Identification and Purification of Functional Human β-cells by a New Specific Zinc-fluorescent Probe

Human Islet Processing

Human pancreata (n = 11; mean age 35 ± 12 years) were harvested from adult brain-dead donors in accord with French Regulations and with the local Institutional Ethical Committee. Pancreatic islets were isolated after ductal distension of the pancreata and digestion of the tissue with Liberase (Roche Diagnostics; Meylan, France) according to the automated method of Ricordi et al. (1988), with modifications (Kerr–Conte et al. 1994). Semipurification was achieved with Histopaque (Sigma; Saint-Quentin Fallavier, France) discontinuous density gradients using a cell separator (Lake et al. 1989). Islet number and diameter were determined on triplicate samples of each preparation after dithizone (DTZ) staining (Latif et al. 1988) and expressed in islet equivalents (IEs) corresponding to islets with a diameter of 150 μm (Ricordi et al. 1990). Islet viability was assessed in each preparation by microscopic examination after staining with DTZ and trypan blue. Preparations used in this study exhibited an 86 ± 5% viability and an average yield of 2320 ± 420 IE/g. Semipurified islets were cultured for 24 hr at 37C in CMRL 1066 medium with 2% Ultroser G (Gibco BRL; Life Technologies, Cergy–Pontoise, France) and antibiotics.

Cell Culture

Pancreatic β-cells of murine origin, including INS-1 kindly obtained from Dr. Wolheim (Asfari et al. 1992) and β-TC3 from Dr. Efrat (Efrat et al. 1988), were used to ascertain some features of the probe. Culture medium for INS-1 cells consisted of RPMI-1640 containing 10% fetal calf serum (FCS), 1.1 mM sodium pyruvate, 50 μM 2-mercaptoethanol, and antibiotics. For β-TC3 cells, DMEM containing 2.5% FCS, 15% horse serum, and antibiotics was used.

Fluorescence Labeling with Newport Green

Newport Green (Molecular Probes Europe; Leiden, The Netherlands) is available as a salt (potassium form, NG-K) or an ester (diacetate form, NG-Ac). The water-soluble potassium form was used for fluorometric studies designed to determine both binding constants from titration curves of ZnCl₂ and specificity from other metal ion solutions. NG-K (6 μM) was added to 96-well microplates and left in the dark for 10 min. Fluorescence was measured at room temperature in a fluorophotometer equipped with microtiter plate accessories (Fluorocount; Packard, Rungis, France). The Newport Green indicator exhibited an increase in fluorescence emission on binding Zn, with a slight shift in wavelength; single excitation and emission spectral peaks were measured at wavelengths of 485 and 530 nm, respectively.

The ester form (NG-Ac) was used for fluorescent studies on living cells. NG-Ac is cell-permeant and can therefore diffuse across cell membranes. Once inside the cells, this ester is cleaved by intracellular esterases to yield a cell-impermeant fluorescent indicator able to bind zinc. Cells in PBS were exposed for 30 min at 37C to 25 μM NG-Ac containing 1.5 μl/ml Pluronic F127 to enhance the penetration of the probe. After washing in PBS, the fluorescence was estimated in microwell plates as described above or analyzed by confocal microscopy or cell sorting. Standardization of fluorescence in spectrofluorometric measures was achieved using a blank consisting of 25 μM NG-Ac in PBS set at a fixed and constant unit of fluorescence.

Cell Distension and Dissociation

Distension of islets was performed for 5 min at 37C in a dispersion buffer consisting of Earle's balanced salt solution without Ca²⁺ and Mg²⁺ and supplemented with 10 mM HEPES, 3 mM EGTA, 2.8 mM glucose, and 2.5 mg/ml BSA.

Total dissociation of cells from islet preparations was achieved on islet pellets (200 × g, 3 min) by gentle pipetting for 10 min in 4.5 ml dispersion buffer. An enzymatic dissociation was then performed by the addition of 100 μg/ml trypsin. The reaction was stopped by 2 volumes FCS and 1 volume 0.035% (w/v) trypsin inhibitor (Type 1-S; Sigma) when about 80% of islet cells appeared as single cells. The cell suspension was centrifuged (500 × g, 5 min) and the pellet suspended in F10 medium without Ca²⁺ and containing 2.8 mM glucose. The cell suspension was then filtered through a 70-μm nylon screen.

Confocal Microscopy

Confocal microscopic analyses were performed using a confocal inverted microcope (Leica TCS-NT; Rueil–Malmaison, France) on dissociated cells or on whole islets plated in Lab-Tek chambers with coverslips (Nalge Nunc International; Naperville, IL) coated with poly-l-lysine. When islets were particularly compact, we performed a mild dissociation of islets before staining for 5 min in the dispersion buffer to prevent defective penetration of the probe. Staining was performed after a period of 90 min in F10 medium containing 2.8 mM glucose to restore cell integrity.

Cell Sorting

Cell sorting was performed after total dissociation of islet cells. Staining was achieved after a period of 90 min in F10 medium without Ca²⁺ and containing 2.8 mM glucose.

We used either an Epics XL-MCL or an Epics Elite Coulter flow cytometer equipped with an argon-ion laser (Coultronics; Margency, France). Excitation was performed with the 488-nm blue line of the laser and emission measured through a 530/30 bandpass filter.

Insulin Content and Secretion

Insulin content in human pancreatic islets was measured after cell pelleting by centrifugation and homogenization by sonication (20 Khz, 30 W) for 15 sec in 200 μl Tris-HCl containing 1 mM EDTA, pH 7.4, and frozen-stored at −20C. The islet insulin content was measured by radioimmunometric assay (Bi-insulin; IRMA Diagnostics, Pasteur, France) after an overnight acid–ethanol (0.18 M HCl in 95% ethanol) extraction of 50-μl aliquots of the homogenates at 4C. Insulin secretory responses were studied as follows: 1 hr preincubation in Kreb's buffer containing 3.6 mM glucose followed by three successive incubation periods of 1 hr each with 3.6, 16.7, and 3.6 mM glucose. Incubation media were collected for assay of insulin secretion from the cells. The stimulation indices were calculated by dividing insulin secretion in the presence of high glucose by mean basal insulin secretion levels.

DNA Assay

DNA was assayed in aliquots of homogenates using Pico Green (Molecular Probes) as the fluorescent probe at a 1:200 dilution in PBS containing 20 mM EGTA. After a brief period of contact (15 min), the samples were excited at 485 nm and the fluorescence emission intensity measured at 530 nm.

OPEN IN VIEWER

Figure 1

Fluorescence intensity of NG as a function of Zn²⁺ concentration. •, Zn²⁺ was in the form of ZnCl₂ in 10 mM Tris-HCl; ▪, Zn²⁺ was in the form of Zn-insulin solution (Actrapid; Novo Nordisk; 40 IU/ml corresponding to 100 μM Zn²⁺); —, incubation with NG potassium salt solution (NG-K, 6 μM); —, incubation with NG esterified form in PBS (NG-Ac, 25 μM). Values are the means ± SEM of 16 experiments (ZnCl₂) or four experiments (Actrapid). Excitation wavelength 485 nm; emission wavelength 530 nm.

Immunocytochemical Analysis

Cells were cultured on glass coverslips, washed with PBS to remove serum, and processed at 4C as described by Brelje et al. (1989). Briefly, they were fixed for 18 hr in Zamboni's fixative (Stefanini et al. 1967) and, after extensive washings in Sorenson's PBS containing 0.1% Triton X-100, they were incubated with 10% normal goat serum in PBS containing 0.1% Triton X-100 (PBS-T) for 4 hr. After washings in PBS-T, cells were incubated for 18 hr with a 1:300 dilution of the primary antibody (monoclonal antibody to insulin/proinsulin; Biogenesis, Newfields, UK). The incubation was stopped by three washings in PBS-T, followed by a second one for 18 hr with either a 1:100 dilution of a fluorescein isothiocyanate-conjugated anti-mouse immunoglobulin (Dako; Trappes, France) or a 1:500 dilution of anti-mouse immunoglobulin revealed with phtaloRed from Kirkegaard & Perry (Gaithersburg, MD). Controls included replacing the primary antobody with PBS containing 1% BSA. Nuclei were counterstained with Carazzi's hematoxylin.

OPEN IN VIEWER

Figure 2

Specificity of NG fluorescence for zinc. Zn²⁺ (6 μM) was incubated with 2.5 μM NG-K (control). The following competitors were added: Mg²⁺ and Ca²⁺ at the final concentration of 1 mM, Fe²⁺, Ba²⁺, Cu²⁺, Cd²⁺, Pb²⁺, Hg²⁺, Co²⁺, Ni²⁺ at the final concentration of 100 μM; and EGTA and EDTA at the final concentration of 1 mM. Values are expressed as a percentage of control values and are the means ± SEM of four experiments.

OPEN IN VIEWER

Figure 3

Correlation between NG fluorescence and number of islets. “Handpicked” islets were incubated in 1 ml NG-Ac (25 μM) after distension of cells for 5 min as described in Materials and Methods. The regression equation was y = 272 × + 15, 203; r = 0.989; p < 0.0001.

OPEN IN VIEWER

Figure 4

NG cytotoxicity. INS-1 cells were incubated with or without 25 μM NG-Ac for 30 min, then cultured for several days and counted as described in Materials and Methods. Values are means ± SEM of four experiments. Filled columns, control cells; open columns, NG-Ac-treated cells.

OPEN IN VIEWER

Figure 5

Confocal microscopic examination of stained islets. (A) Projection of 60 optical sections acquired at 1-μm intervals through a human islet stained with NG-Ac. Bar = 50 μm. (B) Projection of 60 optical sections acquired at 2-μm intervals through a human islet stained with NG-Ac (green living cells) and propidium iodide (red dead cells) after mild distension of cells as described in Materials and Methods. Bar = 50 μm. (C) A section through a particularly big and compact islet, showing defective penetration of the probe. Bar = 50 μm. (D) A section through a standard islet stained with NG-Ac; note the complete penetration of the probe in the core of the islet. Bar = 30 μm. (E) Individual islet cells stained with NG-Ac and settled on a glass coverslip coated with poly-l-lysine. Note the heterogeneous fluorescence of NG-labeled cells with large unstained nuclei. Bar = 15 μm. (F) Individual islet cell monolayer immunostained for insulin. Bar = 40 μm.

OPEN IN VIEWER

Table 1

Effect of NG and DTZ staining on glucose-induced insulin secretion in human pancreatic islets^a

	Control	NG staining (SI ± SEM; n = 6)	DTZ staining
0 hr	2.61 ± 0.28	2.50 ± 0.31	1.90 ± 0.25
72 hr	1.33 ± 0.11	1.27 ± 0.13	1.17 ± 0.04

^aIslets (40 IE) were stained for 30 min by NG-Ac (25 μM) or DTZ (10 μg/ml). They were then immediately stimulated by glucose (3.6 and 16.7 mM) or stimulated after 72 hr of culture in CMRL 1066 medium. Stimulatory indices (SIs) were calculated as described in Materials and Methods.

Properties of Newport Green

Fluorescence of NG-K (6 μM) in 10 mM Tris-HCl was increased at the micromolar range of Zn²⁺ in the form of ZnCl₂ or crystal insulin (Actrapid) and was saturated at about 10 μM Zn²⁺. Half-maximal intensity was achieved at a concentration of 2–3 μM Zn²⁺. Binding studies using NG-Ac confirmed that the ester form was unable to bind zinc (Figure 1).

Zinc-dependent fluorescence of NG-K was unaffected by an excess of the metal ions frequently encountered in living cells, such as Ca²⁺, Mg²⁺, or Fe²⁺ added as salts. Among other metal ions analyzed, Cu²⁺, Cd²⁺, and Pb²⁺ quenched fluorescence, whereas others, such as Co²⁺ and Ni²⁺ enhanced Zn fluorescence. Nevertheless, they can be found only as trace elements in living cells. As expected, EDTA, EGTA, and an acidic pH affected Zn fluorescence of NG-K (Figure 2).

Control of Cell Loading

“Hand-picked” islets loaded with 25 μM NG-Ac demonstrated an excellent correlation between fluorescence intensity and islet number (r = 0.99; p < 0.0001) (Figure 3).

Fluorescence intensity was also increased in murine cells cultured in the presence of ZnCl₂ (25 μM) and pyrithione (4 μM), a zinc ionophore. The ratios of Zn:DNA in arbitrary units were 2.5- and 12-fold higher in treated INS-1 and in β-TC3, respectively, compared to the control cells.

OPEN IN VIEWER

Figure 6

FACS analysis. Overlaid fluorescence histograms of dissociated cells from islet cell preparations. Region A, histogram of autofluorescence; Region B, histogram of NG-Ac-stained cells.

Toxicity Study

INS-1 cells stained with NG-Ac (25 μM) for 30 min and then cultured for 7 days exhibited the same pattern of growth as untreated cells (Figure 4).

As shown in Table 1, human pancreatic islets stained by NG-Ac (25 μM, 30 min) demonstrated the same capacity for glucose-induced insulin secretion as control islets. In contrast, DTZ staining (30 min, 10 μg/ml) resulted in a moderate shift of stimulatory indices.

Cell Imaging

Islet cell staining was visualized by confocal microscopic studies. Cells positively stained by NG appeared green; dead cells stained by propidium iodide (4 μM) appeared red (Figures 5A and 5B). Analyses of individual optical sections of islets indicated a homogeneous staining from the periphery to the center in cases where cells were loosely bound inside the islet (Figure 5D). However, defective penetration of the probe was noticed when islets were particularly compact (Figure 5C). We therefore systematically performed a mild distension of islets for 5 min, as described above, before staining.

Staining of individual cells revealed a heterogeneous intensity of fluorescence with cytoplasmic dots (Figure 5E). Identical pictures were obtained using insulin staining by fluorescein-conjugated antibody to insulin/proinsulin. No nuclear staining was noticed (Figure 5F).

OPEN IN VIEWER

Figure 7

FACS analysis. (A) Histogram of side scattering (SS) activity of dissociated cells from islet cell preparations. Dissociated and NG-Ac-stained cells were distributed in two regions (I and II) according to their side scatter properties. (B) Fluorescence-SS dot-plot of dissociated cells from an islet cell preparation. Two window settings were established for separation of dissociated NG-Ac-stained cells: R2, high scatter and high fluorescence, corresponding to β-cells, and R1, low scatter and low fluorescence, corresponding to non-β-cells.

Quantitative Studies

FACS analyses of NG-stained cells from semipurified preparations (containing endocrine β- and non-β-cells contaminated with exocrine cells) demonstrated a large range (about ×300) of relative fluorescence units (RFUs). In comparison, the range of autofluorescence was more limited (about ×10) (Figure 6). The recovery of cells after sorting was always greater than 85%, and viability assessed by trypan blue exclusion was greater than 95%.

Two populations of cells were observed, according to their side scatter patterns (SS) and RFU, and were gated for sorting (Figure 7). The ratios of mean values (±SEM) in the two populations were 10.5 ± 1.8 (n = 5) for SS and 12.1 ± 1.14 (n = 5) for RFU. Insulin content of the sorted cells was analyzed. The ratios of insulin per DNA in the two sorted populations were 30 ± 9.4 (n = 5), proving a clear distinction between β- and non-β-cells. Immunocytochemical analysis of these two sorted cell populations also confirmed the presence of a great majority of insulin-positive cells (red staining) in the most fluorescent sorted population compared to the less fluorescent one (Figure 8). When the cell populations were compared before sorting, therefore containing endocrine plus exocrine cells, vs the most fluorescent sorted population, the enrichment in β-cells, estimated by the ratios of insulin per DNA, was highly variable (from 1.3 to 7; mean 3.2 ± 1.04; n = 5) because it depended on the purity of the starting preparation, i.e., the more pure the starting preparation was, the less the enrichment ratio was.

OPEN IN VIEWER

Figure 8

Immunocytochemical analysis of sorted cells. Sorted cells from regions R2 and R1 were plated on poly-l-lysine-treated Lab-Tek chambers with coverslips and cultured for 24 hr before immunostaining using anti-insulin antibody revealed with phtaloRed, as described in Materials and Methods. Nuclei were counterstained with Carazzi's hematoxylin. (A) Controls which included replacing the primary antobody with PBS containing 1% BSA. (B) Weakly fluorescent sorted cells. (C) Highly fluorescent sorted cells. Bars = 30 μm.

OPEN IN VIEWER

Table 2

Effect of sorting on β-cell functionality^a

	Insulin secreted (ng/105 cells) for 1 hr at 16.7 mM glucose (mean ± SEM; n = 3)	SI
Control cells	41.4 ± 0.53	1.23 ± 0.16
Highly fluorescent fraction	40.1 ± 1.58	1.51 ± 0.12
Weakly fluorescent fraction	0.44 ± 0.02	0.98 ± 0.31

^aAfter dissociation from semipurified islet preparations, cells were stained by NG-Ac 25 μM for 30 min. A portion of cells was sorted by FACS as described in Materials and Methods. Control and sorted cells were then plated in 48-well plates (10⁵ cells/well) and cultured for 72 hr before stimulation experiments. Control cells, dissociated and stained unsorted cells.

In terms of function, the dissociated cells plated before and after sorting and cultured for 72 hr exhibited the same pattern of glucose-induced insulin secretion. However, the most fluorescent sorted cells secreted about 10-fold more insulin than the weakest fluorescent cells (Table 2).

The determination of zinc content per DNA unit was investigated in islet preparations from different donors to determine if an index designed to estimate the purity plus viability of these preparations could be established. Staining of cells was performed in six replicates of 40 IEs stained with 25 μM NG-Ac after mild distension of islets. As shown in Figure 9, we observed a wide variability of values among the different preparations. Regression analyses demonstrated fair correlations between Zn:DNA values and purity, qualitatively estimated by the percentage of DTZ-stained cells in the preparation (r = 0.63; p = 0.011, n = 15) and between Zn:DNA values and insulin content (r = 0.81; p = 0.007; n = 9).