Functional and genetic studies of isolated cells fromparathyroid tumors reveal the complex pathogenesisof parathyroid neoplasiaYuhong Shia, Joyce Hogueb, Darshana Dixitb, James Kohb,1, and John A. Olson, Jr.a,1,2

aDivision of General and Oncologic Surgery, Department of Surgery, University of Maryland School of Medicine, Baltimore, MD 21201; and bDivision ofSurgical Sciences, Department of Surgery, Duke University Medical Center, Durham, NC 27710

Edited* by Robert C. Gallo, Institute of Human Virology, University of Maryland School of Medicine, Baltimore, MD, and approved January 3, 2014 (receivedfor review November 4, 2013)

Parathyroid adenomas (PAs) causing primary hyperparathyroidism(PHPT) are histologically heterogeneous yet have been historicallyviewed as largely monotypic entities arising from clonal expansionof a single transformed progenitor. Using flow cytometric analysisof resected adenomatous parathyroid glands, we have isolatedand characterized chief cells, oxyphil cells, and tumor-infiltratinglymphocytes. The parathyroid chief and oxyphil cells produceparathyroid hormone (PTH), express the calcium-sensing receptor(CASR), and mobilize intracellular calcium in response to CASRactivation. Parathyroid tumor infiltrating lymphocytes are T cellsby immunophenotyping. Under normocalcemic conditions, oxyphilcells produce ∼50% more PTH than do chief cells, yet display sig-nificantly greater PTH suppression and calcium flux response toelevated calcium. In contrast, CASR expression and localizationare equivalent in the respective parathyroid cell populations. Anal-ysis of tumor clonality using X-linked inactivation assays in apatient-matched series of intact tumors, preparatively isolatedoxyphil and chief cells, and laser-captured microdissected PA speci-mens demonstrate polyclonality in 5 of 14 cases. These data dem-onstrate the presence of functionally distinct oxyphil and chiefcells within parathyroid primary adenomas and provide evidencethat primary PA can arise by both clonal and polyclonal mecha-nisms. The clonal differences, biochemical activity, and relativeabundance of these parathyroid adenoma subpopulations likelyreflect distinct mechanisms of disease in PHPT.

tumor heterogeneity | endocrine neoplasia

The parathyroid glands maintain serum calcium concentrationwithin a narrow physiological range through regulated syn-

thesis and secretion of parathyroid hormone (PTH) (1). Para-thyroid neoplasia results in inappropriate secretion of PTH byone or more glands, leading to hypercalcemia and the diseaseprimary hyperparathyroidism (PHPT) (2). To date, research inthe field has focused on histologic and molecular profiling ofparathyroid tumors and the investigation of calcium sensing indispersed cells from parathyroid tumors and normal bovineparathyroid glands. Few studies have characterized the individualcellular constituents of human parathyroid tumors, and no pub-lications have reported live-cell functional evaluation of the dif-ferent cellular subtypes observed in parathyroid tumors.Parathyroid adenomas, the most common cause of PHPT, are

considered clonal proliferations of a transformed parathyroidcell that has acquired proliferative or survival advantage due toone of several genetic abnormalities including the PRAD1translocation or mutations in the genes encoding menin, P53,and P27 (3-7). Regardless of type, most authors consider para-thyroid tumors clonal (8, 9) although intratumoral heterogeneityhas been observed and polyclonality of microdissected para-thyroid adenomas has been reported (10, 11).Although most data support mutation-driven clonal expansion

of parathyroid tumors, an alternative model for the origin ofparathyroid tumors is that abnormal calcium sensing by para-

thyroid cells leads to abnormal secretion of PTH initially, followedby proliferation of parathyroid cells in response to chronic de-mand for increased PTH (12, 13). An abnormal calcium–PTH setpoint has been well-described in aggregate dispersed cells fromparathyroid adenomas, and most reports attribute the impairedset point in these cells to decreased expression of the calcium-sensing receptor (CASR) (14, 15) or more recently to alteredexpression of downstream molecules linked to CASR signaling,including RhoGEF and RGS5 (16, 17). In contrast to clonal ex-pansion following genomic tumor-initiating events, attenuatedcalcium responsiveness could be expected to drive polyclonalproliferation in the parathyroid gland.To investigate the composition of parathyroid tumors, we

sought to characterize a series of parathyroid adenomas at thecellular and functional level. In this study, we report the isolationand characterization of chief cells, oxyphil cells, and lymphocytespresent in parathyroid adenomas and histologically normalparathyroid glands from patients with PHPT. Our results showthat parathyroid adenomas removed from patients with PHPTare composed of functionally and genetically distinct oxyphil andchief cells and have varying amounts of infiltrating lymphocytes.The chief and oxyphil cells within parathyroid adenomas havediffering ability to respond to changes in ambient calcium andproduce PTH although CASR expression is comparable betweenthese parathyroid cell subtypes. Further, we show that a signifi-cant proportion of PHPT patients have polyclonal tumors. Therelative abundance, functional behavior, and clonal origin of

Significance

Parathyroid adenomas, the main cause of primary hyperpara-thyroidism (PHPT), are thought to result from clonal expansionof tumor cells and to be insensitive to normal calcium feedbackdue to the loss of the calcium-sensing receptor (CASR). Utilizingflow cytometric analysis to isolate and individually studyoxyphil cells, chief cells, and lymphocytes from resected para-thyroid tumors and glands, we now report previously un-recognized heterogeneity in these tissues with respect tocalcium responsiveness, CASR expression, and clonal origin ofparathyroid tumors. Such heterogeneity of parathyroid ade-nomas likely reflects the complex etiopathogenesis and clinicalheterogeneity of PHPT.

Author contributions: Y.S., J.K., and J.A.O. designed research; Y.S. and J.H. performedresearch; Y.S., J.H., D.D., J.K., and J.A.O. contributed new reagents/analytic tools; Y.S., J.K.,and J.A.O. analyzed data; and Y.S., J.K., and J.A.O. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.1J.K. and J.A.O. contributed equally to this work.2To whom correspondence should be addressed. E-mail: jolson@smail.umaryland.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1319742111/-/DCSupplemental.

3092–3097 | PNAS | February 25, 2014 | vol. 111 | no. 8 www.pnas.org/cgi/doi/10.1073/pnas.1319742111

distinct parathyroid-cell subpopulations in parathyroid tumorslikely reflect multiple alternative etiologies of PHPT.

ResultsFlow cytometric analysis of dispersed parathyroid adenoma cellsrevealed discrete subpopulations of cells resolved on the basisof visible-light diffraction detected by the in-line axis [forwardscatter (FSC-A)] and orthogonal axis [side scatter (SSC-A)]sensors. After excluding nonviable and doublet cells, three dis-tinct populations, initially designated P3, P4, and P5, were clearlyand reproducibly resolved as separate peaks in a bivariate SSC-A/FSC-A contour plot (Fig. 1A, Upper Left). Ultrastructuralanalysis by transmission electron microscopy of preparativelyisolated P3, P4, and P5 cells showed clear morphological dif-ferences between the three populations (Fig. 1A). Based uponsize and morphology, P3 cells were provisionally identifiedas peripheral blood lymphocytes. P4 cells displayed highlyinterdigitated plasma membranes, sparse mitochondria, andprominent intracellular secretory granules, all features ofparathyroid chief cells (18). P5 cells were large and containedround nuclei, moderate peripheral chromatin condensation,and numerous mitochondria densely packed throughout thecytoplasm. These characteristics are consistent with theparathyroid oxyphil cell type (19). We used the same sortingstrategy to examine the distribution of these cells types amongcells dispersed from single adenomas from 20 patients withPHPT (Fig. 1B). In these samples, the proportion of lym-phocytes present in the dispersed parathyroid adenoma sam-

ples tested ranged from 2% to 31.9%. The relative abundanceof chief cells ranged from 9.4% to 57.8% of the total viablepopulation. The relative abundance of oxyphil cells rangedfrom 0.9% to 59.8%. The distribution of these cells did notcorrelate with any feature of the disease including sex, age,preoperative calcium or PTH, gland weight, or histology.To compare the cellular composition of histologically normal

and neoplastic parathyroid tissue, we examined a series of fivepaired tumor and histologically normal (but physiologically sup-pressed) parathyroid gland biopsy specimens using flow cytometryto assay for the presence of the three cell populations. The relativedistribution of oxyphil cells, chief cells, and lymphocytes was re-markably consistent between normal tissues. In contrast, adenomasfrom different patients have highly variable distribution patterns(Fig. 2 A and B). These data indicate that oxyphil cells, chief cells,and lymphocytes are present in both normal tissue and adeno-matous tissue, but the relative abundance of these subpopulationscan be markedly variable between different adenomas.To confirm the designation of P3 cells as lymphocytes, we

probed P3 cells with the leukocyte common antigen CD45 an-tibody using standard flow-cytometry methods. P3 cells are uni-formly CD45-positive and are indistinguishable in this assay frompatient-matched isolated peripheral blood lymphocytes. Sub-sequent cell-surface marker studies revealed that the majority ofP3 cells are CD45+/CD3+/CD19−/CD24−/CD44+, an immuno-type consistent with peripheral T cells (Fig. 3A). These resultsunambiguously confirm that P3 cells are lymphocytes. CD3staining on parathyroid adenoma tissue sections showed threepredominant patterns: widely separated single cells (Fig. 3B,Pt1), clustered cells extravasated into the parenchyma of thegland but adjacent to vascular structures (Fig. 3B, Pt2), and fo-cal, densely packed infiltrates (Fig. 3B, Pt3). To verify that the P3cells are tumor-infiltrating lymphocytes as opposed to contami-nating peripheral blood mononuclear cells (PBMCs) adventi-tiously carried over from the tumor vasculature, we examined theCD4+/CD8+ ratio in the P3 population. In contrast to the 2:1ratio of CD4+ to CD8+ cells found in the bloodstream, CD8+

effector cells are highly overrepresented in the P3 lymphocytes(Fig. 3C). Taken together with the physical distribution of lym-phocytes in the parathyroid parenchyma observed by immunohis-tochemistry (IHC), this result confirms the identity of parathyroidadenoma P3 cells as tumor-infiltrating T cells as opposed to con-taminating peripheral blood lymphocytes.We then sought to functionally characterize isolated oxyphil

and chief cells by evaluating basal PTH production and cellularresponse (PTH suppression and intracellular calcium flux) toexogenous calcium. Comparison of PTH production under nor-mocalcemic conditions (1.25 mM) in 11 patient-matched pairs ofoxyphil and chief cells showed significantly higher baseline PTHproduction in the oxyphil cells relative to the chief cells (P =0.026 by paired t test) (Fig. 4A). The paired samples showedsignificant correlation to each other (r = 0.9017), indicating thatadenoma PTH output is strongly influenced by patient origin aswell as cell type (chief vs. oxyphil). In vitro PTH production byisolated cells correlated well with unsorted cells.We next compared PTH production over a 2-h period by

preparatively isolated oxyphil and chief cells under varied calciumconcentrations. Both oxyphil and chief cells secrete PTH whencultured in 0.5 mM, 1.25 mM, and 3.0 mM calcium-containingmedia. Regardless of calcium concentration, oxyphil cells secretedmore PTH per cell than chief cells over the assay interval. Atelevated calcium concentrations, PTH production was inhibitedincompletely in both oxyphil and chief cells. Normalized PTHsecretion profiles of oxyphil and chief cells reveal that bothpopulations show similar PTH suppression at a 3-mM ambientcalcium concentration (Fig. 4B). However, the wide SD clearlyreflects heterogeneous responsiveness among the adenomas. Thesuppressability of PTH secretion by calcium in vitro did not

P3

P4 P5

A

SSC-

A

FSC-A0 200 400 600 800 1000

100

101

102

103

104

P3

P4P5

Oxyphil cell

Chief cell

Lymphocyte

0 20 40 60 80 100Fraction of sorted viable cells (%)

B

Fig. 1. Cellular heterogeneity in parathyroid adenomas. (A, Upper Left) Con-tour diagram showing the distribution of parathyroid cells (P3, P4, and P5) fromdispersed parathyroid adenoma samples based upon the forward and sidescatter channel outputs. Nonviable cells and aggregated cells were excluded bygating for the absence of propidium iodide uptake. In panels labeled P3, P4, andP5 is shown electron microscopy of isolated cells (P3, P4, and P5) from dispersedparathyroid adenomas. (Scale bar: 500 nm.) (B) Box and whisker plots of lym-phocytes, chief cells, and oxyphil cells isolated from a total of 20 cases.

Shi et al. PNAS | February 25, 2014 | vol. 111 | no. 8 | 3093

MED

ICALSC

IENCE

S

correlate with severity of clinical disease (i.e., calcium and PTHlevels in the blood).We also used flow cytometry-based kinetic analysis to evaluate

the relative intracellular responsiveness of the parathyroid cellsto changes in extracellular calcium concentration. Stimulationwith 3 mM extracellular calcium elicited a rapid and robust in-tracellular calcium-flux response in the dispersed parathyroidcells as detected by increased Fluo-4 AM mean fluorescent in-tensity. Clear differences in calcium-flux responses were revealedwhen we compared the relative activity of the oxyphil cells, chiefcells, and lymphocytes (Fig. 4C). Oxyphil cells are the most re-sponsive and account for the great majority of the overall calcium-flux activity in the dispersed parathyroid-cell parental population.Chief cells display a much lower response whereas lymphocytescells are completely unresponsive. These data indicate that readilyidentifiable subpopulations of cells within a given parathyroidadenoma display categorically distinct quantitative responses toextracellular-calcium stimulation.To better understand the heterogeneity of parathyroid ade-

nomas and to explore the question of why parathyroid adenomasdo not sense calcium properly and fail to regulate PTH secretion,we examined CASR expression in parathyroid adenomas froma series of patients with PHPT. First, we used immunofluores-cence detection to evaluate CASR protein expression in isolatedoxyphil and chief cells from a panel of primary parathyroid ad-enoma specimens. As shown in Fig. 5A, CASR immunoreactivityis detected on the surface of both oxyphil and chief cells. Forquantitation of the relative abundance of CASR-positive cellswithin each population, at least 100 cells per field were countedin each of three different fields, and the proportion of CASR-positive cells was calculated as a percentage. Both cell pop-ulations were found to contain similar proportions of CASR-positive cells (39.5 ± 6.3% for chief cells and 41.0 ± 6.2% foroxyphil cells) (Fig. 5B). Second, we examined CASR transcriptabundance in chief and oxyphil cells by quantitative reverse-

transcriptase polymerase chain reaction (qRT-PCR). Both celltypes were found to contain equivalent amounts of the CASRmRNA (Fig. 5C). Because oxyphil and chief cells differ greatly intheir calcium responsiveness as evaluated in the flux assay, thepresence of comparable CASR expression in both populationsindicates that the degree of calcium responsiveness in para-thyroid cells is not solely determined by CASR abundance.The functional and morphological heterogeneity we observed

within primary parathyroid adenomas and their cell isolates raisedthe question of whether the originating tumors were uniformlyclonal as is widely believed (8, 9). To test this concept, we probedfor allele-specific X-inactivation in parathyroid adenoma cells de-rived from female patients at two highly polymorphic imprinted loci,the human androgen receptor (HUMARA) (20) and the phosphateglycerate kinase PGK genes (21). Five female HUMARA locusinformative cases were subsequently assessed for clonality. Two ofthe five cases were found to be monoclonal using both wholeadenoma tissue and laser-capture microdissected frozen-sectionspecimens as input (Fig. 6A). In contrast, the remaining three casesshowed a polyclonal pattern both in whole adenoma tissue and

Fig. 2. Relative abundance and distribution of lymphocytes, chief cells, andoxyphil cells from parathyroid adenomas and matched normal parathyroidgland biopsies from patients with PHPT. (A) Representative contour plots ofside scatter and forward scatter parameters from flow-cytometry analysisof parathyroid adenoma and matched normal tissue. (B) Box and whisker plotsof lymphocytes, chief cells, and oxyphil cells isolated from a total of five cases.

Fig. 3. Immunophenotypic characterization of cells isolated from para-thyroid adenomas. (A) Flow-cytometric analysis of cell-surface markers CD45,CD3, CD19, CD24, and CD44 on tumor-infiltrating lymphocytes (TILs) (Lower)and patient-matched PBLs (Upper). (B) Tissue sections from three indepen-dent parathyroid adenomas were probed with an anti-CD3 antibody, andreactivity was visualized by DAB staining and with hematoxylin/eosin coun-terstaining. (C) CD4 and CD8 cell-surface expression in P3 cells derived fromthree parathyroid adenomas was determined by FACS. The ratio of CD4+ toCD8+ cells is based upon quadrant gating using standard conditions.

3094 | www.pnas.org/cgi/doi/10.1073/pnas.1319742111 Shi et al.

microdissected frozen-section specimens (Fig. 6A). Patient-matchedperipheral blood lymphocyte genomic DNAwas used in as an alleliccontrol to demonstrate the polyclonal status of the HUMARA lo-cus in all five of the cases tested. We next probed for the clonalstatus of isolated oxyphil and chief cells and found that these cellsreflected the same clonality pattern observed in the originatingpatient-matched whole adenoma tissue (Fig. 6A). In the mono-clonal cases, the oxyphil and chief cells both harbored the samematernally imprinted HUMARA allele, indicating that thesepopulations arose from a common progenitor. Conversely, inthe polyclonal specimens, both the oxyphil and chief populationswere found to contain equal representations of each imprintedHUMARA allele, indicating the nonclonal provenance of bothcell types. To confirm the HUMARA findings, we also performedmethylation-sensitive PCR of the imprinted PGK gene (21). Of 8HUMARA locus informative cases, 4 were monoclonal and 4were polyclonal; of 9 PGK locus informative cases, 7 cases weremonoclonal and 2 were polyclonal (Fig. 6B). Three cases wereinformative at both HUMARA and PGK loci, and all 3 yieldedconcordant results (2/3 monoclonal, 1/3 polyclonal). Altogether,14 informative cases assayed by either method or both methodsshowed polyclonal origin in 5 and monoclonal origin in 9 para-thyroid adenomas. A single tumor from a patient with renalfailure-induced secondary HPT was investigated as control andshowed polyclonal status in both HUMARA and PGK assays.

DiscussionParathyroid neoplasias result in abnormal PTH secretion froman expanded population of parathyroid cells. With few excep-tions, a significant body of prior work has shown that parathyroidtumors in PHPT are clonal and that aggregate dispersed parathyroidtumor cells secrete PTH in culture but can be poorly respon-sive to negative feedback by calcium. In our work to iden-tify differential gene expression and mechanisms of abnormalcalcium sensing in parathyroid adenomas, we noticed substantialheterogeneity in these tumors. This finding, coupled with theobserved histologic heterogeneity of parathyroid adenomas, ledus to design the current study to isolate and functionally char-acterize individual populations of cells from parathyroid ade-nomas. We report that parathyroid adenomas comprise variable

proportions of chief cells, oxyphil cells, and infiltrating T lym-phocytes and that the two parathyroid cell types manifest distinctfunctional properties. Parathyroid chief cells and oxyphil cells ex-press similar amounts of CASR, and both exhibit attenuated PTHsecretory responses to changes in ambient calcium. However, therespective cell types are markedly different in their baseline PTHproduction and intracellular flux response to extracellular calciumchallenge. As the basis for this differential calcium responsivenessis not attributable solely to CASR relative abundance, addi-tional mechanisms, including enhanced biochemical oppositionto CASR signaling, altered CASR protein trafficking, or CASR-independent calcium sensing, may prove to be important functionaldeterminants in these cells. Further, we show that a significantproportion of parathyroid adenomas causing PHPT are polyclonalrather than monoclonal. These results indicate that parathyroidtumors causing primary hyperparathyroidism are heterogeneousat the cellular and functional level and further show that somepatients’ tumors are monoclonal whereas other patients havepolyclonal tumors. Expanded studies of a large number of patientswith PHPT to assess frequency of these different tumor types andthe clinical phenotypes associated with them will be needed tofully understand the meaning of these findings. It will be importantin future studies to relate the variable biochemical behavior andclonal status of PHPT tumors to the presence of driver mutationsin known parathyroid tumor suppressor genes and oncogenes suchas MEN1 and CCND1 (22).Earlier investigations using histologic methods have revealed

that parathyroid adenomas comprise mainly chief cells, transi-tional oxyphil cells, and oxyphil cells. These observational studieswere limited to analysis of fixed parathyroid tissue sections. Untilthe current work, the successful preparative isolation of livingoxyphil and chief cells from human parathyroid adenoma tissuehas not been reported. Using a flow cytometry-based approach,we analyzed 20 adenomas resected from patients with PHPT anddetermined the proportional representation of chief cells, oxy-phil cells, and infiltrating T lymphocytes in the primary tumor.Wide variations among adenomas were observed. In contrast,the distribution of these cell types proved remarkably consistent

Fig. 4. Parathyroid hormone (PTH) secretion and intracellular calcium-fluxresponse to 3.0 mM extracellular calcium in dispersed primary cells isolatedfrom parathyroid adenomas. (A) Basal PTH secretion over 2 h by tumor-matched chief and oxyphil cells from 11 PHPT patients under normocalcemic(1.25 mM) conditions. (B) PTH secretion over 2 h under varied calcium con-centration by tumor-matched chief and oxyphil cells (n = 6). (C) Intracellularflux response among isolated lymphocytes, chief cells, and oxyphil cells. Onerepresentative experiment out of seven is shown.

Fig. 5. Expression of CASR in isolated parathyroid cells. (A) Immunofluo-rescence (IF) for CASR protein in chief and oxyphil cells using anti-CASRmonoclonal antibody (20x). (Inset) CASR expression-magnified (63x, oil). (B)Quantitation of CASR IF-positive cells in chief and oxyphil cells from A. Atotal of more than 100 cells per field were counted in each of three differentfields. Data are from three patients. (C) Pairwise comparison of CASR mRNAexpression measured by qRT-PCR in chief and oxyphil cells from parathyroidadenomas from seven patients, normalized to an internal GAPDH control.

Shi et al. PNAS | February 25, 2014 | vol. 111 | no. 8 | 3095

MED

ICALSC

IENCE

S

in a series of five histologically normal parathyroid glands ipsilateralto adenomatous tissue. This result suggests that the variable com-position of parathyroid adenomas relative to normal tissue couldreveal a greater degree of functional heterogeneity among primaryparathyroid adenomas than is currently appreciated.To address this question, we isolated live oxyphil and chief

cells from parathyroid adenomas and compared the behavior ofthe respective cell types in functional assays measuring calciumresponsiveness and PTH secretion, two definitive performancemetrics of parathyroid gland function. Our data indicate that thesubpopulations of chief and oxyphil cells within a given para-thyroid adenoma display clearly different quantitative responsesto extracellular calcium stimulation. In all cases tested, oxyphilcells responded much more strongly than chief cells to increasedambient calcium concentration. Baseline PTH production undernormocalcemic conditions (1.25 mM) in individual patient-matched pairs of chief and oxyphil cells was significantly higherin oxyphil cells relative to the chief cells. The enhanced calciumsensitivity and increased PTH production we observed in oxyphilcells relative to chief cells is consistent with a recent report thatused immunohistochemistry to compare oxyphil and chief cellgene expression in hyperplastic parathyroid tissue from patientswith hyperparathyroidism secondary to chronic kidney disease(23). In their study, Ritter et al. (23) demonstrated elevatedexpression of parathyroid-relevant genes including PTH andPTHrP in oxyphil cells relative to chief cells. In conjunction withthis finding, our data support the notion that oxyphil cells areimportant functional participants in the calcium-sensing and

secretory properties of parathyroid tissue. Although the numbersof patients studied are too small to form clear conclusions, thebehavior of preparatively isolated chief and oxyphil cells appearssimilar to the activity of each of these cell types when analyzed asunsorted cell suspensions. Further, in vitro basal and calciumstimulated PTH production by isolated cells did not correlatewith severity of patient disease, underscoring the complexphysiology of calcium homeostasis in patients.CASR down-regulation is generally believed to be the principal

mechanism for the abnormal calcium-PTH set-point relationshipin patients with parathyroid tumors. If this assumption is true,then CASR expression should be significantly reduced in allparathyroid tumors, and calcium responsiveness should be de-monstrably dependent upon CASR expression. The data in ourstudy challenge this prevailing model. We show that chief andoxyphil cells display comparable CASR expression at both theprotein and transcript levels, yet oxyphil cells are clearly moreresponsive to calcium and exhibit consistently higher basal PTHsecretion. These data indicate that CASR expression is not thesole determinant of PTH production and calcium responsivenessin the parathyroid gland and suggest that the elevated PTH as-sociated with PHPT could be reflective of not only increased tu-mor-cell number but also the higher basal PTH secretion profileof oxyphil cells compared with chief cells.Our study indicates that the presence of lymphocytic in-

filtration is a common feature in parathyroid adenomas. Usingflow cytometry, we identified the parathyroid adenoma-derivedP3 population as T lymphocytes expressing the leukocyte com-mon antigen CD45 and T-cell receptor unit CD3. Additionalmarker studies revealed the T lymphocytes to be CD45+/CD3+/CD19−/CD24−/CD44+. The role of immune cells in the patho-genesis of human parathyroid adenoma is less clear. Immunecells can release inflammatory mediators with proangiogenic andprometastatic effects. Tumor cells can express antigens and be-come targets for a T cell-mediated adaptive immune response.The presence of high numbers of tumor-infiltrating lymphocytes,particularly T cells, has been found to be a major predictor offavorable clinical outcomes in several solid cancers (24–26).These findings support the hypothesis that the adaptive immuneresponse influences the behavior of human tumors. However,lymphocytic infiltration in parathyroid adenoma is rarely repor-ted (27), which suggests that our current findings may representa newly recognized histologic feature. Because we found lym-phocytes in both normal and adenomatous parathyroid glands,the role of these T cells is still unclear. As the ratio of effector(CD8+) to helper (CD4+) T cells in P3 lymphocytes recoveredfrom parathyroid adenoma tissue is consistently distinct fromthe 1:2 ratio maintained in the bloodstream, it is clear that theP3 population represents tumor infiltrating lymphocytes (TILs)rather than adventitious contaminating cells carried over fromvascular elements. It is possible that the TILs may be homing toantibodies targeting CASR, PTH, or other parathyroid antigensand that antibody-mediated inhibition of CASR or clearance ofCASR-expressing or PTH-producing cells could induce compen-satory hyperplasia in the gland leading to adenoma development.Alternatively, the P3 TILs could be responding to inflamma-tory mediators, tumor-specific antigens, or senescence-associatedproteins released upon oncogene-induced senescence and maybe participating in tumor-clearance activities. Further study willbe necessary to verify the functions of these immune cells and todiscover any potential relationships to parathyroid disease.Our results are at odds with the assumption of clonal expan-

sion as the sole tumorigenic mechanism in parathyroid adeno-mas. The detection of 5 polyclonal adenomas out of 14 testedcertainly suggests that monoclonal expansion and multi- or poly-clonal proliferation may coexist as alternative mechanisms in theorigin of primary parathyroid adenomas. Seminal work (8, 9)established that monoclonal proliferation is commonly found in

Fig. 6. Clonality of parathyroid adenomas. (A) Clonality of parathyroid cellpopulations isolated from five parathyroid adenomas. HUMARA alleles wereamplified usingmethylation-sensitive PCR in HhaI undigested (−) and digested (+)genomic DNA from peripheral blood lymphocytes (PBL), whole tumor tissue(Adenoma), preparatively sorted chief and oxyphil cells, or microdissected tumortissue (LCM). The presence of a single band reveals a single methylated allelefrom a monoclonal tumor (Pt3 and -14) whereas two bands reflects randommethylation of alleles in a polyclonal tumor (Pt15, -16, and -17). (B) Clonality of 14parathyroid adenomas identified by HUMARA and PGK methods.

3096 | www.pnas.org/cgi/doi/10.1073/pnas.1319742111 Shi et al.

primary hyperparathyroidism, yet subsequent studies using wholetissue (10, 11) have shown that polyclonal origin can occur in asignificant subset of primary parathyroid adenomas. Thesestudies were limited by the potential for contamination in samplesnot specifically enriched for individual cell populations. In ourstudy, we used multiple independent methods of tissue samplingand two independent methods of clonality determination to es-tablish definitively that both polyclonal and monoclonal mecha-nisms can give rise to primary parathyroid adenomas. Differencesin reported frequencies of polyclonal versus monoclonal originmay be attributable to the lack of definitive pathological criteriafor distinguishing adenoma from hyperplasia, sample size limi-tations in the number of parathyroid glands tested, as well as theapplication of different clonality assays with variations in meth-odology and the ability to exclude contaminating nonadenomaelements (11). Recently, a new model has been described in whichtumor heterogeneity arises through recruitment of proximateuntransformed but hyperplastic cells into aggregation chimerasinduced by a monoclonal population of tumor-initiating cells (28).It is possible that a similar mechanism may contribute to theorigin of the polyclonal adenomas observed in our study. Futurestudies will be required to analyze the genomic compositionof parathyroid adenomas at single-cell resolution to determinewhether polyclonal tumors are constitutively transformed or whethera monoclonal initiating component has induced hyperplastic expan-sion of neighboring untransformed parathyroid cells.The current study demonstrates that flow cytometry can be

used to identify, analyze, and recover functionally distinct sub-populations of live cells from human parathyroidectomy speci-mens. Using this approach, we show that parathyroid oxyphil andchief cells derived from parathyroid adenomas display cleardifferences in calcium responsiveness and PTH secretion. Thedifferential activity and relative abundance of these distinct celltypes could contribute to heterogeneity in clinical presentation,therapeutic response, and outcome in PHPT. Importantly, the

demonstration of polyclonality within primary adenomas raisesthe possibility that such expansions may occur during an un-recognized prodrome phase analogous to the abnormal physio-logical contexts that drive the development of secondaryparathyroid hyperplasia. Further work will be necessary to testthese concepts in expanded patient cohorts to define the re-lationship of these findings to the clinical presentation anddisease course of PHPT and to improve our understanding ofthe causes of primary and secondary parathyroid neoplasia.

Materials and MethodsPatients and Samples. Informed consent was sought from patients scheduledto undergo parathyroidectomy for PHPT. Consenting patients donatedresected parathyroid tissue for study (Protocol Number 00007056, approvedby the Institutional Review Board of the Duke University Medical Center). Allof the patients were diagnosed with sporadic, nonfamilial PHPT on the basisof standard clinical and biochemical parameters, and all had a single para-thyroid tumor removed at curative surgery. Viable dispersed parathyroidcells collected from the surgical specimens were preparatively sorted byflow cytometry prior to further analysis. Cellular morphology was eval-uated by electron and light microscopy. Marker gene expression wasdetected by immunofluorescence, immunohistochemistry, quantitativereverse-transcriptase PCR, and analytical flow cytometry. Calcium re-sponsiveness was detected via kinetic analysis of Fluo-4 AM fluorescenceintensity. Clonality status was determined by PCR assays detecting X-chromosomeinactivation at the highly polymorphic HUMARA and PGK loci. A full de-scription of the experimental procedures may be found in the SI Materialsand Methods.

ACKNOWLEDGMENTS. We thank Dr. Sara Miller and Mr. Phillip Christopher(Duke Electron Microscopy Core Facility) for excellent technical assistance inthe preparation and analysis of the EM data, Dr. Michael Cook and Ms. LynnMartinek (Duke Flow Cytometry Shared Resource, Duke Cancer Institute) foradvice and assistance, Ms. Cynthia Webb for assistance with parathyroidtissue procurement and banking, and Dr. Yasheng Gao (Duke Light Micros-copy Core Facility) for assistance with the live-cell imaging studies. This workwas supported by National Institutes of Health Grant R01 DK088188-03 (toJ.A.O. and J.K.).

1. Brown EM (1991) Extracellular Ca2+ sensing, regulation of parathyroid cell function, androle of Ca2+ and other ions as extracellular (first) messengers. Physiol Rev 71(2):371–411.

2. DeLellis RA, Mazzaglia P, Mangray S (2008) Primary hyperparathyroidism: A currentperspective. Arch Pathol Lab Med 132(8):1251–1262.

3. Arnold A, et al. (1992) PRAD1 (cyclin D1): A parathyroid neoplasia gene on 11q13.Henry Ford Hosp Med J 40(3-4):177–180.

4. Carling T, et al. (1998) Parathyroid MEN1 gene mutations in relation to clinical character-istics of nonfamilial primary hyperparathyroidism. J Clin Endocrinol Metab 83(8):2960–2963.

5. Kishikawa S, et al. (1999) Overexpression and genetic abnormality of p53 in para-thyroid adenomas. Pathol Int 49(10):853–857.

6. Erickson LA, et al. (1999) Parathyroid hyperplasia, adenomas, and carcinomas: Dif-ferential expression of p27Kip1 protein. Am J Surg Pathol 23(3):288–295.

7. Cromer MK, et al. (2012) Identification of somatic mutations in parathyroid tumorsusing whole-exome sequencing. J Clin Endocrinol Metab 97(9):E1774–E1781.

8. ArnoldA,StauntonCE,KimHG,GazRD,KronenbergHM(1988)Monoclonalityandabnormalparathyroid hormone genes in parathyroid adenomas. N Engl J Med 318(11):658–662.

9. Noguchi S, Motomura K, Inaji H, Imaoka S, Koyama H (1994) Clonal analysis of parathyroidadenomas by means of the polymerase chain reaction. Cancer Lett 78(1-3):93–97.

10. Fialkow PJ, Jackson CE, Block MA, Greenawald KA (1977) Multicellular origin ofparathyroid “adenomas” N Engl J Med 297(13):696–698.

11. Sanjuan X, Bryant BR, Sobel ME, Merino MI (1998) Clonality analysis of benign parathyroidlesions by human androgen receptor (HUMARA) gene assay. Endocr Pathol 9(1):293–300.

12. Mayer GP, Hurst JG (1978) Sigmoidal relationship between parathyroid hormone secretionrate and plasma calcium concentration in calves. Endocrinology 102(4):1036–1042.

13. Brown EM (1983) Four-parameter model of the sigmoidal relationship betweenparathyroid hormone release and extracellular calcium concentration in normal andabnormal parathyroid tissue. J Clin Endocrinol Metab 56(3):572–581.

14. Kifor O, et al. (1996) Reduced immunostaining for the extracellular Ca2+-sensingreceptor in primary and uremic secondary hyperparathyroidism. J Clin EndocrinolMetab 81(4):1598–1606.

15. Farnebo F, Höög A, Sandelin K, Larsson C, Farnebo LO (1998) Decreased expressionof calcium-sensing receptor messenger ribonucleic acids in parathyroid adenomas.Surgery 124(6):1094–1098, discussion 1098–1099.

16. Pi M, Spurney RF, Tu Q, Hinson T, Quarles LD (2002) Calcium-sensing receptor acti-vation of rho involves filamin and rho-guanine nucleotide exchange factor. Endo-crinology 143(10):3830–3838.

17. Koh J, et al. (2011) Regulator of G protein signaling 5 is highly expressed in para-thyroid tumors and inhibits signaling by the calcium-sensing receptor.Mol Endocrinol25(5):867–876.

18. Cinti S, Sbarbati A (1995) Ultrastructure of human parathyroid cells in health anddisease. Microsc Res Tech 32(2):164–179.

19. McGregor DH, Lotuaco LG, Rao MS, Chu LL (1978) Functioning oxyphil adenoma ofparathyroid gland: An ultrastructural and biochemical study. Am J Pathol 92(3):691–711.

20. Allen RC, Zoghbi HY, Moseley AB, Rosenblatt HM, Belmont JW (1992) Methylationof HpaII and HhaI sites near the polymorphic CAG repeat in the human androgen-receptor gene correlates with X chromosome inactivation. Am J Hum Genet 51(6):1229–1239.

21. Gilliland DG, Blanchard KL, Levy J, Perrin S, Bunn HF (1991) Clonality in myelopro-liferative disorders: Analysis by means of the polymerase chain reaction. Proc NatlAcad Sci USA 88(15):6848–6852.

22. Costa-Guda J, Arnold A (2013) Genetic and epigenetic changes in sporadic endocrinetumors: Parathyroid tumors. Mol Cell Endocrinol, 10.1016/j.mce.2013.09.005.

23. Ritter CS, Haughey BH, Miller B, Brown AJ (2012) Differential gene expression byoxyphil and chief cells of human parathyroid glands. J Clin Endocrinol Metab 97(8):E1499–E1505.

24. Galon J, et al. (2006) Type, density, and location of immune cells within human co-lorectal tumors predict clinical outcome. Science 313(5795):1960–1964.

25. Dieu-Nosjean MC, et al. (2008) Long-term survival for patients with non-small-celllung cancer with intratumoral lymphoid structures. J Clin Oncol 26(27):4410–4417.

26. Zhang L, et al. (2003) Intratumoral T cells, recurrence, and survival in epithelialovarian cancer. N Engl J Med 348(3):203–213.

27. Lam KY, Chan AC, Lo CY (2000) Parathyroid adenomas with pronounced lymphocyticinfiltration: No evidence of autoimmune pathogenesis. Endocr Pathol 11(1):77–83.

28. Thliveris AT, et al. (2013) Transformation of epithelial cells through recruitment leadsto polyclonal intestinal tumors. Proc Natl Acad Sci USA 110(28):11523–11528.

Shi et al. PNAS | February 25, 2014 | vol. 111 | no. 8 | 3097

MED

ICALSC

IENCE

S