the need for resorption. To distinguish this process from remodeling we suggest it be called renewed modeling. It is notable that a single short exposure to an “osteogenic” loading regime can influence the full cascade of cellular events between quiescence and active bone formation.

INTRODUCTION NE OF FROST'S CLASSIC CONTRIBUTIONS to OUT under-0 standing of bone remodeling was his suggestion that resorption and formation were the result of discrete "packets" of cellular activity, sequential in their appearance and coupled in their effect.''' In adults with a stable or declining bone mass the normal sequence of events is activation followed by resorption and formation (ARF). Naturally, within solid pre-existing cortical bone this is the only sequence of events possible. On the surfacecz) the concept of resorptive or formative drifts has become established to explain the process of modeling during growth. However, such drifts are sometimes considered solely a phenomenon of g r~w t h , '~,~) not to contribute to the pattern of remodeling behavior in normal adult It has been proposed that in the bone "envelopes" where drift is important during growth, normal ARF turnover becomes of principal importance in the adult.'4) Although the ARF sequence is predominant in adult bone, there is a danger of it becoming thought of as the only modeling and remodeling mechanism normally possible on adult bone surfaces. Parfitt's) has suggested that should the quiescent surface of adult bone be able to transform directly to a forming surface without intervening resorption, this would constitute "a special form of modeling." In previous experiments in which we used mechanical loading to stimulate an increase in bone mass in adult animals, there was no histologic evidence of resorption having preceded formation regardless of whether the new bone type formed was woven, laminar (plexiform), primary osteonal, or circumferential lamellar. w.') However, although these bones were fluorescently labeled during the relevant period, the animals were killed weeks or months after the onset of adap-648 PEAD ET AL.
tive change and so the appearance of the periosteum at the time of the stimulus was not known. Evidence of transitory resorption could therefore have been obliterated. For this reason we undertook the experiments reported here in which we observed the response of a quiescent periosteal surface of an adult bone to the application of a stimulus of established osteogenic potential.
The osteogenic stimulus was produced by subjecting the ulna of skeletally mature roosters to a load regime to which they were not adapted. Using a daily period of similar loading in this preparation in adult turkeys we had previously been able to demonstrate a dose-response relationship between peak strain magnitude and change in bone cross-sectional area after 8 weeks. (8) In the experiment reported here we examined the changes in the periosteum 4-7 days after exposure to a stimulus that in our previous studies we had established would produce a long-term increase in bone mass.

MATERIALS AND METHODS
A total of 14 skeletally mature hybrid roosters (Gaffus domesticus), aged 30 weeks, were used to provide functionally isolated, externally loadable, ulna preparations.
As previously described, the preparation is made in anesthetized animals by a transverse osteotomy through the ulna's proximal and distal submetaphyseal region. The diaphyseal surface of each osteotomy is covered by a stainless steel cap, which is then pierced by a Steinman pin. These pins are positioned to protrude through both dorsal and ventral skin surfaces (Fig. 1). Following recovery from anesthesia the protruding sections of the pins are clamped together with external fixators in order to prevent accidental loading of the bone shaft. The preparation was made in the left ulna only. Although experimentally more convenient to have a bilateral comparison, when this procedure is carried out bilaterally some animals have difficulty in balancing and may apply loads to the preparation by bringing the external fixators into direct contact with the ground. When only one wing is prepared it is held close to the body and the preparation is not loaded in this way.
When it is required to apply loads to the bone preparation, the external fixators are removed and the pins engaged between the forks of an Instron machine, which applies a cyclic compressive load between them. Previously, ulnae from similar roosters with the same surgical preparation had been placed in vitro in the forks of the loading machine. These ulnae had rosette strain gauges bonded to each of the three faces of the midshaft of the bone. These gauges allow the recording of the strains engendered by a range of loads, and from these,(Io) the principal strains at any part of the loading cycle may be determined for every FIG. 1. This diagram shows the avian wing bones after the preparation has been made. Note the Steinman pins piercing the ends of the osteotomized bone, which are covered by stainless steel caps. One fixator is shown (above the wing), but in the experiment there are two, one above and one below the wing. When the preparation is loaded these fixators are removed and the pins engaged in the loading forks. point on the surface of the bone at the midshaft. Thus we were able to choose a load for the experiment that placed a peak principal compressive strain of 0.0027-0.003 on the midshaft of the prepared section of bone. Because the bone is slightly curved the strain differs across the cross section of the bone, but at no point on the surface did the peak principal compressive strain exceed 0.003. This figure corresponds to the peak principal strains recorded in similar animals during normal vigorous exercise. (") In this experiment, 48 h after the surgical preparation the left ulna preparation was subjected either to no load (control birds) or to a single period of dynamic loading. In each case the intact right ulna was left undisturbed. The load was cycled at 1 Hz, producing peak strains of up to 0.003 in compression at a maximum strain rate of 0.05 per s. Both the peak strain magnitude and the strain rate were thus within the physiologic range, but as in previous studies the strain distribution was different from normal. Since there were 300 loading cycles the loading period lasted for 5 minutes.
Prior to surgery each animal was given two intravenous injections of 1.5 ml oxytetracycline (50 mg/ml) 6 days apart. This procedure was carried out at the same time of day on each occasion. The second label was given 3 days prior to the surgery. Radiographs of the wing bones were also taken. The animals were divided into three groups, each of which contained both experimental and control animals. These groups were then treated according to the following protocol. Following sacrifice, a transverse section was made across the midshaft of the ulna on the prepared and contralateral (intact) sides of each animal. A full transverse block of the ulna was taken from either side of the initial section. The proximal block was fixed in 10% buffered formal saline, decalcified in EDTA, and embedded in wax. The distal block was fixed in absolute ethanol and embedded in methyl methacrylate (British Drug House Chemicals Ltd., Poole, Dorset). Each block was embedded so that the side toward the original midshaft section was uppermost, and thus both wax and methylmethacrylate sections were cut from similar areas.
Sections 5 pm thick were cut from the wax blocks and stained with hematoxylin and eosin or soluble blue-phosphotungstic acid."') Unstained sections 13 gm thick were prepared for fluorescent microscopy from the methyl methacrylate-embedded blocks. Some of these plastic-embedded sections were stained with new universal bone stain (NUBS).(l3)

Histologic assessment of modeling and remodeling changes
Histologic assessment of modeling and remodeling changes was carried out on two complete transverse sections per animal. The state of the periosteum was assessed at 16 locations around the bones' circumference. These locations were defined by a 16-point "star" grid ( Fig. 2) placed over the section and a measurement made at each point where the periosteum and endosteum was transected by a grid line. The cross-sectional shape of the avian ulna is that of a rounded triangle. By aligning the grid with the angles of the triangle it could be placed in a similar position over each section.
For each section the following measurements were made at the point of intersection of the grid lines with the bony surfaces.
1. Periosteal or endosteal width was the distance from the bone surface to the outer surface of the fibrous layer of the periosteum or endosteum measured in micrometers.
2. Osteogenic cell layers were the number of layers of cells between the mineralized bone surface and the fibrous layer of the periosteum or endosteum. The bone lining

FIG. 2.
A diagram showing the shape of the cross section of the avian ulna with the 16-point counting grid superimposed upon it. All the measurements referred to in the text were made at the points of intersection of the star with the bone surface. The grid was oriented over the section with respect to the angles between the faces of the ulna. Measurements were commenced at the angle of the dorsal and ventral faces and were thus made at equivalent points on the circumference for each section. cells were considered a layer in this calculation. These cells were recognized by their flattened elongated nuclei and their position directly adjacent to the bone surface. Frequently these cells were the only clearly discernible layer of the endosteum. On formative bone surfaces active osteoblasts were recognized adjacent to the surface by their columnar shape. The layers of cells between the surface layer and the inner fibrous layer of the periosteum had a more rounded appearance and were considered mostly preosteoblasts, although it is probable that this area also contains cells of other lineages.
3. Osteoid thickness was the thickness of the osteoid seam measured using the soluble blue-phosphotungstic acid stain. Seams of thickness less than 3 pm in width were not measured.
4. Depth of new bone was the distance from the preexisting bone surface to the outermost new bone. This included the height of fronds of bony tissue in the osteogenic layer. Since the new bone was not present at all points on the circumference of each section, the number of grid points at which new bone was present were recorded.
Periosteal width and number of cell layers in the periosteum are somewhat difficult to measure in areas where the new bone is in the form of fronds. However in this study these measurements were made by taking the tips of the bone fronds as the base of the periosteum.
All measurements were made using an eyepiece microm-eter and placing the scale at right angles to the bone surface. Measurements of the periosteal and endosteal surfaces using bright-field microscopy were carried out at a magnification of x 125. Osteoid seams were measured at x 250. Fluorescent microscopy was performed using blue light (365 nm) at a magnification of x 100.

RESULTS
The absence of fluorochrome label in the undecalcified sections indicated that there was no active formative drift at the bone surfaces immediately prior to the experiment.
The lack of open growth plates on the radiographs and in the postmortem dissections confirmed the skeletal maturity of the animals. The periosteum of the intact and nonloaded ulnae were typical of a quiescent periosteum with an outer fibrous layer and an inner cellular layer (Fig. 2). The thicker fibrous layer consisted of loose connective tissue with collagen fibers and a few fibroblasts. The deeper cellular layer consisted of lining cells whose nuclei were flattened and spindle shaped lying on the outer surface of the mineralized bone. The average number of cell layers between the bone surface and the fibrous layer on the intact bones was 1.1 f 0.03 and the periosteal thickness 10 f 0.3 pm. There was no evidence of new bone formation. The sections from the bones that were prepared but not loaded   showed a similar histology, and there was no significant difference in the periosteal width (1 1 f 0.4 pm) or number of osteogenic cell layers (1.2 f 0.1) between this group and the intact ulnae (unpaired t-test). Ulnae that were prepared but not loaded were indistinguishable from their intact contralateral controls.
All the loaded ulnae showed an increase in periosteal activity compared with the contralateral intact controls. The periosteal activity of the loaded ulnae in groups 2 and 3 (killed 6 and 7 days after loading) was greater than those in group 1 (killed 5 days after loading). The periosteum of the loaded ulnae showed a response that varied in magnitude around the circumference of the bone. The histologic appearance of this periosteum could be divided into five types of increasing activity: 1. A resting periosteum as outlined earlier for the intact and unloaded bones with flattened lining cells adjacent to the bone (Fig. 3).
2. Active osteoblasts. The cells adjacent to the bone surface are not flat but larger with rounded nuclei. The depth of the periosteum is greater than in type 1.
3. Proliferative periosteum. The cellular layer adjacent to the bone surface is deeper than in type 2. The cells adjacent to the bone surface are larger than those closer to the fibrous layer, but all have rounded nuclei. These cells were identified as osteoblasts and preosteoblasts. Osteoblasts adjacent to the bone surface in these areas appear to be actively depositing osteoid. The connective tissue layer is thicker than in type 1 or 2, and the border between fibrous and cellular layers is less obvious (Fig. 4).
4. New bone formation. The periosteum is similar to that in type 3. There is an osteoid seam a maximum of 7 pm thick covering mineralizing new bone formation. Osteocytes are present in round lacunae within the osteoid and newly formed bone (Fig. 5).

5.
New bone fronds. Fronds of newly woven bone covered with osteoid project at right angles to the bone surface. In the areas between the fronds there are many plump round cells; however, there is still a periosteal layer similar to that in type 3 surrounding the area beyond the tips of the bony fronds.
Using a paired t-test, the results for periosteal width and number of osteogenic cell layers from each counting point on the prepared ulnae were compared with those from the contralateral intact side for the individual animals. In the loaded ulnae in all the groups there was a significantly greater periosteal width and number of osteogenic cell layers than in the intact control side. There was no significant difference in the periosteal width and the number of osteogenic cell layers between the prepared and intact sides in the control groups, except in group 3 (See Table 1). There was no new bone formation on any of the intact ulnae or on the prepared nonloaded (control) ulnae, whereas there was new bone on all those that had been loaded.
The only detectable difference between intact and contralateral prepared bones that had not been loaded was a small but statistically significant difference in the thickness of the periosteum in one animal in group 3 (periosteal width, pm): intact 8 f 0.4, prepared 1 1 f 0.6; p < 0.01).
The prepared but not loaded bones in this group were in fact disuse preparations of 10 days' duration.
The results from the prepared loaded ulnae are compared with the prepared control (nonloaded ulnae) in Table 2. In addition to the increase in periosteal width and number of osteogenic cell layers seen in all the loaded bones, new bone was observed on all loaded bones whereas no new bone was observed on the prepared control ulnae. (See Table 2.

)
The endostea showed no osteogenic changes that could be related to the presence or absence of loading. The en-  Fig. 2. The periosteum is thickened, with active osteoblasts and preosteoblasts on the bone surface. A similar cement line is also visible (arrows). Note the difference between this and Fig. 2.  FIG. 5. Photomicrograph (hematoxylin and eosin stain). The periosteum shows new bone formation, which has been laid down on the original bone surface (white arrows). Note the active periosteum on the bone surface with rounded osteoblasts and a thickened connective tissue layer. This was taken from a prepared bone, the contralateral to that used in Fig. 2. Note the similar cement line to that in Fig. 2 (black arrows). dosteum consists of bone lining cells with a sparse layer of fibrous tissue between them and the bone marrow. The entire endosteurn is between 5 and 7 pm thick. In two of the sections from loaded preparations there was evidence of small areas of frondlike new bone formation, but examination of the undecalcified sections corresponding to these areas showed that this modeling had taken place prior to labeling with tetracycline and thus prior to the start of the study. The only other section in which endosteal modeling was present was one of the nonloaded preparations (group C sacrificed on day 10). In this section two areas of the endosteum showed active osteoclastic activity, and at a separate area there was a proliferation of osteoblasts and new bone formation.
Osteoid seams were examined in all the different areas of periosteal activation. In areas of resting periosteum the osteoid could be seen as a line less than 3 pm thick. Seams of this thickness were not measured. In areas with a prolif-

DISCUSSION
The results presented here demonstrate clearly that, in these animals at least, exposure of the adult bone to a single period of osteogenic loading can cause the periosteal surface to transform directly from quiescence to active bone formation. There appears to be no need for resorption to precede or accompany this process.
According to one theory of activation proposed for normal ARF remodeling, the bone surface becomes physically exposed by a change within the lining cell layer. This process allows the recruitment of osteoclast precursor cells and their subsequent fusion at the bone surface to produce any cells within the periosteum that resemble osteoclasts suggests that such exposure did not occur or that it did not have that effect. The bone surface appears to have transformed directly from the quiescent surface seen in the controls to an active state of formation. To achieve such a direct transformation, lining cells must have reverted directly back to osteoblasts, dedifferentiated into osteoblast precursor cells that subsequently produce osteoblasts, or been displaced by osteoblasts produced superficial to them.
Direct bone activation has been demonstrated in other experimental systems,(1S.16' confirming that a direct activation of the osteoblast lineage is possible. However, these studies have involved the effects of prostaglandins or estrogens, which are known to be able to directly affect osteoblastlike cells,(L5) or such situations as fracture, in which there is a strong pathologic stimulus toward bone formation by osteoblasts. In the experiment reported here there was no pre-existing trend toward formation, and the active o~t e o c l a s t s . '~~~~) In this experiment the absence of osteogenic stimulus was a perturbation of the bone's nor-ma1 mechanical input, the strains involved of the same order of size as those produced during natural loading but with a different distribution.
A remarkable feature of these results is the effect on bone cell behavior, lasting 5-7 days, that may be induced by a single 5 minute period of loading. In previous experiments we determined that prolonged isolation results in endosteal resorption and incompletely infilled haversian remodeling, whereas interruption of this isolation by short daily periods of loading produces a strain-related prevention of resorption and increase in bone mass.11o) From our present experiments it can be seen that these long-term effects are the cumulative result of single periods of stimulation each one of which is capable of stimulating the full cascade of events necessary to produce a "packet" of new bone formation. This cascade involves a large number of steps, including strain transduction, osteocyte and periosteal osteoblast activation, and preosteoblast recruitment and proliferation.
In this preparation the loads applied were controlled to produce strains within physiologic levels. However, as in our previous studies the distribution of the strain across the bone was different from normal. We have proposed that it is this difference in strain distribution that is the basis of the adaptive stimulus that leads to change in bone architecture in this preparation. '10 The new bone deposited as a result of the stimulus was clearly woven. Woven bone has been seen as part of the long-term modeling reaction to a change in strain distribu-tionI1' and is the principal bone type in the new bone associated with the architectural change involved in the cumulative response to short daily periods of loading in our previous experiments.'1o) The association of woven bone with both trauma and bone pathology has led some investigators to classify all woven bone formation on diaphyseal shafts as pathologic.'lg zo) Clearly, woven bone is common in many pathologic situations characterized by rapid new bone formation, but we disagree with the view that classifies all woven bone formation as indicative of pathology. First, in previous experiments in which adaptive new bone formation has been stimulated, areas of woven bone have been incorporated into the new cortex, where they clearly play a structural role.'' ") [It is true that disorganized cortical bone incorporated into a cortex in this way is often preferentially replaced by secondary osteones (Fig. 6), but this probably indicates a nonoptimal alignment of the tissue rather than any original pathology.] Second, woven bone does occur as part of the normal process of bone formation. Although most evident in fastgrowing animals, particularly those producing laminar or plexiform bone, significant areas of woven bone have been reported as constituents of normal structural bone in both humans and Third, in this experiment and in previous chronic experiments involving the same preparation,l5 woven bone formation has always been induced by strain levels within the physiologic range. These strain levels, with the remarkably few repetitions we employ, are clearly unlikely t o cause macro-or microdamage. We postulate that the intense osteogenic stimulus we observe using natural strain levels with abnormal strain distribution is a consequence of the exposure of the bone to this abnormal strain distribution. This response is not therefore pathologic but an extreme example of the adaptive response that normally matches structural mass and architecture to the characteristics of each bone's loading.
Thus in our view the presence of woven rather than lamellar bone indicates the speed of the osteogenic response rather than any specific pathologic association. We therefore propose that it should be considered a potential part of the normal adaptive modeling and remodeling process that is invoked when there is a demand for rapid new bone formation. We suggest that the phenomenon of direct periosteal new bone formation without preceding resorption on adult bone surface be termed renewed modeling. "Renewed modeling" distinguishes it from remodeling and adequately describes a process that may happen at any age with the production of any bone type as a normal response to a change in the mechanical demands on the bony tissue.

CONCLUSIONS
In response to a suitable osteogenic stimulus, the quiescent adult periosteum can transform directly to active bone formation. No resorption need precede or accompany this event.
In the absence of any normal load-related stimulus, a single period of loading of 5 minutes' duration engendering strains in the normal physiologic range but with an abnormal distribution can initiate a series of events resulting in recruitment and division of cells of the osteoblastic lineage and formation of periosteal new bone evident 7 days later.