Breast Cancer Research and Treatment 22: 141-151, 1992. © 1992 KluwerAcademic Publishers. Printed in the Netherlands.

Report

Epidermal growth factor receptor in breast cancer: Storage conditions affecting measurement, and relationship to steroid receptors W.R.B. McLeay ~, D.J. Horsfall ~, R. SeshadrF, D.A. Morrison 1and G.T.R Saccone ~ 1Departments of Surgery and 2Haematology, Flinders Medical Centre, Bedford Park, S.A. 5042 Australia

Key words: breast cancer, epidermal growth factor receptors, steroid receptors, quality control Summary This study investigates the effect of freezing and storage of tissue and subcellular fractions on the measurement of epidermal growth factor receptors (EGF-r); compares competition binding and single saturating dose assays (SSD) for quantitating EGF-r levels; investigates several tissues as potential quality control; and examines the relationship between EGF-r and hormone receptor expression in human breast cancers. Mouse and calf uterine cell membranes were preferred sources of quality control tissue with similar levels of high affinity EGF-r to human breast cancer tissue (< 150-200 fmol/mg membrane protein). Studies using pooled mouse uterine tissues indicated a loss of 40% in EGF-r activity following a single - 2 0 ° C freeze/ thaw cycle, while a breast cancer tissue showed a 75% loss, independent of storage temperature (liquid nitrogen, - 7 0 ° C, - 2 0 ° C). A single freeze/thaw cycle of mouse uterine broken cell pellets (nuclei plus membrane fraction) again indicated a loss of EGF-r irrespective of storage temperature (43 % loss a t - 70 ° C, 52% loss a t - 20 ° C). In most cases irrespective of the tissue type or tissue fraction being stored, the length of storage had little impact on the extent of the loss in activity. A second freeze/thaw cycle of intact tissue, or freezing of broken cell pellets from a previously-frozen tissue, led to a further major or total loss of the remaining EGF-r. Overall these results are commensurate with the published effects of freezing and storage on estrogen receptor measurement. In addition, our studies suggest that the most suitable procedure for assaying frozen breast cancer specimens for EGF-r levels in conjunction with steroid receptor quantitation is to prepare and assay both cytosol and membrane fractions for their respective receptor content without further storage. A concordance of 86% was found in 44 breast cancers assayed for EGF-r by saturation analysis and SSD. Statistically significant inverse relationships were found between EGF-r and estrogen and progesterone receptor levels in a study of approximately 350 breast cancer patients. No association was found with tumor stage or diameter, axillary node involvement, or patient age.

Introduction The value of EGF-r measurement of breast cancer management is still a matter of some debate. Most studies indicate a relationship between high EGF-r levels and low or absent estrogen receptor (ER)

[1-7]. However, while some prospective studies indicate that high EGF-r levels are associated clinically with poor prognosis and tumor progression [1, 8], and that EGF-r may be the most important prognostic indicator for node negative patients [9, 10], others find no relevance for EGF-r determina-

Address f o r offprints: D.J. Horsfall, Department of Surgery, Flinders Medical Centre, Bedford Park, 5042, South Australia

142

W McLeay et al.

tion in prognosis [11, 12]. The reasons for such divergent conclusions on the prognostic implications of tumor EGF-r levels need to be determined, but could be related to differences in either EGF-r assay technique, patient selection, or treatment protocols. Because of the extensive and very efficient tumor collection systems which have been developed for the regional steroid receptor laboratories that exist in most Australian states and internationally, these laboratories are well placed to analyze EGF-r levels in breast cancer tissues in conjunction with steroid hormone receptors. Unfortunately, however, minimal effort has been made to establish guidelines for the standardization of storage and assay conditions between laboratories. While a number of studies have reported the effects of different storage conditions on the biochemical measurement of steroid receptors in breast cancer and other human tissues [13, 14], similar studies are lacking for EGF-r assays. Furthermore, no consensus exists regarding an appropriate 'cut-off' value below which a sample is considered to be EGF-r negative. Arbitrarily chosen cut-off levels have ranged from 1-30 fmol/mg of membrane protein in various publications [3-5, 11, 15]. The proven costeffectiveness of batch-testing of specimens for hormone receptor assays led us to investigate whether it is feasible to routinely store broken cell pellets, obtained at the time of cytosol preparation for steroid receptor analysis, for later batchwise determination of EGF-r, as has been reported [16]. This, of necessity, involves an additional freeze/ thaw cycle, a manoeuvre which we anticipated might have a detrimental effect on EGF-r quantitation. This study reports the examination of various tissue sources for the purpose of developing a suitable quality control, and the investigation of the effects of storage conditions on tissues and subcellular fractions prior to EGF-r measurement. In addition, we confirm the inverse relationships between the expression of EGF-r and steroid receptors in human breast cancers.

Patients and methods

Patients In this study EGF-r levels were determined in tumor biopsies from 352 women diagnosed with primary breast cancer in South Australia between November 1989 and December 1990. All patients were treated by either partial (25% of patients) or total (75% of patients) mastectomy plus axillary lymph node clearance. Since one of the aims of this study was to evaluate the association of EGF-r with pathological determinants of poor prognosis, including axillary lymph node involvement, patients with stage IV disease or in whom axillary clearance was not performed were excluded. All patients were staged using UICC criteria [17].

Tissues Except where noted in 'Results', samples of primary breast cancer tissues were routinely snap-frozen and stored at - 2 0 ° C within 30min of biopsy, and assayed for steroid hormone and EGF receptors within 7 days of collection. Human placentae were obtained on ice from the Department of Obstetrics and Gynaecology, Flinders Medical Centre. Uteri from 5-day old calves were obtained on ice from the local abattoir. Mouse uteri were obtained from freshly-killed young adult Balb/C mice. The breast cancer cell line MDA-MB-231 was obtained from ATCC (Bethesda, MD) and cultured in RPMI 1640 medium containing 5% fetal bovine serum (Flow Laboratories, Australia).

Preparation of broken cell pellets Samples of fresh tissue were homogenized in icecold cytosol buffer (containing 10 mM Tris pH 7.4, 10mM sodium molybdate, 1.5 mM ethylenediaminetetraacetic acid (EDTA), 10% glycerol, and 1mM dithiothreitol) with an Ultra-Turrax (IKAWerk, Staufen i. Breisgau, Germany) for two 10second bursts. Samples of frozen tissue were pulverized before homogenization using a stainless

Epidermal growth factor receptor in breast cancer

steel mortar chilled by immersion in liquid nitrogen. Following homogenization, broken cell pellets containing membrane were separated from the cytosol fraction by centrifugation (100,000g 30min, 4 ° C). Except where indicated in 'Results', broken cell pellets were immediately used to isolate the membrane fraction.

Preparation of membrane fraction

Broken cell pellets were resuspended in ice-cold buffer containing 10 mM Tris and 50 mM NaC1 (pH 7.4) and homogenized as described above. The suspensions were then centrifuged at 800g for 10 min at 4°C and the supernatants containing the membrane fraction were separated from the residual nuclear pellets.

EGF-r assay

A competition binding assay similar to that described by Nicholson et al. [5] was used in this study. Briefly, saturation analyses were performed in duplicate using 6 concentrations of 125I-labelled mouse EGF ranging from 0.25 nM to 3 nM incubation concentrations, with and without the presence of a 100-fold excess of unlabelled EGE The reaction mixture contained 50 gl of membrane fraction with a mean protein concentration of 2mg/ml, 50 gl of Tris/NaC1 buffer containing 2% bovine serum albumin (BSA), and 10gl of 125I-EGE After one hour incubation at 22 ° C, the reaction was terminated by the addition of 0.5 ml of ice-cold 0.5% BSA in Tris/NaC1 buffer. Bound and free ~25I-EGF were separated by centrifugation at 12,000g for 5 min. Total and non-specific binding data were determined in an NEC 1600 gamma counter, and the results were calculated by Scatchard plot with nonlinear least squares regression using software for the analysis of multiple binding sites (Beckman Accufit TMsaturation two-site software, Fullerton CA). In some studies, single saturating dose (SSD) incubations were performed in quadruplicate using a 1nM final concentration of 125I-EGF with and without 100-fold excess of unlabelled EGE The re-

143

mainder of the assay was performed as described for the saturation analysis. EGF-r concentration was expressed in fmol/mg membrane protein, and a _>10 fmol/mg protein cut-off taken for receptor positivity. Protein concentrations were determined by the method described by Lowry et al. [18]. Lyophilized mouse EGF (Sigma, St. Louis MO) was stored at - 2 0 ° C, reconstituted with distilled water (200 gg/ml), aliquoted, and refrozen until required. EGF was radio-iodinated by the chloramine T method [19] with specific activity of 81+ 12.8 gCi/ gg (mean+ SEM, n= 10). Iodinated EGF was aliquoted and frozen (-20°C) until required. Working dilutions for radioligand binding assays were kept at 4°C for 3 weeks.

Steroid receptor assay

Estrogen receptor and progesterone receptor (PR) levels were determined in breast cancer tissues using saturation analysis as reported previously [20]. Briefly, the assays employed 5 incubation concentrations ranging from 0.05 to 2nM [3H]-estradiol (Amersham Aust.) and 0.08 to 4nM [3H]R5020 (promegestone, NEN-Dupont Aust.) to determine total ER and PR binding respectively. Parallel tubes containing an additional 100-fold excess of unlabelled steroid were used to estimate the levels of non-specific binding. Bound and free ligand were separated using dextran-coated charcoal, and results were analysed by Scatchard plot and least squares regression using software for the analysis of a single binding site (Beckman Accufit TMsaturation one-site software). Hormone receptor status was determined using a cut-off for positivity of concentration > 10 fmol/mg cytosol protein.

Statistical methods

Comparison of EGF-r status with specified pathological variables was performed by Chi-squared analysis of contingency tables. Quantitative EGF-r and steroid receptor data were compared using Kendall's rank correlation test. Statistical analysis

144

0O, . W McLeay et al.

levels. Five parallel Scatchard estimations of EGF-r level in a pool of mouse uterine tissue used to estimate the intra-assay variability recorded a mean+ SEM of 66.6+ 9.1 fmol/mg membrane protein and CV of 30.7%. Interassay variability using individual fresh mouse uteri on a weekly basis recorded a mean+ SEM and % CV of 43.8+ 3.5 and 35.6% respectively (n= 20).

0.04

m g

.

100

200

300

EGF bound fmel/mg membrane protein

Fig. 1. Curvilinear Scatchard plot of human breast cancer cell membrane EGF-r. No. of high affinity binding sites (Bmax) = 151 fmol/mg membrane protein; dissociation constant (Kd)= 0.23 nM.

between groups of continuous variables was carfled out by Wilcoxon rank sum test.

Results

Choice of suitable quality control tissue Several tissues were examined for their suitability as internal controls for EGF-r assays and for assessing appropriate conditions for storage of breast cancer tissues prior to EGF-r quantitation. Tissues tested were neonatal (5-days old) calf uterus, young adult mouse uterus, human placenta, and the breast cancer cell line MDA-MB-231. Membrane preparations from these unfrozen tissue sources were assayed without delay by saturation analysis. Excluding the two samples of calf uterus, curvilinear plots illustrating the presence of both high and low affinity binding components were obtained for all normal tissues tested, and for 80% (16/21) of breast cancers, (Table 1, Fig. 1). Membrane preparations from mouse and calf uterine tissue were deemed the most suitable sources of routine quality control tissue since they contained levels of high affinity receptor sites similar to those observed in the majority of human breast tumors (Table 1). For this reason, mouse uterine tissue was used as a source of EGF-r positive tissue with which to investigate the effect of tissue storage on EGF-r

Lability of EGF-r following thawing and refreezing of stored tissues In view of the lability of steroid receptors following multiple freeze/thaw cycles [21], we investigated whether thawing and refreezing of tissue samples was also detrimental to EGF-r. Uteri from 28 mice were pooled and divided to form 4 large representative aliquots of uterine tissue. One aliquot was processed fresh and membrane EGF-r determined. The other 3 aliquots were frozen at - 2 0 ° C. Two of these aliquots were assayed for EGF-r content after 7 and 21 days of storage, while the third aliquot underwent an additional freeze/thaw cycle before EGF-r assay. As shown in Table 2, freezing and subsequent storage of mouse uterine tissue for

Table 1. Measurement of high affinity EGF-r binding sites in various tissues" considered as candidates for quality control Membrane preparation

Scatchard kinetics

Placenta

curvilinear

MDA-MB-231

curvilinear

Calf uterus

linear

Mouse uterus

curvilinear

Breast cancer

curvilinear

EGF-r binding sites Bmax b

Kdc

128(7) 74-322 320(2) 143-404 16(2) 8-24 61(5) 47-100 38(21) 7-151

0.18 0.04-0.60 0.35 0.17-0.52 0.27 0.18-0.35 1.51 1.10-2.13 1.04 0.08-3.12

Tissues were fresh on ice, membranes prepared and assayed within 2 hours after surgery. ~Bmax= no. of binding sites in fmol/mg protein; median and range, no. of values in parenthesis. cKd= dissociation constant (riM).

Epidermal growthfactor receptor in breast cancer up to 7 days appeared to have a slight effect on EGF-r levels, with a total loss of 60% after 21 days of storage. The dissociation constant was also reduced, indicating an increased lability of lower affinity binding sites to freeze/thawing and storage at - 2 0 ° C . No EGF-r could be detected in tissue which experienced two freeze/thaw cycles.

30-

g

20-

-6

_

,'v ii

10-

? Lability of EGF-r during freezing and storage of broken cell pellets prior to batch analysis

145

0

'

-5

I

0

'

o I

5

9 '

I

'

10

I

15

'

I

'

20

I

25

Time (days)

To assess whether broken cell pellets resulting from tissue cytosol extraction procedures can be stored for later batchwise determination of EGF-r levels, the lability of EGF-r following storage of broken cell pellets at different temperatures was examined. Uterine tissue from 40 mice was pooled and homogenized, and a broken cell pellet was prepared. An aliquot was used to separate the cell membrane fraction and immediately assayed in triplicate using the SSD method. The remainder of the broken cell preparation was aliquoted and pellets were stored at - 2 0 ° C or - 7 0 ° C for 4-21 days prior to membrane preparation and triplicate EGF-r assay. A significant loss of EGF-r was observed (Fig. 2) as a result of freezing of broken cell pellets (p < 0.05, Wilcoxon rank sum test). The loss was 52% a t - 20 ° C and 43 % a t - 70 ° C compared to the value obtained for fresh tissue. After this initial fall, there was no appreciable loss over 21 days of storage.

Table

2.

Fig. 2. The effect of freezing and storage of broken cell pellets of mouse uteri prior to EGF-r determination. ~ Broken cell pellet from unfrozen tissue; O Broken cell pellet frozen at-20°C; • Broken cell pellet frozen at - 7 0 oC. Receptor analysis by SSD assay; median and range of 3 values per point plotted.

Lability of EGF-r during freezing and storage of human breast tumor tissue, and of residual broken cell pellets obtained after cytosol extraction of frozen breast cancer tissues The majority of cancers delivered to regional diagnostic receptor laboratories are of necessity frozen at - 2 0 ° C. Since broken cell pellets from unfrozen uterine tissues can be frozen and the presence of EGF-r determined albeit at reduced levels, we investigated whether this procedure could be adopted to batch broken cell pellets from frozen tissues, for later EGF-r analysis. In order to investigate the viability of this approach and to confirm the appropriate storage temperature for breast tumor tissue in addition to broken cell pellets, the experiment outlined in Table 3 was performed. The results in-

Effect of freeze/thaw cycles on stored mouse uterine tissues before EGF-r assay

Storage condition of tissue prior to assay

EGF-r concentration fmol/mg protein

Dissociation constant nM

Fresh tissue Frozen tissue stored at - 2 0 ° C for 7 days Frozen tissue stored at - 2 0 ° C for 21 days Tissue thawed at 7 days, refrozen and stored - 2 0 ° C for an extra 7 days

61(47-100) a 57(44-69) b 24(18-29) b 0~

1.51(1.10-2.13)

"Median (range) of 5 Scatchard analyses of pooled mouse uteri. bMean of duplicate analyses. cDuplicate estimations, both negative.

1.05(l.03-1.07) 0.52(0.74-1.29)

146

W McLeay et al.

Table 3. The effect of freezing human breast cancer tissue and subsequent freezing of broken cell pellets after cytosol preparation prior to EGF-r measurementa Storage condition

Storage temperature

EGF-r concentration fmol/mg protein

Dissociation constant nM

Fresh tissue Frozen tissue Frozen tissue Frozen tissue Frozen pellet Frozen pellet Frozen pellet

Liquid nitrogen - 70 ° C - 20 ° C Liquid nitrogen - 70 ° C - 20 ° C

53 12 14 16 0 0 0

2.00 0.46 0.30 0.67

"A large unfrozen breast biopsy was divided into seven pieces. One aliquot was immediately homogenized for cytosol extraction without freezing and the membrane fraction prepared for EGF-r assay by saturation analysis. The other 6 tissue aliquots were frozen in pairs and stored at the 3 indicated temperatures for 7 days before preparation of broken cell pellets. One of the paired tissue pellets for each temperature was used to prepare membranes and immediately assayed for EGF-r, while the second pellet was refrozen for 7 days at the same temperature before membrane preparation and EGF-r assay.

dicate that freezing of a large breast cancer tissue for 7 days resulted in loss of low affinity receptors, as evidenced by the fall in Kd and a corresponding fall in the total number of EGF-r. The reduction in EGF-r level was similar (median 74%) between the three storage temperatures (liquid nitrogen, -70°C, -20°C). Most important, subsequent freezing of broken cell pellets containing membrane fractions led to a total loss of EGF-r regardless of storage temperature. Confirmation of the loss of EGF-r induced by freezing broken cell pellets containing membrane

fraction was sought by assay of EGF-r in a series of breast cancer specimens which had been frozen for less than 7 days. As outlined in Table 4, residual broken cell pellets after cytosol preparation for hormone receptor determination were divided into 2 aliquots; membrane prepared from one aliquot was assayed fresh, while the other was frozen at -20°C. In those biopsies (n= 6) with measurable EGF-r, membrane was prepared from the frozen pellets and assayed for EGF-r levels after 2-21 days of storage. The results indicate that freezing of broken cell pellets resulted in either a complete

Table 4. EGF-r quantitation in frozen human breast cancer tissues, and in the residual broken cell pellets frozen for a further 2-21 days at - 2 0 ° C Patient tissue biopsy

1 2 3 4 5 6

EGF-r determination on frozen breast cancer tissue

EGF-r determination on frozen broken cell pellet

Storage time of broken cell pellet

Bmax"

Kd b

B max

Kd

Days

% loss of binding

97 120 28 231 17 53

3.69 1.43 3.34 3.42 0.68 0.29

33 37 0 67 0 17

2.00 0.36

2 5 7 14 14 21

66 69 100 71 100 68

1.20 0.02

"EGF-r concentration (fmol/mg protein). bDissociation constant (riM). EGF-r quantitation was performed using Beckman AceufitTM saturation two-site software. Frozen breast cancer tissues from individual patients were powdered in a stainless steel mortar at liquid nitrogen temperatures, homogenized in buffer, and centrifuged to obtain cytosolic extracts and broken cell pellet as per Patients and Methods. Pellets were divided, and cell membranes prepared immediately and assayed for EGF-r by saturation analysis in one portion, while the remaining pellet was frozen and stored before membrane preparation and EGF-r assay.

Epidermal growthfactor receptor in breast cancer

or a major loss of measurable EGF-r (median toss 70%), and was independent of storage time.

147

of _>10 fmol/mg protein. EGF-r status was not related to any other pathological determinant studied, including stage, tumor diameter, axillary node involvement, or patient age (Table 6).

Validity of SSD method to quantitate EGF-r in breast cancers Discussion

EGF-r levels were measured in 44 breast cancers by both the saturation analysis and SSD methods. While the SSD method in most cases underestimated the level of EGF-r by 50% when compared with saturation analysis (data not shown), the concordance of the 2 methods, using a cut-off of >_10 fmol/mg protein for positivity in both cases, was high (38144, 86%, Table 5). Sensitivity and specificity were also high (89% and 84% respectively). The predictive values for EGF-r positive and negative tumors were also high (91% and 81% respectively). SSD assays were performed in a total of 352 breast cancers with a median EGF-r concentration of 5.2 fmol/mg protein (range 0-468 fmol/ mg; first quartile 1.1 fmol/mg; third quartile 12.4 fmol/mg protein).

Association of EGF-r concentration with pathological features of breast cancer

An inverse correlation of EGF-r concentration with both ER and PR concentration is shown in Fig. 3A & B. Kendall rank correlation (R) for ER = -0.2682, p< 0.0001, and for PR= -0.1689, p< 0.0001, using arbitary cut-off values for positivity

The importance of EGF-r determination for breast cancer prognostication remains a contentious issue (1,8-12,15) and it is not possible currently to fully evaluate the influence that different methodologies for measuring EGF-r have had on the highly variable incidence (22-57%) and mean level (4-72 fmol/mg protein) reported for EGF-r positive human breast cancers [2-5, 7, 15]. There is little information available on the stability of EGF-r during freezing or storage of tissues, and some studies [8] have extrapolated from experience obtained with steroid hormone receptor measurements [13]. Tissue storage times vary from several days [this study] to years [16], or more often are not stated [4, 6, 7, 22]. Others report the use of unfrozen tissues [2, 11]. The differences in methodological approach to EGF-r determination, i.e. saturation analysis versus displacement [8, 11] or SSD assays, the variation in concentration of radioligand used in the SSD assay (0.3-5nM) [3, 5, 6, 15, 23], and the wide range of EGF-r cut-offused for positivity (1-30 fmol/mg) [3, 5, 15, 22], will all contribute to the enormous variability in reported associations of EGF-r with ER, PR, and pathological indices such as tumor grade and axillary node involvement

Table 5. Sensitivity, specifcity, and predictive value of SSD assay compared to saturation analysis for EGF-r quantitation in 44 breast cancers Concordance~

Sensitivityb

Specificity ~

Predictive value + r e a

Predictive vaIue - v e ~

38/44 86%

17/19 89%

21/25 84%

17/21 81%

21/23 91%

Cut-off value for EGF-r positivity by both SSD and saturation analysis >10 fmol/mg protein. Cancers were frozen at -20°C, and assayed within 7 days. aConcordance = (TP+ TN)/(TP+ FP + TN + FN) × 100%. bSensitivity = TP/(TP+ k'N) × 100%. ~Specificity= TN/(TN+ FP) x100%. dPredictive value of a positive result= TP/(TP+ FP) × 100%. ~Predictive value of a negative result= TN/(TN+ FN) x 100%. TP, true positive; FP, false positive; TN, true negative; FN, false negative.

148

W McLeay et al. (xl000 2

c

0

=0

1.6

1.2

--0 0

0

O d~

0

E

0.8

_

.O

O

W

0.4 £3

[] i

t

i

i

0

i

L

i

i

i

100

E

i

i

i

I

200

T

i

i

I

300

i

i

i

i

|

4O0

50(

400

500

EGF-r fmol/mg membrane protei n (xl000) i

2

[

'

~

'

~--"1

¸

,

L

,

,

I

'

'

'"'

~

'

'

~

D

D

c

1.6

--

[] £3 D

Q. 0

o

1.2

D []

0

E] D

E

0.8

__

g O

.

.

.

.

.

.

.

O

O

E n" n

0.4 3£3

D % 0

100

~z

o• 200

.

.

.

.

.

.

.

.

.

.

300

EGF-r fmol/mg membrane protein Fig. 3. Correlation of quantitative EGF-r and steroid receptor binding. A) ER, Kendall rank correlation R= -0.2682, p < 0.0001 (n = 348); B) PR, Kendall rank correlation R= -0.1689, p < 0.0001 (n= 352). ER, PR, EGF-r cut-off values for positivity of _>10 fmol/mg protein.

Epidermal growth factor receptor in breast cancer

[2, 4, 5-8, 22]. Since no two studies have apparently used identical storage and assay methodologies, some standardization is required, as has been recognised recently [5, 15, 16]. Our study methodology was based on that of Nicholson et al. [5] but is also similar to that recently reported by Formento et al. [6]. Because of tumor size limitations, all three studies have initially validated SSD assays against saturation analysis and then assayed a larger number of tumors using an SSD and with an incubation concentration of 1nM, an EGF-r positivity cut-off of > 10 fmol/mg protein. Of the breast cancers studied by saturation analysis in this study, 80% had curvilinear plots denoting high and low affinity binding sites. As also reported by the other 2 groups, we found that the SSD assay underestiTable 6. Association o f E G F - r concentration with pathological features of breast c a n c e r Pathological feature

E G F - r concentration ~ 10

20

6

n = 352

4.71

A g e (years) 50

165

67

p = N.S.

3.69

10

182

54

p =