Further observations on mechanisms of bone destruction by squamous carcinomas of the head and neck : The role of host stroma

Mechanisms of bone invasion by squamous carcinomas of the head and neck have, been investigated using fresh tumours and established tumour cell lines in an in vitro bone resorption assay with 45Ca-labelled mouse calvaria. Fresh tumours regularly resorb bone in vitro. Activity is consistently reduced by indomethacin. The tumours release E2 prostaglandins (PGE2) in amounts sufficient to account for -50% of the bone resorption observed. Small amounts of non-prostaglandin (indomethacin-resistant) osteolytic factors are also produced. Control non-neoplastic tissues show a variable capacity to resorb bone in vitro; PGE2 levels in these tissues may be related to their content of inflammatory cells. Tumour cell lines also resorb bone in vitro but, for most lines, activity is not significantly blocked by indomethacin and PGE2 levels are generally insufficient to account for the osteolysis observed. Non-prostaglandin bone resorbing factors thus predominate. It is concluded that most squamous cancers of the head and neck are osteolytic in vitro and release a mixture of prostaglandin and non-prostaglandin factors which stimulate osteoclastic bone resorption. These factors are derived from both neoplastic and stromal elements, and are "tumour-associated" rather than "tumour-specific". In vitro bone resorption and prostaglandin release does not correlate with pathological features of the tumour or with post-operative survival.

Osteoclasts accumulate at sites of bone invasion by squamous carcinomas of the head and neck, and appear to play an important role in the destructive process (Carter, 1982). Prostaglandins are known stimulants of osteoclastic activity, and raised levels of extractable prostaglandin-like material were demonstrated by bioassay in a series of squamous cancers from the head and neck region (Bennett et al., 1980). Subsequent work showed that fresh tumour tissues and tumour cell lines were osteolytic in vitro and that bone resorption could be partly blocked by indomethacin (Tsao et al., 1981). These studies have been extended and information is now presented on the identification and quantitation of the prostaglandins involved and the contribution made by host tissues as a source of osteolytic factors. Observations on bone resorption by xenografts of squamous carcinomas have been reported separately (Tsao et al., 1983).
Bone resorption assay The methods used were based on the procedure devised by Reynolds (1968) and have been fully described by us (Tsao et al., 1981). In brief, calvaria were dissected from 5 to 7 day old BALB/c mice previously injected with 45CaCl2, and cultured on metal grids in modified Bigger's medium, supplemented with heat-inactivated rabbit serum and antibiotics, at 37°C in 5% CO2 in air. Paired half-calvaria were used, one half of each serving as a control. After a preliminary incubation period of 24h to equilibrate calcium exchange between bone and culture medium, the calvaria were cultured for 3 days either with fresh tissue fragnents or with various test and control media-see below. Release of 45Ca was estimated by a liquid scintillation system. The percentage of isotope released from each bone was calculated and osteolysis was expressed in a standard manner as the ratio of the % of 45Ca release from test and control cultures. The values of each bone resorption ratio were recorded as the mean + s.e. of 4 pairs of bone cultures.
In vitro osteolysis by fresh tissues Twenty-nine squamous carcinomas were obtained from patients admitted to the Royal Marsden Hospital for major surgery. The tumours were from the following primary sites: tongue 6, hypopharynx 6, larynx 4, floor of mouth 3, oropharynx 3, maxillary antrum 1 and nasal septum 1 together with nodal metastases from primary carcinomas of the tongue 3, larynx 1 and nasal septum 1.
Control (non-neoplastic) tissues were taken from macroscopically uninvolved regions near the resection lines of the same surgical specimen. Observations were also made on normal breast skin from patients admitted to Queen Mary's Hospital, Roehampton and the Royal Marsden Hospital for reduction mammaplasties.
Two methods were used to study in vitro osteolysis. The fresh tissues were either directly cocultured with radiolabelled bone, or incubated alone with culture medium and the conditioned medium then assayed for osteolytic activity. 1) Co-culture experiments Tumour fragments (_ 1 mm3) were weighed, washed with culture medium and incubated with the 45Ca-labelled calvaria for 3 days. Three-A pieces of tumour tissue (net weight 9.5-28.2mg, mean 15.8) were either placed round the bone or cultured on a separate grid. The distance between the bone and the tissues in both instances was 3mm. Five ml of culture medium was used in each tumour-bone culture. Full details of culture conditions were given previously (Tsao et al., 1981) The release of 45Ca in test and control bone cultures was estimated and bone resorption ratios calculated. Control cultures consisted of bone incubated without tumour. Control (non-neoplastic) tissues were treated in the same way.
2) Conditioned-medium experiments Fresh tissues were incubated alone on metal grids for 3 days ( -20 mg/5 ml culture medium). Cell-free supernatants were prepared after incubation by Millipore filtration (0.45 pm). Culture medium incubated alone under the same conditions served as a control. Media were stored at -40°C if not assayed immediately, and aliquots of 1.5 ml were preserved for prostaglandin assays.
Indomethacin treatment Indomethacin (Sigma) was included in some of the culture media at a concentration of 1 pg ml -1.
In vitro osteolysis by tumour cell lines Thirteen cell lines were examined. Ten of the lines (LICR/HN 1-10) were established by Easty et al. (1981a, b) and the other 3 (LICR/HN 12, 13 and 15) by one of us (S.-W.T.) using the same techniques. Validation of the tumour cell lines in terms of their karyotypes, ultrastructure and growth as xenografts is provided in the papers by Easty and her colleagues. Subconfluent cultures of carcinoma cells were incubated with modified Bigger's medium (supplemented with 5% heat-inactivated rabbit serum and antibiotics) for 24 h at 37°C in 5% CO2 in air (40 ml per culture flask, 174 cm3, Nunc). Cell free supernatants were prepared by filtration (0.45pm Millipore filter). After adjusting the pH to the same value as control medium (-7.4), the conditioned media were added to cultures of 45Calabelled calvaria. The pH of the culture medium was measured again at the end of experiment.
Eight control fibroblastoid cell lines were also established from primary explant cultures of squamous carcinomas which had been overgrown by fibroblastoid cells. They were examined in the same way as the carcinoma cell lines.
Indomethacin was included in some of the culture media at a concentration of 1 pgml--.

Radioimmunoassay ofprostaglandins
Prostaglandins present in culture media were extracted in ether, and purified by thin-layer chromatography (Eastman & Dowsett, 1976). Total recovery after these steps was 50%, calculated by adding a radiolabelled prostaglandin tracer. The purified prostaglandins were then quantitated by standard radioimmunoassay using two antisera raised separately against PGE2 and PGF2a from Steranti laboratories and [3H]-labelled PGE2 (16OCimM-1) and PGF2a (l80 CimM-1) from Amersham International Ltd. Technical details are given in Tsao (1982).

Histology
Fresh soft tissues were fixed after culture in formol saline, embedded in paraffin wax, processed by routine histological techniques and stained with H and E. Incubated calvaria were fixed in formol saline, decalcified in EDTA, embedded in methacrylate resin and cut at 1-2 pm.

Observations with fresh tumour tissues
In vitro osteolysis by tumours co-cultured with bone Osteolytic activity in 16 unselected squamous carcinomas is summarized in Table I. Twelve of the tumours examined showed bone resorption with 45calcium release test/control ratios . 1.3. Differences in pH between test and control culture media were small (usually <0.1). No consistent association was established between in vitro osteolysis and a tumour's site, size, degree of differentiation or the presence of bone invasion or lymph node metastases. Indomethacin was added to 14 of the co-culture experiments; the concentration used (1 ugml -1) has been shown to suppress prostaglandin synthesis to insignificant levels in this assay (<1 ngml -'see Tsao (1982). The results are included in Table I. Osteolysis was blocked to varying degrees by indomethacin and, in 10 tumours (71%), the level of inhibition was near to or greater than 50% of the total osteolytic activity. Inhibition was statistically significant in 6 tumours (P < 0.05, Student's t-test). This consistent but incomplete blocking of tumour-associated bone resorption by indomethacin suggests that both prostaglandins and indomethacin-resistant (non-prostaglandin) osteolysins are produced.
In order to show that indomethacin affected the synthesis of new osteolytic factors rather than the action of osteolytic factors already released into the medium, the drug was also added to 3 tumourconditioned media after instead of during the incubation period with tumour. Osteolysis was only slightly affected, confirming that indomethacin acts mainly by inhibiting prostaglandin formation.
In vitro osteolysis and prostaglandin release in tumour-conditioned media Synchronous observations on in vitro osteolysis and prostaglandin release were made with tumour-conditioned media (3 days incubation) from a further 12 tumours. The results are shown in Table II. Varying degrees of bone resorption were detected. Prostaglandins, especially PGE2, were demonstrated in the conditioned media but the levels varied widely (42-120 ngml-1, mean + s.d. = 32.9 + 36.7 ngml-1) and bore no direct relationship to the osteolysis observed. The amount of PGE2 required to stimulate in vitro bone resorption in this system is -5 to lOngml-l (Tsao, 1982) so the PGE2 detected in the tumour- conditioned media would account for at least part of the osteolytic activity demonstrated. PGF2a was present in only small amounts and is unlikely to stimulate bone resorption to any extent. Histological structure was well-preserved in the tumours after 3 days co-culture with calvaria. All the fragments contained intact carcinoma. Foci of necrosis, inflammation and fibrosis were also present.
Observations with control tissues Two sets of control tissues were examined: uninvolved tissues from the head and neck removed from surgical specimens at the same time as the tumours, and normal skin from reduction mammaplasties.
Tissues from the head and neck Osteolytic activity in 12 paired tumours and non-neoplastic tissues is shown in Table III. Ten of the control tissues resorbed bone in vitro. Activity was usually greater in the corresponding tumour, but the differences were small and reached statistical significance in only one instance (P<0.01, Student's t-test). Prostaglandins were measured in conditioned media from 4 of the pairs. Similar levels of PGE2 were found in tumour and control tissues in 3 of the 4 pairs, the fourth showing an excess of PGE2 in the tumour. Levels of PGF2a followed no consistent pattern in the material examined. Sections from the cultured tissues showed no evidence of tumour but a variable amount of focal necrosis and mixed inflammatory infiltrates with mononuclear cells, lymphocytes and sometimes polymorphs.

Observations with carcinoma cell lines
In vitro osteolysis Eleven carcinoma cell lines were tested for bone resorbing activity using conditioned culture media (24 h incubation) obtained from subconfluent monolayer cell cultures. The results are shown in Table IV. Varying degrees of osteolysis were detected. Cell lines LICR/HN 2, 4, 6, 7, 10, 12 and 13 were moderately active and LICR/HN 1, 3, 5, and 9 were less active. The variations in activity noted among individual cell lines in repeated assays probably reflect variations in culture conditions such as cell density and batches of serum used. The pH differences between the test and control media were low (<0.1).
Indomethacin (1 4gml -) was added to cultures of five of the active cell lines (LICR/HN 2, 4, 6, 7 and 12). Control medium contained the same amount of indomethacin. The results are shown in Table IV. In contrast to the findings with freshly   (Tsao, 1982)  Histological changes were examined in paired test 0.89+0.04 and control calvaria from experiments with tumour -127+007 cell lines LICR/HN 1, 2, 4 and 6. All slides were -1.27 + 0.07 coded beforehand, but differences between the two in original surgical specimens. groups were readily apparent. Bone incubated in control media felt firm when handled, and sections subsequently showed smooth intact trabeculae with only occasional multinucleate osteoclasts. Bones able I), no consistent effects incubated with media from the carcinoma cell lines uced osteolysis seen with cell felt soft and sections showed a loss of bone nd 12 is not statistically substance with thin and irregularly outlined trabeculae and increased numbers of osteoclasts on or near the internal bone surface. The osteoclastic )f prostaglandins In 5 response was most marked in calvaria exposed to ,ICR/HN 1, 3, 4, 5 and 6), media from tumour cell lines LICR/HN 4 and 6; atrations and osteolytic these two lines showed high levels of in vitro lined in the same culture osteolysis and produced predominantly nonire shown in Table V. The prostaglandin osteolysins (see Table IV). taglandins detected were low GF2a <1.3ngml-1) and did levels of osteolytic activity Observations with control (fibroblastoid) cell cultures Fibroblastoid cell lines were examined for in vitro osteolysis and release of prostaglandins.
The results are shown in Table VI. Osteolytic activity was detected in 4/8 cultures. Differences in pH between test and control media were insignificant (pH <0.01). Prostaglandins were determined in media from two osteolytically active lines (FB 1 and 6) and two inactive lines (FB 2 and 4). The concentrations detected were low (PGE2 < 2.0 ngml -'; PGF2a <0.7 ngml-1) and were too small to stimulate bone resorption in vitro.

Discussion
Three main groups of findings have emerged from this work. Freshly excised squamous carcinomas of the head and neck resorb bone in vitro. Osteolytic activity, mediated by local osteoclasts, is consistently reduced (though not abolished) by indomethacin. The tumours release E2 prostaglandins (PGE2): the levels are variable but are sufficient to account for at least 50% of the bone resorption observed in most instances. Cell lines derived from squamous cancers are also osteolytic in vitro but differ from fresh tumours in two respects. Bone resorption by most of the cell lines is largely unaffected by indomethacin and production of PGE2 is low, suggesting that prostaglandins are mainly implicated in osteolysis by fresh tumours. No differences in activity were observed between tumours treated pre-operatively by irradiation and/or cytotoxic drugs and tumours treated by primary surgery. Non-neoplastic tissues show a variable capacity to resorb bone in vitro, indicating that osteolysis is not a tumour-specific activity.
Several human tumours produce prostaglandins in vitro, notably carcinomas of the breast (Bennet et al., 1975(Bennet et al., , 1976(Bennet et al., , 1977aPowles et al., 1976;Dowsett et al., 1976;Greaves et al., 1980) kidney (Atkins et al., 1977) and large intestine (Bennett et al., 1977b). Prostaglandins from human tumours resorb bone in vitro, and extensive studies of prostaglandinmediated osteolysis have been made with experimental tumours (Tashjian et al., 1972(Tashjian et al., , 1982Voelkel et al., 1978;Tashjian, 1978). Activation of local osteoclasts has been observed in the in vitro model systems (Schelling et al., 1980). Investigators have been unable to establish a linear relationship between levels of prostaglandins released and the mass of bone resorbed, and it has become apparent that additional non-prostaglandin osteolytic factors are also involved. Particular discussion continues in relation to two topics-the cellular origin of the osteolysins and their nature, especially with respect to non-prostaglandin substances.
(1) The complementary use of fresh tissues and cell lines provides a starting point for separating tumour and host cells as potential sources of osteolytic agents. In the experiments described here with paired tumour and control tissues, most of the fresh control tissues resorbed bone in vitro and released prostaglandins at levels not significantly less than the corresponding tumour. The absence of tumour in control tissues was always confirmed histologically, but they were invariably inflamed and had usually been exposed to pre-operative irradiation and/or chemotherapy; although such tissues form acceptable controls for the tumours from the same patients, they clearly cannot be regarded as normal. By contrast, histologically normal breast tissue with no necrosis or inflammatory infiltrates showed negligible osteolysis and prostaglandin release. Mononuclear macrophages can synthesise prostaglandins and resorb bone in vitro (Mynett et al., 1975;Humes et al., 1977;Mundy et al., 1977;Kahn et al., 1978;McArthur et al., 1980) and there is evidence that bone resorption in rheumatoid arthritis and peridontal inflammation may be partly due to local production of prostaglandins (Robinson et al., 1975;Harris, 1978). Some of the control fibroblastoid cell lines resorbed bone in vitro but PGE2 levels were always low. Taken together, these findings suggest an association between the capacity of non-neoplastic tissues to manifest prostaglandinmediated bone resorption in vitro and the presence of inflammatory cells within them. This is a difficult topic to pursue in intact tissues as the inflammatory infiltrates would need to be quantified and each of the cell constituents accurately identified.
(2) The nature of non-prostaglandin (indomethacin-insensitive) osteolysins remains obscure. Candidates include ectopic parathyroid hormone, osteoclast activating factors and certain other ill-defined products (Mundy et al., 1974a, b;Josse et al., 1981;Nimberg et al., 1982). The nonprostaglandin osteolysin associated with squamous cancers of the head and neck is uncharacterized at the present time.
No consistent relationship has emerged between bone resorbing activity and prostaglandin release in vitro, clinicopathological features of the tumours (including the presence of bone invasion in the surgical specimens) and post-operative survival of the patients. A similar lack of correlation has been reported in patients with breast cancer (Dady et al., 1981) though prognostic significance for raised prostaglandin levels has been claimed by other workers (Fitzpatrick & Stringfellow, 1979).