• Open Access

Decreased Ratio of CD8+ T Cells to Regulatory T Cells Associated with Decreased Survival in Dogs with Osteosarcoma

Authors

  • B.J. Biller,

    1. Department of Clinical Sciences, Animal Cancer Center, Colorado State University, Fort Collins, Colorado. Portions of this material were submitted for presentation in abstract form to the 32nd Annual Conference of the Veterinary Cancer Society, Austin, TX, October 2009.
    Search for more papers by this author
  • A. Guth,

    1. Department of Clinical Sciences, Animal Cancer Center, Colorado State University, Fort Collins, Colorado. Portions of this material were submitted for presentation in abstract form to the 32nd Annual Conference of the Veterinary Cancer Society, Austin, TX, October 2009.
    Search for more papers by this author
  • J.H. Burton,

    1. Department of Clinical Sciences, Animal Cancer Center, Colorado State University, Fort Collins, Colorado. Portions of this material were submitted for presentation in abstract form to the 32nd Annual Conference of the Veterinary Cancer Society, Austin, TX, October 2009.
    Search for more papers by this author
  • S.W. Dow

    1. Department of Clinical Sciences, Animal Cancer Center, Colorado State University, Fort Collins, Colorado. Portions of this material were submitted for presentation in abstract form to the 32nd Annual Conference of the Veterinary Cancer Society, Austin, TX, October 2009.
    Search for more papers by this author

Corresponding author: Dr Steven W. Dow, DVM, PhD, DACVIM, Department of Clinical Sciences, Colorado State University, Fort Collins, CO 80523; e-mail: sdow@colostate.edu.

Abstract

Background: Increased numbers of regulatory T cells (Treg) and decreased ratios of CD8+ T cells to Treg have been shown to correlate with decreased survival times (ST) in humans with certain malignancies. A possible connection between Treg and ST in dogs with cancer has not been investigated previously.

Hypothesis: The purpose of this study was to compare numbers of Treg and T lymphocyte subsets in dogs with osteosarcoma (OSA) to those of healthy dogs and to determine whether pretreatment values were associated with disease-free interval or with ST. We hypothesized that Treg numbers would be increased in dogs with cancer and that dogs with a high percentage of Treg would have a poorer prognosis.

Animals: Twelve client-owned dogs with appendicular OSA were entered into a prospective clinical trial. Twenty-two healthy dogs were used as controls.

Methods: The percentages and numbers of Treg and CD4+ and CD8+ T cells in blood, lymph nodes, and tumors were determined with flow cytometry and compared between dogs with OSA and control dogs.

Results: Dogs with OSA had significantly fewer circulating CD8+ T cells and significantly more Treg compared with healthy dogs. The CD8/Treg ratio also was significantly lower in dogs with OSA compared with control dogs. In dogs with OSA, a decreased CD8/Treg ratio was associated with significantly shorter STs.

Conclusions: These data support a role for Treg in the immune control of canine OSA and suggest that determination of the CD8/Treg ratio may be useful for assessing outcomes.

Abbreviations:
OSA

osteosarcoma

Treg

regulatory T cells

Alterations in the normal balance of T lymphocyte subpopulations are common in advanced cancers. These changes may occur in multiple compartments such as blood, lymph node and tumors and include decreases in effector T cells (CD4+ and CD8+) and increases in regulatory T cells (Treg).1–4 Treg, a distinct subset of the CD4+ T-cell population, are of particular concern in cancer patients because they play an important role in negatively regulating the development of antitumor immune responses. Normally functioning to prevent harmful autoimmune responses, Treg have been shown to directly suppress immune responses to tumors in mouse tumor models.5–7 Treg-induced immune suppression also appears to occur in humans as evidenced by clinical studies that correlate high Treg numbers with impaired immune function.4,8–10 In people with metastatic melanoma, Treg are present in high numbers in tumor-draining lymph nodes where they directly inhibit the function of infiltrating T cells.10

In addition to an association with immune dysfunction, Treg numbers correlate with clinical outcomes for many tumors. In people with head and neck cancer, the percentage of circulating Treg increases with stage of disease and is predictive of early recurrence.11 The magnitude of Treg infiltration within tumors is of prognostic importance for human malignancies such as melanoma, lymphoma, and carcinomas of breast and ovaries.1,12,13 For some tumors, the ratio of CD8+ T cells to Treg in tumor tissues is a better predictor of outcome than measurement of Treg alone.14 Therefore, there is considerable evidence to support the prognostic value of Treg measurement as well as their critical role in regulating immune responses to tumors.

There has been little research investigating the role of Treg in diseases of companion animals. In a study of cats with chronic viral infections, Treg were identified based on surface expression of CD4 and CD25 (the high-affinity interleukin-2 receptor) and were found to be increased compared with healthy animals.15,16 Similar studies in dogs have been limited by the availability of Treg-specific cell surface markers. However, we previously reported that intracellular detection of FoxP3, a transcription factor expressed in high levels within human and murine Treg, can be used to identify canine Treg.17–20 FoxP3 protein expression in feline and porcine CD4+CD25+ lymphocytes also was recently reported, suggesting that this marker may be a useful tool for Treg identification in multiple species.21,22

The results of our previous work show that Treg are increased in dogs with cancer compared with healthy dogs and suggest that tumor-specific differences may influence Treg numbers in different tumor types.17,23 However, too few dogs with osteosarcoma (OSA) were included in these studies to determine the importance of Treg in this disease. In addition, evaluation of the association between Treg numbers and clinical outcome has not been previously reported for dogs. Therefore, we designed a prospective study to evaluate T cell subpopulations in the blood, lymph nodes, and tumors of dogs with OSA and to investigate their association with clinical outcomes.

Material and Methods

Inclusion Criteria

Dogs presenting to the Colorado State University Veterinary Teaching Hospital (CSU-VTH) with histopathologically confirmed OSA of the appendicular skeleton were eligible for enrollment in this prospective clinical trial. All dogs underwent amputation of the affected limb followed by an adjuvant chemotherapy protocol containing either carboplatin alone or alternating carboplatin and doxorubicin. Dogs receiving carboplatin chemotherapy were treated at a dosage of 300 mg/m2 body surface area for a total of 4 treatments over a 12-week period. Dogs receiving multidrug therapy were treated with alternating doses of doxorubicin and carboplatin according to a previously described protocol.24 After completion of chemotherapy, all dogs were reevaluated every 8 weeks by physical examination and thoracic radiography to monitor for the development of metastatic disease. Dogs were excluded from the study if they had radiographic evidence of pulmonary metastatic disease at the time of diagnosis or if they had received previous chemotherapy or radiation therapy within 3 weeks of amputation. This study was approved by the Animal Care and Use Committee at Colorado State University.

Sample Collection

For analysis of Treg and effector T cells in normal dogs, blood and lymph node samples were collected from 22 healthy, age-matched dogs. When possible, lymph node samples were obtained by fine-needle aspiration of a popliteal lymph node. For analysis of dogs with OSA, blood was collected the day before surgery and again 18–24 hours after amputation. During surgery, a small section of the tumor and a regional lymph node were collected. In addition, a second lymph node sample was obtained by fine-needle aspiration from a distant (nontumor draining) lymph node.

Sample Preparation

Peripheral blood mononuclear cells (PBMC) were obtained from blood samples collected in EDTA tubes after lysis of red blood cells. PBMC or lymphocytes were added at a concentration of 5 × 105 to 1 × 106 per well in 96-well round bottom plates and then immunostained for surface expression of CD4 and CD8 using appropriate concentrations of FITC-conjugated anti-canine CD4 mAB (clone YKIX302.9)a and Alexa 647-conjugated anti-canine CD8 mAB (clone YCATE55.9)b anti-dog CD8 following the method described previously.25 Lymph node tissues were processed before immunostaining by washing through a 70-μm cell strainer with a PBS buffer containing 2% fetal bovine serum; tumor tissues were incubated in a collagenase buffer for 20 minutes before processing through the strainer. The cells collected from these specimens then were added to 96-well plates and immunostained as described above.

Immunostaining for FoxP3 expression was done as described previously.17 Briefly, after washing to remove unbound antibodies, intracellular detection of FoxP3 was performed with a cross-reactive, PE-conjugated murine FoxP3 antibodyc and the buffers that accompany the antibody staining kit. A directly conjugated rat IgG2A antibody was used as the isotype control.

Flow Cytometric Analysis

Flow cytometry was performed with a CyAn ADP flow cytometerd and Summit software for data analysis. Analysis gates were set on the live lymphocyte population based on typical forward and side scatter characteristics.26 The percentage of Treg was calculated by determining the percentage of FoxP3+CD4+ cells within the CD4+ T-cell population. The percentages of CD4+ and CD8+ T cells also were determined. Absolute numbers of Treg, CD4+ and CD8+ T cells in peripheral blood were calculated based on the total lymphocyte count determined from a CBC performed with an automated cell counter.

Statistical Analyses

Patients were censored from disease-free interval (DFI) and survival time (ST) analysis if they died from an unrelated cause or were lost to follow-up. DFI was defined as the time from amputation until the development of metastasis and ST was defined as the time from amputation until death. Median DFI and ST were calculated by the Kaplan-Meier product limit method and compared between groups by log rank analysis. Variables assessed for value as predictors of outcome included percentages and absolute numbers of Treg, CD4+ T cells and CD8+ T cells, the CD8/Treg ratio, and type of postoperative chemotherapy used. Mean percentages and numbers of Treg, T effector cells and the mean CD8/Treg ratio were compared between normal and OSA-bearing dogs by an unpaired, 2-tailed Student's t-test. Comparisons between multiple groups were done by ANOVA with Tukey's multiple means comparison. Statistical calculations were performed by a commercial software package.e A P value of <.05 was considered significant for all analyses.

Results

Study Patients

Twelve patients fulfilling entry criteria were included in this study; all underwent amputation of the affected limb. Adjuvant chemotherapy consisted of 4 cycles of carboplatin in 6 dogs and 3 cycles each of alternating carboplatin and doxorubicin in 6 dogs. The overall median DFI was 235 days (95% CI, 164–306 days) and the overall median ST was 281 days (95% CI, 192–377 days). There were no significant differences in DFI and ST in dogs receiving single-agent carboplatin chemotherapy versus dogs receiving alternating therapy (P= .90 and P= .98 for DFI and ST, respectively). Two dogs received a metronomic chemotherapy protocol consisting of low doses of oral cyclophosphamide, piroxicam, and doxycycline at the time of detection of metastatic disease; the ST for each of these patients was <130 days. All 12 OSA dogs were euthanized because of clinical signs attributable to progression of metastatic disease including respiratory distress, lethargy and inappetence (n = 11), and hindlimb paralysis (n = 1). T cell analyses from the dogs with OSA were compared with those of 22 age-matched control dogs. The average age of dogs with OSA was 8.5 years (range, 5–14 years), whereas the average age of the control dogs was 8.0 years (range, 5–11 years); this difference was not statistically significant (P= 1.0).

Treg Are Significantly Increased in the Blood of Dogs with OSA

Blood and lymph node samples were collected from control dogs and dogs with OSA and evaluated by flow cytometry to determine the percentage of CD4+ and CD8+ T cells and Treg. Representative dot plots for Treg analysis from the blood of a healthy dog and a dog with OSA are shown in Figure 1. For this analysis, the percentage of Treg was expressed as the percentage of FoxP3+CD4+ cells within the overall CD4+ T cell population. For blood samples, absolute numbers of Treg and CD4+ and CD8+ T cells were determined by multiplying the percentage of each cell type by the total number of lymphocytes determined from the patient's CBC.

Figure 1.

 Flow cytometric analysis of regulatory T cells in blood of dogs. Peripheral blood mononuclear cells from a healthy dog (A) and a dog with osteosarcoma (B) were immunostained for expression of CD4 and FoxP3, as described in “Materials and methods.” (C) Shows staining of blood from the same dog as in (A) with an irrelevant isotype control antibody to control for specificity of FoxP3 staining. The percentages of lymphocytes within each quadrant are given.

Both the percentage and absolute number of Treg were significantly higher (P= .01) in the pretreatment blood samples of dogs with OSA than in blood samples from healthy dogs (Fig 2). The mean percentage of Treg in the dogs with OSA was 6.4 (± 2.6) versus 4.5 (± 1.6)% in the normal dogs. The total number of Treg in peripheral blood also was significantly increased (50 ± 20 versus 32 ± 14 cells/μL, P= .01). Interestingly, despite removal of their primary tumor, the percentage of Treg was similar in pre and postoperative blood samples (6.4 ± 2.1 versus 6.6 ± 1.5%)(6.4 + 2.1 versus 6.6 + 1.5%) from the dogs with OSA.

Figure 2.

 Relative and absolute numbers of regulatory T cells (Treg) in peripheral blood in dogs with osteosarcoma (OSA) and healthy control dogs. Blood of control dogs (n = 22) and dogs with OSA (n = 12) was analyzed with flow cytometry to determine the percentage and absolute numbers of Treg, as described in “Materials and methods.” The mean percentage (long horizontal line) of Treg in the blood of normal dogs and dogs with OSA was plotted in (A) and the mean of the absolute numbers (long horizontal line) of Treg was plotted in (B). Error bars show SD. Both the percentage and absolute numbers of Treg were significantly increased (P= .01) in dogs with OSA compared with healthy control dogs as assessed by unpaired, 2-tailed Student's t-tests.

Because Treg preferentially accumulate within lymph tissues in some types of cancer, we investigated whether this was the case in the dogs of this study. When possible, fine-needle aspirates of a peripheral lymph node distant from the tumor and the lymph node nearest to the tumor were collected at the time of limb amputation and percentages of Treg were determined. Percentages of Treg were similar in the regional lymph node and in distant lymph nodes in the OSA dogs and also similar to those found in healthy dogs (Fig 3).

Figure 3.

 Percentages of regulatory T cells (Treg) in lymph nodes of dogs with osteosarcoma (OSA) and healthy dogs. The percentage of Treg was determined in lymph node (LN) aspirates from normal dogs (n = 9) and dogs with OSA (n = 12). A portion of the LN nearest the tumor (regional LN) was available for 10 of the 12 dogs with OSA; in addition, a 2nd sample was obtained by fine-needle aspiration from a distant LN for 7 of the 12 dogs with OSA. The mean percentages (long horizotal lines) of Treg in the regional LN, the distant LN and in LNs from normal dogs were compared. Error bars show SD. No significant differences in Treg percentages were found when assessed by a 1-way ANOVA with Tukey's multiple means comparison.

The Distribution of T Cell Subsets Is Altered in Dogs with OSA

To determine whether dogs with OSA had pretreatment changes in the distribution of their T-lymphocyte subsets, we compared pretreatment percentages and overall numbers of CD4+ and CD8+ T cells in the blood to levels in healthy dogs. Neither the percentage nor total number of CD4+ T cells was significantly different in control dogs versus dogs with OSA dogs (Table 1). However, we found that dogs with OSA had significantly fewer CD8+ T cells, both on a relative and absolute basis, than the control dogs (Table 1). Despite this decrease, total lymphocyte counts were within the normal range for all but 2 of the 12 dogs in this study (not shown). No significant differences in CD4+ and CD8+ T cell percentages were observed in the lymph nodes of OSA versus healthy dogs (Table 1).

Table 1.   Lymphocyte subpopulations in control dogs and dogs with OSA.
ParameterSiteControl (Mean ± SD)OSA (Mean ± SD)
  • DLN, distant lymph node; RLN, regional lymph node; OSA, osteosarcoma.

  • *

    P < .05

% CD4+ T cellsBlood48.8 (10.3)56.8 (15.2), P= .08
%CD8+ T cellsBlood22.0 (9.4)15.2 (8.9), P= .04*
Total lymphocytes (cells/μL)Blood1,614 (615)1,404 (796), P= .45
# CD4+ T cells (cells/μL)Blood752 (373)745 (391), P= .96
# CD8+ T cells (cells/μL)Blood382 (195)218 (205), P= .03*
% CD4+ T cellsDLN38.8 (10.6)37.6 (11.1), P= .89
% CD8+ T cellsDLN13.1 (4.5)13.8 (3.4), P= .59
% CD4+ T cellsRLNNA39.7 (11.1)
% CD8+ T cellsRLNNA10.5 (4.8)
% CD4+ T cellsTumorNA19.2 (10.0)
% CD8+ T cellsTumorNA8.6 (6.4)

The Ratio of CD8 to Treg in the Blood Is Significantly Lower in Dogs with OSA Compared with Control Dogs

We next examined whether the CD8/Treg ratio in the blood and lymph nodes of the dogs in this study was different than in healthy dogs. This was done by dividing the percentage of blood or lymph node CD8+ T cells by the percentage of blood or lymph node Treg to determine CD8/Treg ratios for each site. [Correction added after online publication 28 July 2010: ratio calculation corrected.] When we compared the CD8/Treg ratio in the lymph nodes of dogs with OSA to that of control dogs, no significant differences were found (Fig 4A). However, the CD8/Treg ratio in the blood was markedly lower in dogs with OSA compared with controls (2.8 ± 1.7 versus 5.1 ± 2.2, P= .004) (Fig 4B).

Figure 4.

 Comparison of CD8/regulatory T cells (Treg) ratios in healthy dogs and dogs with osteosarcoma (OSA). (A) The CD8/Treg ratio (long horizontal line) in lymph nodes was determined by dividing the mean percentage of CD8+ T cells in the regional lymph nodes of 10 dogs with OSA by the mean percentage of Treg, and then compared with the CD8/Treg ratio (long horizontal line) in the lymph nodes of normal dogs (n = 9). No significant differences were found when assessed by an unpaired, 2-tailed Student's t-test. (B) The blood CD8/Treg ratio (long horizontal line) was calculated by dividing the percentage of CD8+ T cells in blood by the percentage of Treg and then compared between healthy dogs (n = 22) and dogs with OSA (n = 12). Error bars show SD. The CD8/Treg ratio in blood was significantly lower (P= .004) in dogs with OSA compared with control dogs.

Treg Percentages Are Highest in Tumor Tissues

Treg generally are present in highest levels within tumors compared with other compartments such as the spleen, lymph nodes, and blood circulation. To investigate this relationship, we compared the percentages of Treg, CD4+, and CD8+ T cells in tumor samples to samples collected from tumor-draining and distant lymph nodes and blood. When the percentage of Treg in each compartment was evaluated by ANOVA there was a significant difference between groups (P= .001), with a mean percentage of Treg in tumors of 21(±8.0)% followed by a mean of 18(±5.0)% in distant lymph nodes, 14(±5.0)% in regional lymph nodes and 6.4(±2.6)% in the blood.

OSA Dogs with a Low CD8/Treg Ratio Have Significantly Shorter Survival than Dogs with a Higher CD8/Treg Ratio

Both high numbers of Treg and low CD8/Treg ratios have been found to be negatively associated with prognosis in human cancer patients.1,11–14 Therefore, we investigated whether Treg or the CD8/Treg ratio was associated with the DFI and ST for the dogs with OSA in our study. To do these analyses, we first determined the mean Treg percentage and mean CD8/Treg percentage ratio in the tumor, blood, and regional lymph nodes of the OSA patients. Kaplan-Meier product limit analyses for DFI and ST times then were performed by assigning outcome based on whether dogs were above or below the Treg and CD8/Treg ratio mean values. Although dogs with a high percentage of Treg in their blood and tumor tissues had a shorter ST than dogs with low percentages of Treg, these differences were not statistically different. However, when outcome was analyzed based on a high or low CD8/Treg ratio in the blood, we observed a significantly shorter (P= .05) ST for dogs with low ratios compared with those with high ratios (Fig 5A). The DFI also was decreased in dogs with low CD8/Treg ratios; but this difference was not statistically significant (P= .10) (Fig 5B). Potential confounding factors such as tumor location, serum alkaline phosphatase, and type of chemotherapy were not associated with a high or low CD8/Treg ratio.

Figure 5.

 Correlation of the CD8/regulatory T cells (Treg) ratio in blood to survival and disease free interval in dogs with osteosarcoma (OSA). Kaplan-Meier plot depicting survival times (A) and disease-free interval (B) in dogs with OSA with a blood CD8/Treg ratio higher (n = 6) or lower (n = 6) than the mean value.

Discussion

Several important findings emerged from this study. First, when compared with a control population of healthy dogs, the blood of dogs with OSA contained significantly increased numbers of Treg, decreased numbers of CD8+ T cells, and a markedly decreased CD8/Treg ratio. In contrast, the percentages of Treg and T cell subpopulations in lymph nodes were similar to those of healthy dogs. We also observed a negative association between a low CD8/Treg ratio in the blood and the DFI and ST, suggesting that the CD8/Treg ratio may be a useful prognostic tool.

Previous work has established that Treg are increased in most human patients with solid tumors. In nearly all reported investigations in mice and humans, the tumor itself is the site with the highest level of Treg accumulation, whereas Treg frequencies in blood and lymph nodes vary considerably depending on tumor type.27,28 This was also true for the dogs of our study in which the highest percentages of Treg were found in tumors compared with lymph nodes and blood. The lack of published studies of Treg in human OSA patients precludes direct comparison with our data, but the Treg percentages reported here are consistent with those of other bone tumors such as Ewing's sarcoma.29

In agreement with previously published data, we found that the number of Treg in the blood of dogs with OSA was significantly greater than in healthy dogs.17,23 In contrast, however, the percentages of Treg within lymph nodes for dogs with OSA did not significantly differ from that of control animals. Although it is possible that a nondraining lymph node was inadvertently sampled rather than the tumor-draining lymph node, histopathology indicated that lymph node metastasis occurred in only 1 of the 12 dogs in this study (not shown). Interestingly, the percentage of Treg was slightly higher in a distant lymph node than in the tumor-draining lymph node for this particular patient. This suggests that Treg may accumulate in tumors and in blood but not in lymph nodes in dogs with OSA. Clearly, larger numbers of dogs will need to be evaluated to better understand differences in Treg distribution among various compartments in patients with OSA.

The inverse association between numbers of Treg and effector T cells is a well-described characteristic of solid tumors in human patients.9,13,14,27,30,31 Although there were no differences between CD8/Treg ratios in the lymph nodes of dogs with OSA and in healthy dogs, there was a striking difference in the blood. In a previous study of dogs with different malignancies, we found that the CD8/Treg ratio was significantly decreased in tumor-bearing dogs compared with healthy dogs.23 In that report, however, when dogs were evaluated by tumor type, those with a diagnosis of sarcoma (which included OSA, soft tissue sarcoma, hemangiosarcoma, and melanoma) had a CD8/Treg ratio similar to that of healthy dogs. The results of the current study are in contrast to that report, perhaps reflecting differences in the numbers of dogs with OSA available for each study. Whether the decrease in CD8+ T cells observed here was a direct consequence of high Treg numbers or because of factors that act on each cell type is not known. These findings suggest, however, that systemic alterations in the normal balance of T lymphocyte subsets exist in dogs with cancer before initiation of anticancer therapies.

Another key finding of this study was the correlation between a low blood CD8/Treg ratio and decreased ST. To our knowledge, this is the first report of an association between CD8/Treg levels in the blood and outcome in cancer patients; however, the human literature contains numerous reports describing the correlation between Treg, tumor-infiltrating lymphocytes and prognosis.13,27,28,32 Interestingly, neither Treg nor CD8+ T-cell levels in blood were significantly correlated with outcome when assessed separately. One intriguing explanation for the negative correlation between the blood CD8/Treg ratio and ST is that dogs presenting with a low CD8/Treg ratio have more advanced-stage disease. Although thoracic radiographs were negative for pulmonary metastases before amputation, advanced imaging techniques such as computed tomography might have better identified metastatic disease in these patients. Alternatively, a low pretreatment CD8/Treg ratio could signal a greater degree of tumor-associated immune evasion and a greater likelihood of developing metastatic disease more rapidly. The inability to monitor Treg and T cell subsets over time in this study prevented assessment of the relationship between these cells and the onset of detectable metastatic disease.

In conclusion, we found that there are significantly more Treg and fewer CD8+ T cells in circulation in dogs with OSA than in healthy dogs. Furthermore, dogs with a low pretreatment CD8/Treg ratio had a significantly decreased ST compared with dogs with higher CD8/Treg ratios. Owing to the small number of dogs, these results must be considered preliminary. The study of Treg in dogs remains limited by the lack of reagents available to evaluate the suppressive function of this unique T cell population. Additional studies should include evaluation of Treg and effector T cells over time in larger numbers of dogs with OSA treated with a standardized protocol.

Footnotes

aSerotec, Raleigh, NC

bSerotec

cMouse FoxP3 staining kit, eBioscience, San Diego, CA

dCyan ADP flow cytometer, Dako-Cytomation, Fort Collins, CO

eGraphPad Prism version 5 for Windows, GraphPad Software, San Diego, CA

Acknowledgments

The authors thank Dahlia Rice for her excellent technical assistance. This study was supported by Grant #778 from the American Kennel Club Canine Health Foundation.

Ancillary