Human breast cancer metastases to the brain …Brain Breast Metastasis * Fig. 1. A 47-y-old female with breast cancer metastasis to the brain. (Lower) Preoperative T1 gadolinium-enhanced

Human breast cancer metastases to the brain displayGABAergic properties in the neural nicheJosh Nemana, John Terminib, Sharon Wilczynskic, Nagarajan Vaidehid, Cecilia Choya,e, Claudia M. Kowolikc, Hubert Lid,e,Amanda C. Hambrechta,f, Eugene Robertsg,1, and Rahul Jandiala,f,1

Divisions of aNeurosurgery and cPathology, Departments of bMolecular Medicine, dImmunology, and gNeurobiochemistry, and eIrell and Manella GraduateSchool of Biological Sciences, City of Hope, Duarte, CA 91010; and fDepartment of Biology, University of Southern California, Los Angeles, CA 90089

Contributed by Eugene Roberts, November 27, 2013 (sent for review October 16, 2013)

Dispersion of tumors throughout the body is a neoplastic processresponsible for the vast majority of deaths from cancer. Despitedisseminating to distant organs as malignant scouts, most tumorcells fail to remain viable after their arrival. The physiologic mi-croenvironment of the brain must become a tumor-favorable mi-croenvironment for successful metastatic colonization by circulatingbreast cancer cells. Bidirectional interplay of breast cancer cells andnative brain cells in metastasis is poorly understood and rarelystudied. We had the rare opportunity to investigate uncommonlyavailable specimens of matched fresh breast-to-brain metastasestissue and derived cells from patients undergoing neurosurgical re-section. We hypothesized that, to metastasize, breast cancers mayescape their normative genetic constraints by accommodating andcoinhabiting the neural niche. This acquisition or expression of brain-like properties by breast cancer cells could be a malignant adaptationrequired for brain colonization. Indeed, we found breast-to-brainmetastatic tissue and cells displayed a GABAergic phenotype similar tothat of neuronal cells. The GABAA receptor, GABA transporter, GABAtransaminase, parvalbumin, and reelin were all highly expressed inbreast cancer metastases to the brain. Proliferative advantage wasconferred by the ability of breast-to-brain metastases to take up andcatabolize GABA into succinate with the resultant formation of NADHas a biosynthetic source through the GABA shunt. The results suggestthat breast cancers exhibit neural characteristics when occupying thebrain microenvironment and co-opt GABA as an oncometabolite.

brain metastasis | tumor microenvironment

Metastases are responsible for 90% of all cancer deaths, andpatients diagnosed with brain metastases have a dismal

20% probability of 1-y survival (1–3). The brain is increasinglythe first site of recurrence after treatment of stage IV advancedbreast cancer, even when disease in other sites is in remission.This emerging clinical problem significantly limits the survivalgains made from recent advances in systemic therapy for breastcancer (4). Breast cancer metastasizes to the brain in ∼40% ofpatients who have a tumor that is HER2+ (>30% of tumor cellshave complete membrane staining for the tyrosine kinase re-ceptor erbB2) or triple negative (TN) (negative for the estrogenand progesterone receptors and have reduced expression ofHER2+) (5). Ninety percent of patients with these breast cancersubtypes will die of metastasis to the brain (1). Currently,treatment options beyond radiotherapy and neurological surgeryare limited, underscoring the need for research into the biologyof these clinically recalcitrant tumors (6).Breast cancer patients typically develop brain metastases

months to several years after their initial diagnosis (6). Thisunique clinical latency occurs despite the early presence of cir-culating tumor cells, often detectable at the time of primary di-agnosis (7–9). These observations suggest that the final step ofthe metastatic cascade to colonize the brain may be the rate-limiting biological step that determines whether or not clinicalprogression will occur. Primary cancers, especially those ofbreast origin, have numerous changes in transcriptome net-works that can lead to clones with additional survival gains at

secondary sites (10–14). We previously showed that metastaticcells have the ability to alter the cellular milieu of the brainfor growth advantage (3).γ-Aminobutyric acid (GABA) was first identified in the

mammalian brain over one-half a century ago and subsequentstudies have demonstrated its relevance to various medical andscientific paradigms (15–18). In addition to its role in neuro-transmission, GABA can act as a trophic factor during nervoussystem development to influence cellular events including pro-liferation, migration, differentiation, synapse maturation, andcell death (19, 20). Various cells can catabolize GABA to meettheir metabolic needs. Under conditions that inhibit the tri-carboxylic acid cycle, impair respiration, and enhance the accu-mulation of reactive oxygen intermediates, GABA can be used asan energy source for growth through a pathway known as theGABA shunt (21). The expression of GABA in adult tissues isspecific and tightly regulated. Abnormal GABA expression orGABAergic participation has been described in primary colon,gastric, ovarian, pancreatic, and breast cancers (22).We obtained surgical specimens of HER2+ and TN subtypes

from patients undergoing neurological surgery. This offereda rare opportunity to examine the histological, cellular, andmolecular features of the tumor microenvironment that cannotbe recreated using highly passaged cell lines and derived xeno-grafts. Results showed that breast-to-brain metastatic (BBM)cells overexpressed integral proteins from the GABA-relatedvariables, including GABA transaminase (ABAT), GABAAreceptor (GABAAR), glutamate decarboxylase (GAD67), GABAtransporter, reelin, and parvalbumin. Brain metastases from bothtypes of breast cancer (HER2+, TN) metabolized GABA in vitro

Significance

Breast cancer patients typically develop brain metastases yearsafter their initial diagnosis. During this clinical latency, cancercells must evolve and adapt to the neural microenvironment tocolonize. We hypothesized that breast cancer cells may assumebrain-like properties to survive in the brain. Our results suggestthat metastases overexpress many variables related to theγ-aminobutyric acid (GABA) and were able to proliferate bymetabolizing GABA as a biosynthetic energy source. The ex-pression of brain-like properties by breast cancer cells could bea malignant adaptation required for metastasis to the brain,which could potentially be exploited to develop new therapy forbreast cancer patients.

Author contributions: J.N., J.T., E.R., and R.J. designed research; J.N., C.C., C.M.K., H.L.,and A.C.H. performed research; J.N., N.V., and R.J. contributed new reagents/analytictools; J.N., J.T., S.W., N.V., H.L., E.R., and R.J. analyzed data; and J.N., J.T., E.R., and R.J.wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. E-mail: [email protected] or [email protected].

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1322098111/-/DCSupplemental.

984–989 | PNAS | January 21, 2014 | vol. 111 | no. 3 www.pnas.org/cgi/doi/10.1073/pnas.1322098111

Dow

nloa

ded

by g

uest

on

Oct

ober

15,

202

0

and used it as biosynthetic source to enhance cell growth. Theexpression of GABAergic features by BBM cells demonstratesthat, despite their origins in distinct and disparate germ-line tis-sues, cancers have the ability to adapt to and colonize newmicroenvironments.

ResultsBBM Cells Exhibit GABA-Related Variables. The tumor and brainmicroenvironment is a highly specialized niche determined by itstissue-specific location and cell-derived contents. Unlike tumorsoriginating within the brain (glioblastoma multiforme), meta-static breast tumors are noninfiltrating and retain their originaltissue morphology. The demarcation of the brain–tumor in-terface can be clearly discerned histologically or clinically onmagnetic resonance imaging (Fig. 1). In patients undergoingneurosurgical resection of BBMs, specimens from the tumor andbrain interface demonstrated few healthy neurons. Metastaticbreast cancer cells were juxtaposed to astrocytes expressing glialfibrillary acidic protein (GFAP) both in the peritumoral brainand tumor regions (Fig. 1). Furthermore, these reactive astro-cytes had a hypertrophic morphology comparable to the glialscars that form in response to traumatic brain injury (23).Resected BBM specimens were classified by their clinical

subtypes (HER2+ and TN) and Bloom–Richardson pathologicalgrade (24). Biopsies of brain metastases revealed ductal carci-nomas with high-grade Bloom–Richardson score (BRG3) (Fig.S1). Low-passage cells derived from metastatic brain tumorresections maintained their original tissue morphology whengrown in serum-free “brain-like” conditions and 3D Matrigelcultures (Fig. S1) (25). Within the intratumoral space in meta-static brain tissue, cells of glial origin interdigitated with thebreast tumor cells (26). Investigating the tumor cell and stromalcompartments separately revealed differences in the micro-environments (12). Staining with H&E combined with stainingfor GFAP (expressed in the intratumoral glial cells) and HER2+

(expressed in tumor cells) allowed us to definitively observe thebrain–tumor boundary (Figs. S1 and S2). There was a largedifference in volume between tumor nuclei (523 ± 43 μm3) andnontumor nuclei (58 ± 8 μm3). Cultured BBM cells organized intomicrocolonies with cells retaining multiple nucleoli (Fig. S1).Because of the abundant expression of GABA receptors and

GABA in the brain microenvironment (27), we investigatedwhether BBM cells had any GABAergic properties by comparingthe expression of their genes and proteins to the highly passagedprimary breast cancer cell lines, SkBr3 and MDA-MB-231. AllGABA receptor mRNA isoforms were highly up-regulated inHER2+ BBMs relative to HER2+ primary breast cancer cell lineSkBr3 (Fig. 2A). The expressions of 12 of 15 GABA receptormRNA isoforms were up-regulated in TN BBM cells relative toMDA-MB-231 (Fig. 2A and Table S1). GABAAR protein wasexpressed by both HER2+ and TN breast cancer subtypes in vitro(Fig. S3A).The tumor compartments of both BBM HER2+ and TN

subtypes had significantly increased levels of GABAAR relativeto the primary tissue tumor correlates (Fig. 2B, Fig. S3B, andTable S2). The fold increase was 1.46 in HER2+ cancer cells and3.12 in TN cancer cells, relative to matched clinical subtypeprimary breast tissue (Fig. S3C). A survey of the nontumor mi-croenvironment also revealed GABAAR expression was in-creased in the brain region of TN BBMs relative to TN primarybreast tumor stromal tissue (Fig. S3B). Expression of GABAAResubunit mRNA was decreased 20-fold in TN BBM cells relativeto TN primary breast cancers (Fig. 2A), and the expression ofthe protein was low in both primary and metastatic tissues (Fig.S4). Variance between mRNA and protein level expression inBBMs could be due to differential micro-RNA–mediated si-lencing, proteasome degradation, or RNA turnover resulting inaugmented translation (28, 29).

The effect of exogenous GABA on the growth of BBM cells inculture was investigated next. GABA induced dose-dependentincreases in cellular proliferation rates; the average EC50 was 0.8μM for HER2+ BBM cells (Fig. 2C) and 0.4 μΜ for TN BBMcells (Fig. S3D). This increase in proliferation could be mediatedeither by GABAAR signaling or by increased GABA metabo-lism. To distinguish between these two possibilities, we tested theproliferative capacity of cells with muscimol, a potent GABAmimetic selective for GABAAR that cannot be used as a meta-bolic resource (30, 31). When BBM cells were cultured withmuscimol, there was no change in proliferation relative to non-treated cells (Fig. 2D and Fig. S3E), indicating that the observedgrowth rate of the cells treated with exogenous GABA was notmediated through GABAAR signaling.

GABA Metabolism Promotes Tumor Cell Proliferation. BBM cellsmay express highly specific GABA transporters that take upGABA from the brain microenvironment. HER2+ BBM cellshad elevated mRNA levels for vesicular GABA transporter(VGAT), GABA transporter 1 through 3 (GAT1–3), and the

Brain

Breast

Metasta

sis

*

Fig. 1. A 47-y-old female with breast cancer metastasis to the brain. (Lower)Preoperative T1 gadolinium-enhanced magnetic resonance imaging and 3Dreconstruction on a 3-T magnet scanner shows a left parietal lobe braintumor (*). (Upper) After surgical resection, a specimen incorporating thetumor/brain boundary (perforated white line) was stained, and 3D renderingdemonstrates hypertrophic GFAP+ reactive astrocytes (red) and HER2+ breastcancer cells (green). Nuclear DNA is counterstained with DAPI (blue) (40×magnification).

Neman et al. PNAS | January 21, 2014 | vol. 111 | no. 3 | 985

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

15,

202

0

GABA-betaine transporter (BGT) relative to HER2+ SkBr3primary breast cancer cells. TN BBM cells had a twofold orhigher increase in VGAT, GAT3, and BGT relative to TNMDA-MB-231 primary breast cancer cells (Table S2 and Fig.S5A). GAT1 protein was expressed by both BBM subtypes invitro (Fig. S5B). Staining of tissue specimens revealed the ex-pression levels of GAT1 in HER2+ BBM, and adjacent braintissue regions were equivalent to primary tissue. However, theexpression of GABA transporter in TN BBMs and adjacentbrain regions was higher relative to primary tissue regions (Fig.S5C). These results suggest that metastatic breast cells in thebrain have acquired the ability to take up extracellular GABA.Neurons and astrocytes express GABA transaminase (ABAT),

which converts GABA to succinate, resulting in subsequentproduction of NADH to satisfy energy and growth requirements(32). Accordingly, BBM cells may also use ABAT to metabolizeGABA. Datasets from 357 patients show BBMs have increasedABAT expression relative to primary breast cancer (1.9-foldchange, P = 0.020) and normal breast tissue (4.3-fold change, P =4.9 × 10−7, Fig. S6A). Our patient-derived HER2+ and TN BBMcells also had increased mRNA copy numbers for ABAT relativeto primary breast cancer cell lines (Fig. 3A, Fig. S6B, and TableS2). ABAT protein was expressed by both BBM subtypes in vitro(Fig. 3B). Tissue specimens revealed greater than twofoldincrease in ABAT protein expression in the tumor compart-ments of both HER2+ and TN BBMs relative to primarymatched subtype tissues (Fig. 3C and Fig. S6C). There was nosignificant difference of ABAT levels in the adjacent braintissue versus primary stromal compartments for either tumorsubtype (Fig. S6C).To test whether GABA can act as an energy source to enhance

proliferation, BBM cells were incubated with vigabatrin, an

irreversible inhibitor of ABAT. Cells were cultured in mediacontaining increasing concentrations of GABA alone or GABAplus 300 μM vigabatrin for 96 h. Vigabatrin abolished the pro-liferative effect of GABA on BBM cells (Fig. 3D and Fig. S6D).We determined the intracellular levels of NADH in BBM cellscultured (i) alone, (ii) with 1 μM GABA or muscimol, or (iii)with 1 μM GABA or muscimol and 300 μM vigabatrin. BBMcells cultured in the presence of exogenous GABA produced47% more intracellular NADH than control (284 pmol versus193 pmol, respectively), whereas BBM cells cultured with mus-cimol or exogenous GABA plus vigabatrin did not have elevatedNADH levels (Fig. 3E and Fig. S6E). These results suggest thatGABA is a source of energy that BBM cells can use to fuelproliferation via the GABA shunt.Glutamate catabolism can also increase NADH levels in the

GABAergic environment. Glutamate is decarboxylated by the en-zyme GAD67 to produce GABA. GAD67 is specifically expressedin GABAergic neurons and certain peripheral tissues that are alsoGABA dependent (33). BBMs have an increase in GAD67 mRNAand protein levels relative to primary breast tumors cells and tissue(Fig. S7 and Table S2). Therefore, BBM cells could use GAD67 toconvert glutamate to GABA and thus provide an additional met-abolic source to promote tumor cell proliferation.

BBM Cells Express GABAergic Interneuron-Specific Proteins. Parval-bumin, a calcium-binding protein, is an important modulator ofGABAergic neuronal synaptic plasticity (34). Cellular stainingrevealed that HER2+ and TN BBM subtypes express this proteinin vitro (Fig. S8A). In tissue specimens, parvalbumin expressionwas significantly increased in the tumor and nontumor com-partments of both BBM subtypes, relative to their primary tissuetumor correlates (Fig. S8B). Collectively, the data indicate that

Fig. 2. GABAA receptors are up-regulated in BBMs. (A) Quantitative RT-PCR of GABA receptor subunits in HER2+ and TN BBM cells (COH-BBM 1–3, n = 3). (B)Immunohistochemistry of HER2+ and TN primary breast cancer (n = 8) and BBM tissue specimens (COH-BBM 1–3) at 40× magnification stained with HER2(green), GABAAR-a3 (red), and nuclear DNA (DAPI; blue). Primary and metastatic breast tumors (T) with bordering primary stroma (S) and brain tissue (B) areseparated by white dashed lines. Proliferation assays of HER2+ BBM cells treated with (C) GABA or (D) muscimol for 96 h. Cell proliferation was determined bythe percentage increase relative to nontreated cells (n = 3).

986 | www.pnas.org/cgi/doi/10.1073/pnas.1322098111 Neman et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

15,

202

0

BBMs possess GABAergic characteristics and have a GABAmetabolic phenotype.Reelin, a large secreted extracellular matrix glycoprotein, is

associated with GABAergic neurons that is crucial for thecytoarchitecture of laminated brain structures (35). During de-velopment, reelin is produced by a subset of neurons calledCajal–Retzius cells and is integral for their migratory capacity(36). Reelin has also been implicated in the migratory capacity ofprimary cancer cells (37–39). To further characterize GABAergicproperties, we determined the expression of reelin in BBM cellsrelative to primary breast cancer. Staining of tissue specimensshowed reelin expression was increased more than sixfold both inHER2+ primary and metastatic tumor cells relative to TN cells(Fig. 4A and Fig. S9 A and B). Because reelin and HER2 werecoexpressed in the HER2+ tumors, the spatial relationship ofthese two proteins was examined. We found that 99% of HER2protein colocalized with reelin in BBM tissue specimens (Fig. 4B).This was also the case for HER2+ BBM cells in vitro (Fig. S9 Cand D). TN breast cancer subtypes could have up to 30% of cellswith HER2 membrane staining but can still be diagnosed as HER2negative (40–42). Heterogeneity for HER2 expression in TNBBMs was also detected. Within the TN subtype, HER2 positivitywas associated with reelin expression, with 99% of HER2 protein

colocalizing with reelin (Fig. S9 C and D). Three-dimensionalrenderings from immunofluorescence of tissue and cells furthersupport reelin and HER2 colocalization (Fig. 4C, Fig. S9E, andMovie S1). Reelin and HER2 were coimmunoprecipitated to de-termine whether there was an interaction between the two pro-teins. Indeed, coimmunoprecipitates indicated HER2 directlyassociates with the reelin protein (Fig. 4D). We generated a 3Dstructural model of the HER2–reelin complex to identify putativeinteraction sites (Fig. 4E). The results suggested aromatic aminoacids that form aromatic cages, as well as polar residues that in-teract electrostatically at the interface between HER2 and reelin(Fig. 4E and Table S3).

DiscussionBBM cells were examined for expression of GABA-relatedvariables because GABAergic communication occurs in ∼40%of all synapses in the brain (43). Despite overexpression ofGABAAR mRNA and protein, tumor cell proliferation was notreceptor mediated. GABA signaling is known to modulate cellcycle and promote synergistic interactions with glutamate sig-naling, which could confer a survival advantage (44). AlteredGABA receptor or function could also affect neoplastic cellmigration and invasion, which are archetypal features of metas-tasis (45). Ultimately, the role of up-regulated GABAAR inBBM cells requires further study.We next investigated a potential metabolic mechanism un-

derlying enhanced tumor growth by GABA. The results showthat uptake of GABA and its subsequent catabolism via theGABA shunt increased NADH levels in the tumor microenvi-ronment, which conferred a proliferative advantage to tumors.GABA is abundant in brain cells and may be functioning as anoncometabolite. Noninvasive clinical metabolic imaging withmagnetic resonance spectroscopy in patients has shown the dis-tribution of GABA to be 1 μM/cm3 in human brain tissue (46). Atthis concentration, cultured BBM cells exhibited maximal pro-liferation. Total GABA levels can range even higher than reportedbecause GABA is present as the precursor for several other mol-ecules found in nervous tissue and cerebrospinal fluid, includingGABA-histidine (homocarnosine). Homocarnosine is present ex-clusively in the brain and cerebrospinal fluid where it functions as anantioxidant and modifies brain excitability (47). In vivo, homo-carnosine can be hydrolyzed by the enzyme homocarnosine/carno-sinase synthetase to form GABA and histidine (48). High levels ofGABA are also found in vitro; intracellular levels of GABA inastrocytes can exceed ∼2 mM, and these cells can release ∼700 μMGABA into the brain microenvironment (27).BBM cells from the brain microenvironment also display other

neuronal-like properties. A strong association was observedbetween HER2 and reelin, a protein expressed by a subset ofGABAergic neurons. It is possible that HER2 and reelin areinvolved in redundant or convergent pathways that control mo-tility and cytoskeletal shape. The reelin signaling cascade acti-vates phosphatidylinositide 3-kinase (PI3K), which triggers severaldownstream signaling molecules, including AKT, and results inthe activation of actin cytoskeletal proteins (49). Pathway analysisof Purkinje progenitor cells expressing reelin have identified ENAH/hMena as a stable gene required for regulating the actin cyto-skeleton (50). Interestingly, hMena is frequently overexpressed inthe HER2+ breast tumor subtype (51). The colocalization ofHER2 with reelin may facilitate the arrangement of HER2+

metastatic cells in the extracellular matrix similar to perineuronalnets that promote synaptic stabilization within the brain micro-environment. Furthermore, HER2 and reelin interaction couldenable cooperative modification of the cytoskeleton to promotetumor–stroma interface stability. Through reciprocal interactionfrom both neoplastic cells and the extracellular matrix, reelin couldbe responsible for tumor-cell masking to avoid immune surveillance.

Fig. 3. GABA transaminase promotes tumor cell proliferation in patientBBMs. (A) Quantitative RT-PCR of ABAT in HER2+ BBM cells (COH-BBM 1–2,n = 3). (B) Immunocytochemistry of HER2+ (COH-BBM1) and TN (COH-BBM3)BBMs at 63× magnification stained for cell membrane (WGA; white), ABAT(red), HER2 (green), and nuclear DNA (DAPI; blue). (C) Fluorescence in-tensities of ABAT in the tumor compartments of HER2+ and TN BBM tissuespecimens were quantified relative to the levels in the tumor compartment ofprimary breast cancer tissue specimens. (D) Proliferation assay of HER2+ BBMcells treated with increasing concentrations of GABA combined with 300 μMofthe GABA transaminase inhibitor, vigabatrin, for 96 h. Cell proliferation wasdetermined by percentage increase relative to nontreated cells (n = 3). (E)NADH content of control (untreated) and treated HER2+ BBM cells with 1 μMGABA and 1 μM muscimol with or without 300 μM vigabatrin.

Neman et al. PNAS | January 21, 2014 | vol. 111 | no. 3 | 987

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

15,

202

0

The cooperation of reelin and HER2 in brain metastases is an areaof ongoing investigation in our group.It is unknown whether metastases arise from an aggressive

progenitor pool that is present at the primary site early duringtumorigenesis or from the dissemination of cells that later adaptto the secondary site to colonize. Our results do not exclude theformer but do provide evidence for the latter. The need to un-derstand the biology of metastasis is increasingly relevant asclinical oncology strives for survival gains by preventing advancedstage IV breast cancer. The “seed and soil” hypothesis that framescurrent investigation of metastasis is uniquely exemplified by thecolonization of the brain by circulating breast cancer cells (52).Accordingly, therapy for brain metastases should use not onlycytotoxic approaches against the tumor cell but also perturbation of

tumor microenvironment that facilitates cancer cell growth andresilience. We may achieve the essential goal of improving andextending the lives of cancer patients by understanding the dynamicneoplastic ecosystem of brain metastases.

Materials and MethodsExperimental methods are detailed in SI Materials and Methods and includethe following: (i) generation of patient-derived tumor lines, (ii) gene ex-pression profiling, (iii) evaluation of patient datasets, (iv) proliferation assay,(v) NAD/NADH measurements, (vi) immunofluorescence staining and quan-tification, and (vii) computational methods for generating the structuralmodel of HER2–reelin complex.

Data are represented as mean values ± SEM. Statistical significance wasassessed using Student t test and one-way ANOVA ± Bonferroni’s multiple-comparison test.

A

B

Reelin

HE

R2

1% Reelin 60% HER2 99%

40%

D

E ReelinHer2C

HE

R2/

Ree

linD

AP

I

BBM2 HER2+

00

1800

*

*

*

*

1o TN BBM TN

HE

R2

Ree

lin

DA

PI

Mer

ge

T

S T T T

B

B

T

T

S T T T

B

B

T

T

S T T T

B

B

T

T

S T T T

B

B

T

Fig. 4. Patient BBMs express reelin. (A) Immunohistochemistry of HER2+ and TN primary breast cancer (n = 8) and BBM tissue specimens (COH-BBM 1–3) at40× magnification stained for HER2 (green), reelin (red), and nuclear DNA (DAPI; blue). Primary and metastatic breast tumors (T) with bordering primarystroma (S) and brain tissue (B) are separated by the white dashed lines. (B) Colocalization and quantification of HER2 and reelin in HER2+ BBM tissuespecimens. Yellow indicates complete colocalization. (C) A 180° 3D rendering from HER2+ BBM tissue stained for HER2 (green), reelin (red), nuclear DNA(DAPI; blue). The asterisk (*) highlights colocalization of HER2 and reelin. (D) Patient-derived BBM cells (n = 3) were used for coimmunoprecipitation of HER2or IgG and then blotted for HER2 and reelin. (E) Predicted 3D structural model of the HER2 (green) and reelin (red) complex. The amino acids (shown as stickmodel) in the interacting interface, which are outlined and enlarged in the Inset, show both aromatic clusters and charged interactions. Amino acid des-ignation: aspartic acid (D), glutamic acid (E), phenylalanine (F), histidine (H), lysine (K), glutamine (Q), arginine (R), tryptophan (W), and tyrosine (Y).

988 | www.pnas.org/cgi/doi/10.1073/pnas.1322098111 Neman et al.

Dow

nloa

ded

by g

uest

on

Oct

ober

15,

202

0

ACKNOWLEDGMENTS. We thankMargaret Morgan, PhD, and Nicola Solomon,PhD, for helpful comments and careful editing of this manuscript. We thankCharles Warden of the Bioinformatics Core at City of Hope for help withdata acquisition. We also thank Mrs. Ruth Roberts for her organizationalassistance to the project. R.J. is supported by National Institutes of Health

(NIH) Grant 2K12CA001727-16A1. J.N. is supported by California Institute forRegenerative Medicine Grant TG2-01150. J.T. is supported by NIH GrantR01CA176611-01. N.V. is supported by NIH Grants R01-GM082896 and R01-GM097261. Additional project support was provided by National CancerInstitute Grant P30 CA033572.

1. Lin NU, Winer EP (2007) Brain metastases: The HER2 paradigm. Clin Cancer Res 13(6):1648–1655.

2. Gupta GP, Massagué J (2006) Cancer metastasis: Building a framework. Cell 127(4):679–695.

3. Neman J, et al. (2013) Co-evolution of breast-to-brain metastasis and neural pro-genitor cells. Clin Exp Metastasis 30(6):753–768.

4. Percy DB, et al. (2011) In vivo characterization of changing blood-tumor barrierpermeability in a mouse model of breast cancer metastasis: A complementary mag-netic resonance imaging approach. Invest Radiol 46(11):718–725.

5. Wolff AC, et al. (2007) American Society of Clinical Oncology/College of AmericanPathologists guideline recommendations for human epidermal growth factor re-ceptor 2 testing in breast cancer. Arch Pathol Lab Med 131(1):18–43.

6. Patel SH, et al. (2012) ACR Appropriateness Criteria� follow-up and retreatment ofbrain metastases. Am J Clin Oncol 35(3):302–306.

7. Giuliano M, et al. (2011) Circulating tumor cells as prognostic and predictive markersin metastatic breast cancer patients receiving first-line systemic treatment. BreastCancer Res 13(3):R67.

8. Botteri E, et al. (2010) Modeling the relationship between circulating tumour cellsnumber and prognosis of metastatic breast cancer. Breast Cancer Res Treat 122(1):211–217.

9. Yu M, et al. (2013) Circulating breast tumor cells exhibit dynamic changes in epithelialand mesenchymal composition. Science 339(6119):580–584.

10. Cancer Genome Atlas Network (2012) Comprehensive molecular portraits of humanbreast tumours. Nature 490(7418):61–70.

11. Curtis C, et al. (2012) The genomic and transcriptomic architecture of 2,000 breasttumours reveals novel subgroups. Nature 486(7403):346–352.

12. Neman J, Somlo G, Jandial R (2010) Classification of genomic changes in breast cancerbrain metastasis. Neurosurgery 67(2):N18–N19.

13. Park ES, et al. (2011) Cross-species hybridization of microarrays for studying tumortranscriptome of brain metastasis. Proc Natl Acad Sci USA 108(42):17456–17461.

14. Magbanua MJ, et al. (2013) Genomic profiling of isolated circulating tumor cells frommetastatic breast cancer patients. Cancer Res 73(1):30–40.

15. Roberts E (2007) Gamma-aminobutyric acid. Scholarpedia 2(3356). Available at www.scholarpedia.org/article/Gamma-aminobutyric_acid.

16. Roberts E (1986) What do GABA neurons really do? They make possible variabilitygeneration in relation to demand. Exp Neurol 93(2):279–290.

17. Roberts E (1992) GABA and inhibition: Command-control in nervous system function.The Neurosciences: Paths of Discovery II, eds Samson F, Adelman G (Birkhauser, Boston),pp 87–105.

18. Roberts E (2000) Adventures with GABA: Fifty Years On (Lippincott Williams & Wilkins,Philadelphia).

19. Spoerri PE (1988) Neurotrophic effects of GABA in cultures of embryonic chick brainand retina. Synapse 2(1):11–22.

20. Ring H, Mendu SK, Shirazi-Fard S, Birnir B, Hallböök F (2012) GABA maintains theproliferation of progenitors in the developing chick ciliary marginal zone and non-pigmented ciliary epithelium. PLoS One 7(5):e36874.

21. Ludewig F, Hüser A, Fromm H, Beauclair L, Bouché N (2008) Mutants of GABAtransaminase (POP2) suppress the severe phenotype of succinic semialdehyde de-hydrogenase (ssadh) mutants in Arabidopsis. PLoS One 3(10):e3383.

22. Young SZ, Bordey A (2009) GABA’s control of stem and cancer cell proliferation inadult neural and peripheral niches. Physiology (Bethesda) 24:171–185.

23. Herrmann JE, et al. (2008) STAT3 is a critical regulator of astrogliosis and scar for-mation after spinal cord injury. J Neurosci 28(28):7231–7243.

24. Elston CW, Ellis IO (2002) Pathological prognostic factors in breast cancer. I. The valueof histological grade in breast cancer: Experience from a large study with long-termfollow-up. C. W. Elston & I. O. Ellis. Histopathology 1991; 19; 403-410. Histopathology41(3A):151–152, discussion 152–153.

25. Vik-Mo EO, et al. (2010) Brain tumor stem cells maintain overall phenotype and tu-morigenicity after in vitro culturing in serum-free conditions. Neuro-oncol 12(12):1220–1230.

26. Zhang C, Yu D (2011) Microenvironment determinants of brain metastasis. Cell Biosci1(1):8.

27. Lee M, McGeer EG, McGeer PL (2011) Mechanisms of GABA release from human as-trocytes. Glia 59(11):1600–1611.

28. Greenbaum D, Colangelo C, Williams K, Gerstein M (2003) Comparing proteinabundance and mRNA expression levels on a genomic scale. Genome Biol 4(9):117.

29. Roberts E (1990) Dehydroepiandrosterone (DHEA) and its sulfate (DHEAS) as neuralfacilitators: Effects on brain tissue in culture and on memory in you and old mice. Acyclic GMP hypothesis of action of DHEA and DHEAS in nervous system and othertissues. The Biological Role of Dehydroepiandrosterone (DHEA), eds Kalimi M, RegelsonW(de Gruyter, Berlin), pp 13–42.

30. Lüddens H, et al. (1990) Cerebellar GABAA receptor selective for a behavioural al-cohol antagonist. Nature 346(6285):648–651.

31. Yazulla S, Brecha N (1981) Localized binding of [3H]muscimol to synapses in chickenretina. Proc Natl Acad Sci USA 78(1):643–647.

32. Larsson OM, Schousboe A (1990) Kinetic characterization of GABA-transaminase fromcultured neurons and astrocytes. Neurochem Res 15(11):1073–1077.

33. Chattopadhyaya B, et al. (2007) GAD67-mediated GABA synthesis and signalingregulate inhibitory synaptic innervation in the visual cortex. Neuron 54(6):889–903.

34. Caillard O, et al. (2000) Role of the calcium-binding protein parvalbumin in short-termsynaptic plasticity. Proc Natl Acad Sci USA 97(24):13372–13377.

35. Franco SJ, Müller U (2011) Extracellular matrix functions during neuronal migrationand lamination in the mammalian central nervous system. Dev Neurobiol 71(11):889–900.

36. Tissir F, Goffinet AM (2003) Reelin and brain development. Nat Rev Neurosci 4(6):496–505.

37. Yuan Y, Chen H, Ma G, Cao X, Liu Z (2012) Reelin is involved in transforming growthfactor-β1-induced cell migration in esophageal carcinoma cells. PLoS One 7(2):e31802.

38. Dohi O, et al. (2010) Epigenetic silencing of RELN in gastric cancer. Int J Oncol 36(1):85–92.

39. Stein T, et al. (2010) Loss of reelin expression in breast cancer is epigenetically con-trolled and associated with poor prognosis. Am J Pathol 177(5):2323–2333.

40. Bastien RR, et al. (2012) PAM50 breast cancer subtyping by RT-qPCR and concordancewith standard clinical molecular markers. BMC Med Genomics 5:44.

41. de Ronde JJ, et al. (2010) Concordance of clinical and molecular breast cancer sub-typing in the context of preoperative chemotherapy response. Breast Cancer ResTreat 119(1):119–126.

42. Skarlos P, et al. (2012) Triple-negative phenotype is of adverse prognostic value inpatients treated with dose-dense sequential adjuvant chemotherapy: A translationalresearch analysis in the context of a Hellenic Cooperative Oncology Group (HeCOG)randomized phase III trial. Cancer Chemother Pharmacol 69(2):533–546.

43. Roberts E (1986) Failure of GABAergic inhibition: A key to local and global seizures.Adv Neurol 44:319–341.

44. Ikeda Y, Nishiyama N, Saito H, Katsuki H (1997) GABAA receptor stimulation pro-motes survival of embryonic rat striatal neurons in culture. Brain Res Dev Brain Res98(2):253–258.

45. Hanahan D, Weinberg RA (2011) Hallmarks of cancer: The next generation. Cell144(5):646–674.

46. Ke Y, Cohen BM, Bang JY, Yang M, Renshaw PF (2000) Assessment of GABA con-centration in human brain using two-dimensional proton magnetic resonance spec-troscopy. Psychiatry Res 100(3):169–178.

47. Grove J, et al. (1982) Concentration gradients of free and total gamma-aminobutyricacid and homocarnosine in human CSF: Comparison of suboccipital and lumbarsampling. J Neurochem 39(6):1618–1622.

48. Kish SJ, Perry TL, Hansen S (1979) Regional distribution of homocarnosine, homo-carnosine-carnosine synthetase and homocarnosinase in human brain. J Neurochem32(6):1629–1636.

49. Chai X, Förster E, Zhao S, Bock HH, Frotscher M (2009) Reelin stabilizes the actin cy-toskeleton of neuronal processes by inducing n-cofilin phosphorylation at serine3.J Neurosci 29(1):288–299.

50. Paul A, Cai Y, Atwal GS, Huang ZJ (2012) Developmental coordination of gene ex-pression between synaptic partners during GABAergic circuit assembly in cerebellarcortex. Front Neural Circuits 6:37.

51. Di Modugno F, et al. (2010) The cooperation between hMena overexpression andHER2 signalling in breast cancer. PLoS One 5(12):e15852.

52. Fidler IJ (2003) The pathogenesis of cancer metastasis: The “seed and soil” hypothesisrevisited. Nat Rev Cancer 3(6):453–458.

Neman et al. PNAS | January 21, 2014 | vol. 111 | no. 3 | 989

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Oct

ober

15,

202

0