IRDye78 Conjugates for Near-Infrared Fluorescence Imaging

Abstract

The detection of human malignancies by near-infrared (NIR) fluorescence will require the conjugation of cancer-specific ligands to NIR fluorophores that have optimal photoproperties and pharmacokinetics. IRDye78, a tetra-sulfonated heptamethine indocyanine NIR fluorophore, meets most of the criteria for an in vivo imaging agent, and is available as an N-hydroxysuccinimide ester for conjugation to low-molecular-weight ligands. However, IRDye78 has a high charge-to-mass ratio, complicating purification of conjugates. It also has a potentially labile linkage between fluorophore and ligand. We have developed an ion-pairing purification strategy for IRDye78 that can be performed with a standard C18 column under neutral conditions, thus preserving the stability of fluorophore, ligand, and conjugate. By employing parallel evaporative light scatter and absorbance detectors, all reactants and products are identified, and conjugate purity is maximized. We describe reversible and irreversible conversions of IRDye78 that can occur during sample purification, and describe methods for preserving conjugate stability. Using seven ligands, spanning several classes of small molecules and peptides (neutral, charged, and/or hydrophobic), we illustrate the robustness of these methods, and confirm that IRDye78 conjugates so purified retain bioactivity and permit NIR fluorescence imaging of specific targets.

Introduction

The detection of cancer cells and their products in vivo is dependent on both the signal generated by the contrast agent and the inherent background or “noise” of the tissue being imaged. The technique of near-infrared (NIR) fluorescence imaging provides extremely low background since living tissue has minimal absorption and autofluorescence in the NIR wavelength range of 700–900 nm, especially when compared to visible (400–700 nm) light [1]. Introducing an exogenous NIR fluorophore, which has a relatively high extinction coefficient and quantum yield, can produce a signal adequate for imaging. By conjugating a cancer-specific ligand to such an NIR fluorophore, a novel contrast agent is generated that has the potential for highly specific and sensitive detection of human malignancy.

Over the last several years, NIR fluorophores with optimal photoproperties and pharmacokinetics have been developed [2–5]. Of the heptamethine NIR fluorophores, the indolyl (indo) cyanines show particular promise, and in certain micro-environments, are less likely to self-aggregate than heptamethines such as the benzothiazoles [6]. Indocyanines also have a long history of clinical use. In 1956, indocyanine green (ICG; IC-GREEN^™) was FDA-approved for use in indicator-dilution studies in humans and has had a remarkably good safety profile.

Of the many indocyanines described, we have focused on IRDye78 (LI-COR, Lincoln, NE). IRDye78 is available as an N-hydroxysuccinimide ester (NHS) for conjugation to ligands having one or more primary amines. Four sulfonate groups impart an aqueous solubility of greater than 10 mM to IRDye78, and make the fluorophore itself highly solubilizing when conjugated to lipophilic ligands. Poly-sulfonation also greatly increases plasma half-life compared with nonsulfonated precursors [5]. In the mouse, we have measured early and late phase serum half-lives for IRDye78 of 7.2 and 24.7 min, respectively [7].

IRDye78 also has certain photoproperties that are ideal for in vivo imaging. Its peak absorption (771 nm) and emission (806 nm) are located in an area of the NIR window [8] that maximizes tissue penetration and subsequent recovery of fluorescent photons. IRDye78 has a relatively high extinction coefficient (°_771nm = 145,650) and quantum yield (14.2%) in 10 mM HEPES, pH 7.4, and a photobleaching threshold that permits continuous excitation at 50 mW/cm² without loss of fluorescent signal (A. Nakayama and JVF, manuscript in preparation). Taken together, these properties make IRDye78 an excellent choice as an NIR fluorophore for in vivo imaging.

Although IRDye78 can be conjugated to small molecule and peptide ligands, there is a pressing need to purify such conjugates conveniently and reliably. Partly due to poly-sulfonation, purification of IRDye78 conjugates can be difficult, especially when using low-molecular-weight ligands with their own net charge. Since IRDye78 is unstable under strongly acidic and basic conditions [7], conjugation and purification should be performed at near-neutral pH. Finally, if care is not taken during the purification process, IRDye78 conjugates will undergo chemical conversion. At least one of these conversions (described below) is catastrophic, resulting in loss of the ligand but retention of NIR absorption and fluorescence.

In anticipation of using IRDye78 and similar NIR fluorophores as contrast agents for the in vivo detection of human cancer, we have developed a simple and robust ion-pairing HPLC strategy for purifying conjugates under neutral conditions, with preservation of conjugate stability. Ion pairing is a chromatographic method suitable for the purification of highly charged molecules, such as DNA, which would otherwise not have retention on a C18 column. Indeed, ion pairing has even been used previously to purify DNA conjugated to di-sulfonated indocyanines [9]. In this article, we attempt to teach the reader how to start simply with a tumor-targeting ligand, and to end with a contrast agent ready for in vivo characterization. We support the general applicability of our methods by purifying seven different IRDye78/low-molecular-weight ligand conjugates, and discuss potential problems that can arise during the process.

Materials and Methods

Reagents

The NHS ester of IRDye78 (IRDye78-NHS) was purchased from LI-COR and stored in the dark, under nitrogen, at −80°C until use. Immediately before conjugation, it was resuspended at 2 mM in dimethylsulfoxide (DMSO). The carboxylic acid of IRDye78 (IRDye78-CA) was a generous gift from LI-COR. β-l-Asp-l-Glu was synthesized by Genemed Synthesis (South San Francisco, CA). Pamidronate disodium was a generous gift from Interchem (Paramus, NJ). The synthesis of peptide KBP 1.66 has been described previously [10]. HPLC grade water was purchased from Alfa Aesar (Ward Hill, MA). HPLC grade methanol (MeOH) was purchased from Mallinckrodt Baker (Phillipsburg, NJ). HPLC grade triethylammonium acetate (TEAA), pH 7.0 was purchased from Glen Research (Sterling, VA). Hydroxyapatite (HA, #391947) was purchased from Calbiochem (La Jolla, CA). All other chemicals were purchased from Sigma (St. Louis, MO).

Conjugation of IRDye78-NHS to Low-Molecular-Weight Ligands

All steps were performed under reduced light conditions. The conjugation of l-glutamate, β-l-Asp–l-Glu, KBP 1.66, and l-tryptophan was performed in DMSO. Each 50 μL reaction contained 20 mM triethylamine (TEA), 1 mM ligand, and 1 mM IRDye78-NHS, added in this order. The conjugation of ethanolamine and l-arginine was performed in 20 mM sodium bicarbonate, pH 8.5. Each 50 μL reaction contained 5 mM ligand and 1 mM IRDye78-NHS, added in this order. Conjugation was performed at room temperature, in the dark, and with constant agitation for 18 hr. The conjugation of pamidronate has been previously described in detail [7].

An HPLC System for the Purification of IRDye78 Conjugates

The HPLC system was purchased from Waters (Milford, MA) and consisted of a vacuum degasser (#WAT079700), model 7725i manual injector (#I86000872), 2 mL stainless steel sample loop (#700000563), model 1525 binary HPLC pump (#I86001525), model 2487 dual wavelength absorbance detector (#WAT081110), and Fraction Collector II (#725000126). The absorbance detector was set to its highest possible wavelength (700 nm), although peak absorption from IRDye78 occurs at 771 nm [7]. Separation was performed on a 4.6 by 150 mm Symmetry C18 column (#WAT045905). The column eluate was split into two using a model P-451 flow splitter (Upchurch Scientific, Oak Harbor, WA). A portion of the eluate flowed into a Sedex model 75 evaporative light scatter detector (ELSD, Richards Scientific, Novato, CA), with the nebulizer modified to reduce band broadening at low flow rates. The ELSD was set to 40°C, with a nitrogen pressure of 3.5 bar and a gain of 6. Control of system components and data acquisition was performed with a Compaq computer using the Breeze (Waters) software package.

Purification of IRDye78/Low-Molecular-Weight Conjugates

Buffer A consisted of 10 mM TEAA, pH 7.0. Buffer B was absolute MeOH. To avoid ELSD artifacts, HPLC grade solvents were not refiltered. Each conjugation reaction was diluted at least 20-fold in Buffer A prior to injection into the sample loop. The mobile phase gradient for all separations presented in this study was 85% A/15% B to 20% A/80% B, over 30 min, at a flow rate of 1 mL/min. Since collected peaks were usually 2 mL, they were concentrated on an Oasis HLB desalting cartridge (Waters, #186000383) by diluting to a total volume of 15 mL in Buffer A, applying to cartridge, washing 3 by 1 mL with H₂O, and eluting with 1 mL of MeOH. Depending on the experiment, final material was dried to powder either by lyophilization under high vacuum, but without prefreezing or heating, or by slow evaporation under a stream of nitrogen gas. The latter method of evaporation is highly preferred (discussed below). Compounds were stored as dry powders, in a desiccator, at −80°C, and in the dark until use. Product yields were calculated by area-under-the-curve analysis on the above HPLC Breeze system using IRDye78-CA standards, and were confirmed off-line using IRDye78′s °_771nm of 145,650 M⁻¹ cm⁻¹. Products were also subjected to matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) mass spectroscopic analysis. Mass spectroscopy was performed on a Voyager-DE STR (Applied Biosystems, Framingham, MA) in reflector mode with positive polarity and an accelerating voltage of 20,000 V. Sample (100 pmol) was diluted 1:10,000 in the matrix (10 mg/mL α-cyano-4-hydroxy-trans-cinnamic acid in 70% acetonitrile/30% H₂O) and was spotted onto the sample plate. A standard mixture of angiotensin-I and ACTH was spotted in a nearby location to calibrate the instrument and to act as a control.

Spectral Measurements

Absorbance was measured on a Model DU-600 spectrophotometer (Beckman, Fullerton, CA) using 1 μM of each compound in 10 mM HEPES, pH 7.4. Fluorescence was measured on a SPEX Fluoromax-1 (Jobin Yvon, Edison, NJ) spectrofluorometer using 500 nM of each compound in 10 mM HEPES, pH 7.4. Emission scans were performed with a 7-nm slit width.

HA Binding Experiments

Binding experiments were performed in 20 mM N-ethylmorpholine, pH 7.4, 150 mM KCl (NK buffer) supplemented with 5 mg/mL HA. Each molecule to be tested was diluted to a concentration of 1 μM in a 200-μL reaction. After continuous vortexing for 30 min, HA crystals were washed three times with NK buffer, transferred to a 96-well plate, centrifuged, and visualized using a previously described small animal NIR fluorescence imaging system [7], at a fluence rate of 5 mW/cm².

KIX Domain Binding Experiments

Purification of glutathione S-transferase (GST) and the fusion of GST with the KIX domain of p300/CBP were performed as described previously [11]. Both proteins were bound to glutathione agarose at a final concentration of 1.6 mg/mL of packed beads. Each 65 μL binding reaction contained 25 μL of packed beads and 11.4 μM IRDye78 conjugate in phosphate-buffered saline, pH 7.4 (PBS). After a 20-min incubation with constant agitation, beads were washed three times quickly in PBS and cytospun onto glass slides. Cytospun beads were resuspended in 20 μL of H₂O and photographed 30 sec later using NIR fluorescence or bright field microscopy as described previously [7].

Results

Covalent Conjugation of IRDye78-NHS to Low-Molecular-Weight Ligands

The NHS ester of IRDye78 (IRDye78-NHS) was conjugated to low-molecular-weight ligands containing a single primary amine as described in Materials and Methods, and as shown in Figure 1A. Whenever possible, conjugation was performed in nonaqueous solvent, in the presence of an excess of TEA. The solvent for ligand conjugation must be chosen carefully to avoid self-aggregation of IRDye78-NHS and/or the ligand of interest. In our experience, IRDye78-NHS is soluble in DMSO, a preferred solvent, to at least 10 mM. Solubility in other solvents is variable. Nucleophilic attack by the primary amine of the ligand results in displacement of the NHS group and formation of a stable amide linkage between IRDye78 and the low-molecular-weight ligand. When the conjugation reaction was performed in aqueous buffer (e.g., for ligands ethanolamine, l-arginine and pamidronate), nucleophilic attack by the hydroxyl ion of H₂O can also occur, resulting in cleavage of the NHS ester bond and formation of the carboxylic acid form of IRDye78 (IRDye78-CA; not shown). For aqueous conjugation, vigorous vortexing should be used throughout to maximize fluorophore/ligand collisions. After ion-pairing HPLC purification in a molar excess of TEAA, pH 7.0, sodium and other salts have the potential to be converted to TEA salts (Figure 1A).

The low-molecular-weight ligands conjugated to IRDye78 in this study are shown in Figure 1B. They were purposely chosen to include commonly encountered classes of molecules. Ethanolamine is one of the simplest possible ligands, having only two carbons, an alcohol, and no net charge after conjugation. l-Arginine is a zwitterion after conjugation, having a cationic guanidinium group and an anionic carboxylic acid group. l-Arginine was also chosen to test the hypothesis that cationic groups, which are themselves poor nucleophiles, do not interfere with the NHS ester/primary amine reaction. The ligands l-glutamate, β-l-Asp–l-Glu (a β-linkage of l-aspartate and l-glutamate), and pamidronate form a series of compounds with increasing anionic character. After conjugation, these ligands impart an additional net charge of −2, −3, and −4, respectively, to IRDye78. KBP 1.66 is a previously described 11 amino acid peptide that has a low but measurable affinity for the KIX domain of the coactivator proteins p300 and CBP [10]. This peptide is extremely hydrophobic and has a net charge of −1 after conjugation. Finally, l-tryptophan is a small aromatic molecule with a net charge of −1 after conjugation. These molecules span a molecular weight range from 61.1 to 1148.3 Da, and include many possible combinations of charge and hydrophobicity.

Ion-Pairing HPLC Purification Strategy for IRDye78 Conjugates

A schematic of the HPLC purification strategy used for IRDye78 conjugates is shown in Figure 2 and is described in detail in Materials and Methods. Buffer A was composed of the ion-pairing agent 10 mM TEAA, pH 7.0, which was degassed before use. Buffer B was absolute MeOH, also degassed. IRDye78-CA was found to be soluble to at least 10 mM in both Buffers A and B, thus permitting preparative purification without loss of resolution due to self-aggregation. However, some IRDye78/ligand conjugates may be more prone to self-aggregation, which will limit the amount of material than can be concentrated with any single purification.

Figure 1.

Conjugation and purification of low-molecular-weight ligands to the NHS ester of IRDye78. (A) Low-molecular-weight ligands (R) having a single primary amine (II) were conjugated to the sodium salt of the NHS ester of IRDye78 (I) as described in the Materials and Methods, yielding free NHS (III) and the IRDye78/low-molecular-weight ligand conjugate (IV). Carbons 2 and 4′, to be discussed herein, are indicated. R′ is ligand R without its primary amine (NH₂), and is shown as an amide linkage to IRDye78. After treatment with 10 mM TEAA, pH 7.4 during the HPLC purification process, sodium and other salts may be converted to TEA salts. For the final product (V), the exact number of associated TEA molecules (n) will depend on the chemical composition of R′. The molecular weight (M.W.) of each compound is shown. (B) The ligands conjugated to IRDye78 in this study are grouped by class, with the number and type of charges on each R′ group postconjugation shown in parentheses (i.e., the two negative charges of l-glutamate are shown as —). The M.W. of each ligand prior to conjugation is given. The chemical structure of each ligand is shown in its ionic state at the time of conjugation. For the hydrophobic peptide KBP 1.66, the amino acid sequence, in one letter code, is oriented from amino terminus (NH₂) to carboxyl terminus (COO⁻). For each ligand R, conjugation is through its single primary amine.

After separation of reaction components on a standard C18 column, a portion of the eluate was directed to an ELSD. An ELSD is a detector whose signal is proportional to the concentration of any nonvolatile molecule present, regardless of the molecule's polarity, UV absorbance, or visible absorbance. It therefore permits detection of all reactants, products, and byproducts that could be formed in the desired coupling reaction. The remainder of the eluate flowed through a dual-wavelength absorbance detector, where the presence of any NIR fluorophore was detected. After collection of the desired peak(s) on a fraction collector, samples were concentrated using an Oasis HLB cartridge and dried as described in Materials and Methods (see also below).

Figure 2.

An ion-pairing HPLC purification strategy for IRDye78 conjugates. After separation of reaction components on a C18 column, a fraction of the eluate was diverted to an ELSD for identification of nonoptically active species. The remainder of the eluate flowed to an absorbance detector and fraction collector. All system components are under computer control (not shown; described in Materials and Methods).* The stability of IRDye78 and its conjugates is critically dependent on the method used for sample concentration (discussed in detail herein).

Detection of Reactants and Products and Mass Spectroscopic Analysis

In order to maximize IRDye78 conjugate purity, we determined the elution time of every reactant and product using the ELSD and absorbance detectors. For demonstration purposes, all ligands and IRDye78 conjugates are shown in Figure 3 on the same mobile phase gradient. First, the ELSD detector was used to determine that NHS, the major metabolite of the conjugation reaction, was not retained on the column, and was always present in the void volume (data not shown). In Figure 3 (left panels), we show the retention time for each unconjugated ligand, as detected by the ELSD. Most ligands were not retained on the column and appeared in the void volume, however, the peptide KBP 1.66 and the amino acid l-tryptophan had high retention, eluting at 27.58 and 5.22 min, respectively. Using the ELSD, we were able to ensure that these retained ligands were well separated from the desired IRDye78 conjugate (see below).

The top two panels of Figure 3 show the retention times of the reactive compound IRDye78-NHS, and the parent compound IRDye78-CA. Using the above gradient, these retention times were found to be 15.73 and 13.30 min, respectively. It should be noted that the small peak at 13.55 min on the 700-nm absorbance tracing for IRDye78-NHS (Figure 3, top right panel) is at present unidentified. It is unlikely to be IRDye78-CA, as might be expected with the reaction of IRDye78-NHS with the hydroxyl ion of H₂O, since it is volatile (compare OD_700nm tracing to ELSD tracing) and has a slightly different retention time than IRDye78-CA. There were also minor impurities in IRDye78-NHS seen with the ELSD eluting at 9.93 and 26.52 min (Figure 3, top panels).

After determining the retention times for each ligand (using the ELSD), NHS (using the ELSD), IRDye78-NHS, and IRDye78-CA, we were able to purify each of the desired IRDye78/ligand conjugates. In Figure 3 (column labeled OD_700nm conjugate) are shown the chromatograms of each such purified compound. By convention, the IRDye78 conjugate of each ligand will be referred to as the ligand's name with suffix “−78”. The measured retention times for ethanolamine-78, l-arginine-78, l-glutamate-78, β-l-Asp–l-Glu-78, pamidronate-78 (previously referred to as Pam78 in Ref. [7]), KBP 1.66-78, and l-tryptophan-78 were 14.30, 14.25, 11.93, 11.12, 13.10, 21.43, and 14.72 min, respectively. Although the chromatograms presented in Figure 3 used the same mobile phase gradient, it is highly desirable to optimize the gradient for each IRDye78 conjugate so as to maximize any differences in retention time between reactants and products. For example, by changing the mobile phase gradient from (85% A/15% B to 20% A/80% B) to (70% A/30% B to 35% A/65% B), the separation between IRDye78-CA and IRDye78-NHS is changed from 2.43 to 4.09 min.

After concentration (discussed below), each sample was subjected to MALDI-TOF mass spectroscopic analysis as described in Materials and Methods. The results are shown in Figure 3 (right column). Not unexpectedly, the TEA salts of these compounds were not visualized by MALDI-TOF, and only the protonated neutral compounds were seen. For each compound, the measured mass was the same as the expected mass, within the calibration error of the instrument (Figure 3).

Reversible Conversions of IRDye78 and its Conjugates

In early experiments, peaks purified using the fraction collector were concentrated on an Oasis HLB cartridge as described in Materials and Methods, eluted with absolute MeOH, and lyophilized (without freezing) under high vacuum, and without heating. The powder so formed had a dark blue tone, and when resuspended in either aqueous buffer, or aprotic solvents such as DMSO, immediately turned a light blue color, most like the gem colors aquamarine or blue topaz (Figure 4A). By testing each component, systematically, it was determined that this conversion from green to light blue could be reproduced in the absence of any purification or concentration simply by lyophilizing IRDye78-CA in increasing concentrations of TEAA, pH 7.0. The higher the molar ratio of TEAA to IRDye78-CA present during lyophilization, the more pronounced the blue color. If this light blue material is diluted into H₂O, and especially if heated, it turns back to green, although the shade of green has slightly more yellow tone (Figure 4A).

Figure 3.

Ion-pairing HPLC purification of IRDye78 conjugates and detection of optically inactive/active species. Prior to conjugation, the ELSD was used to identify the elution time for each ligand R (left chromatogram). After conjugation, and in parallel with the ELSD, the absorbance detector was used to detect conjugated and unconjugated IRDye78 and to trigger fraction collection. For reference, the chromatograms for IRDye78-NHS and IRDye78-CA are shown at the top of the figure. The final chromatogram (OD_700nm conjugate) of the IRDye78/ligand R conjugate after peak purification is shown, along with the expected and found molecular weight (M.W.) of salt-free conjugates. The mobile phase gradient for all separations shown was 85% A/15% B to 20% A/80% B, over 30 min, at a flow rate of 1 mL/min. Buffer A consisted of 10 mM TEAA, pH 7.0. Buffer B was absolute MeOH.

These observations are supported by the absorbance wavelength scans shown in Figure 4B. All three compounds have characteristic major and minor peaks at 771 nm and 708 nm, respectively [7]. However, the “blue” compound has a new broad absorbance from 550 to 650 nm, which likely contributes to its bluish appearance (Figure 4B; *a). The compound converted from blue back to green has a new absorbance between 425 and 490 nm, which likely contributes to its slightly yellowish appearance (Figure 4B; * b), but does not have the shoulder from 550 to 650 nm.

Figure 4.

Reversible chemical conversion of IRDye78 and its conjugates. (A) The sodium salt of IRDye78-CA is light green in color (1), however, after lyophilization in the presence of TEAA, pH 7.0 and resuspension in H₂O, the fluorophore is light blue (2). Prolonged incubation in water, especially after dilution or with mild heating, converts the fluorophore back to a light green color (3). All compounds were diluted in 10 mM HEPES, pH 7.4 to a final concentration of 50 mM prior to being photographed. (B) Absorbance wavelength scanning of the three compounds shown in A was performed as described in Materials and Methods. Scans are normalized to the peak NIR absorbance to highlight changes in the spectrum. A broad absorbance from 550 to 650 nm (*a) is seen with the blue compound (2) but not with the parent compound (1) or the compound converted from blue to green (3). A broad absorbance from 425 to 490 nm (*b) is also seen with compound (3) but not with the others. (C) Fluorescence emission scanning of the three compounds in A was performed using an excitation wavelength of 575 nm (left graph) or 771 nm (right graph). Emission curves are normalized to the peak NIR emission to accentuate the hypsochromic shift from 806 nm (Compounds 1 and 3) to approximately 799 nm (Compound 2) and the significant broadening of the NIR peak for Compound 2 that is seen with 575 nm excitation.

The fluorescence emission from these three compounds was also different. When excited at 575 nm, the sodium salt of IRDye78-CA and the compound converted from blue to green had the characteristic peak NIR emission at 806 nm, however, the blue compound had a hypsochromic shift of this peak to 799 nm, and significant broadening of the emission curve (Figure 4C, left panel). It should also be noted that the compound converted from blue to green had loss of the minor emission peak at 663 nm. When excited at 771 nm, all three compounds had superimposable emission curves, centered at 806 nm (Figure 4C, right panel).

Remarkably, however, when these three compounds were analyzed by MALDI-TOF mass spectroscopy, the major peaks had the identical mass (1014.2 Da expected and found), corresponding to IRDye78-CA. The blue compound and the compound converted from blue to green had a minor mass peak at 1088.2 Da, which is presently unidentified. The significance of these results is discussed below.

Irreversible (Catastrophic) Conversion of IRDye78 and its Conjugates

During preparation of KBP 1.66-78, the conjugate of IRDye78-CA and the hydrophobic peptide KBP 1.66, it was noticed that lyophilization in the presence of TEAA resulted in a green-to-blue transition, but unlike the case given above, it was not always possible to convert the compound back to green with heating or dilution in H₂O (data not shown). MALDI-TOF mass spectroscopy of the blue material revealed a major peak at 1297.5 Da, and a secondary peak at 867.1 Da. One of many possible mechanisms generating such products would be a nucleophilic attack at carbon 4⁰ (Figure 1A), which would release the entire phenoxy/propionic acid/peptide group. Although the 867.1 peak is consistent with the remaining heptamethine indocyanine backbone, the mass does not match either amination or oxygenation of carbon 4′. The precise identification of the 867.1 Da compound will be the subject of future investigation.

By systematically analyzing each step of the purification and concentration process, it was determined that loss of ligand (i.e., loss of the phenoxy/propionic acid/ligand) was most pronounced with lyophilization in the presence of TEAA, and could be minimized or eliminated by slow evaporation under N₂ gas. Concentration on the Oasis HLB cartridge had no effect on fluorophore, ligand, or conjugate as analyzed by MALDI-TOF mass spectroscopy (data not shown).

NIR Fluorescence Imaging using IRDye78 Conjugates

The purification and concentration strategy presented above is valuable only if the final IRDye78 conjugates retain binding to the target normally bound by the unconjugated ligand. Two of the seven ligands shown in Figures 1B and 3 have known targets. Pamidronate is a bisphosphonate with high-affinity binding to HA crystals. We have shown previously that an IRDye78 conjugate of pamidronate purified using thin-layer chromatography (TLC) retains HA binding, and can even be used in vivo as an imaging reagent for osteoblastic activity [7]. In Figure 5A, we show the binding of all IRDye78 molecules presented in this article to HA crystals. Only pamidronate-78 had detectable binding to HA, and this compound permitted NIR fluorescence imaging of the target.

KBP 1.66 is a phage display-derived peptide with low-affinity (55.1 mM) binding to the KIX domain of the coactivators p300/CBP [10]. We conjugated and purified an IRDye78 conjugate of KBP 1.66 using the above strategy. As shown in Figure 5B, KBP 1.66-78 binds to the KIX domain of p300/CBP expressed as a GST fusion protein, but not to GST itself. Importantly, the conjugate of l-tryptophan with IRDye78 (tryptophan-78) does not bind to either protein, despite being hydrophobic and carrying the same net charge as KBP 1.66-78. Taken together, these data suggest that the purification and concentration strategy presented in this study produces IRDye78 conjugates that retain ligand binding, and which can be used to perform NIR fluorescence imaging of desired targets.

Figure 5.

NIR fluorescence imaging with HPLC-purified IRDye78 conjugates. (A) IRDye78 conjugates (1 μM) were incubated with HA (5 mg/mL) in NK buffer as described in Materials and Methods. After incubation and washing, a thin layer of HA crystals were pelleted to the bottom of a 96-well plate and visualized by bright field (left panel) and NIR fluorescence (right panel) as described in Materials and Methods. A scale bar is shown for reference. Wells and compounds were: A1 (NK buffer only), B1 (IRDye78-CA), C1 (ethanolamine-78), A2 (l-arginine-78), B2 (l-glutamate-78), C2 (β-l-Asp–l-Glu-78), A3 (pamidronate-78), B3 (KPB 1.66-78), and C3 (l-tryptophan-78). (B) GST alone or the fusion of GST with the KIX domain of p300/CBP (GST/KIX) were bound to glutathione agarose beads and were incubated with either KBP 1.66-78 or trypotophan-78 as described in Materials and Methods. After washing, beads were visualized by bright field or NIR fluorescence microscopy as indicated. A scale bar is shown for reference. All NIR fluorescence photomicrographs have equivalent exposure time and normalization.

Discussion

For NIR fluorescence imaging of cancer cells to become a clinical reality, convenient, reliable, and generally applicable methods are needed for producing stable fluorophore/ligand conjugates in high yield and with high purity. This study presents such methodology for IRDye78, a tetra-sulfonated heptamethine indocyanine with excellent pharmacokinetics and photoproperties. The seven ligands conjugated to IRDye78 in this study span the classes of molecules that one might use routinely: uncharged but polar (ethanolamine), zwitterionic (l-arginine), anionic (l-glutamate, β-l-Asp–l-Glu, pamidronate), hydrophobic (KBP 1.66, l-tryptophan), and peptide (KBP 1.66). Despite vast differences in chemical character, each IRDye78/ligand conjugate was purified to homogeneity, shown to have the expected molecular weight, and shown (in the case of pamidronate and KBP 1.66) to retain ligand/target binding. It should be noted that conjugation of molecules such as pamidronate to IRDye78 yield nine charges (eight negatives, one positive) on a final mass of only 1234 Da. The satisfactory performance of the above ion-pairing purification strategy with such extreme charge-to-mass ratios suggests that it should also perform well for most low-molecular-weight ligands.

The importance of complementing conventional absorbance monitoring with an ELSD in this purification strategy cannot be overstated. The ELSD detects all nonvolatile reactants, products, and by-products of the conjugation reaction, and is especially useful because the elution time of molecules on a C18 column in the presence of ion-pairing agents such as TEAA is often nonintuitive. This fact is highlighted by the data shown in Figure 3. The hydrophobic molecule KBP 1.66 elutes at a later time than its IRDye78 conjugate, whereas the hydrophobic molecule l-tryptophan elutes at a much earlier time than its corresponding conjugate. Since many potential ligands have no absorbance, the ELSD is necessary to monitor the separation of unconjugated from conjugated ligand. Such separation is essential for future clinical applications, which will require the highest possible purity of injected materials. It should be emphasized, however, that the ELSD is only compatible with volatile solvent systems. It cannot be used in the presence of nonvolatile salts, like those frequently used with ion exchange chromatography, since the salt itself will be registered as a detectable mass. It will also underestimate mass if the mobile phase itself renders a molecule volatile during the separation.

Five caveats must be considered when working with IRDye78-NHS and the above purification strategy. The first is that ligands (small molecules and peptides) with more than one primary amine must be used with caution. The proper ratio of fluorophore to ligand in the conjugation reaction must be determined empirically such that only one fluorophore per ligand is conjugated. Even if this is accomplished, it will likely be extremely difficult to purify to homogeneity molecules with IRDye78 conjugated to a particular primary amine. The final product will be a mixture of compounds, each having a single IRDye78 conjugated at a different position. One way around this problem, of course, is to protect all primary amines not directly involved in target binding, or to change primary amines to other chemical groups that will not interfere with the NHS ester reaction. For peptides, a good choice for this type of substitution is l-arginine, which, as we demonstrate, does not interfere with IRDye78 conjugations (Figure 3).

A second caveat involves ligands that contain strong nucleophiles that are not amines. Thiol groups and imidazole groups are capable of attacking the NHS ester of IRDye78-NHS as well, if not better, than primary amines. Also, it should be understood that IRDye78 is created from a non-phenoxy-containing precursor by nucleophilic attack at carbon 4′ ([12]; see Figure 1A). Nucleophilic attack of indocyanines at other positions such as carbon 2 has also been described [13]. In this article, we show that under certain circumstances, nucleophiles can displace the entire phenoxy/ligand group resulting in decoupling of fluorophore and ligand. If the ligand itself contains a strong, non-amine nucleophile, it is possible that attack of the heptamethine backbone could occur during the conjugation reaction. If nucleophilic attack occurs at carbon 4′, there will be displacement of the phenoxy/propionic acid/NHS ester group from the molecule. Such a conjugate may still function as desired, but its photoproperties will likely differ from those presented in this study. The problems associated with strong nucleophiles are exemplified by our attempts to conjugate l-cysteine or l-histidine to IRDye78-NHS. In both cases, the conjugation reaction turned blue quickly, and ELSD/absorbance analysis revealed multiple fluorescent peaks (data not shown). MALDI-TOF mass spectroscopy confirmed that IRDye78 had been destroyed. However, we have been able to conjugate a peptide cyclized by oxidation of two internal cysteine residues without difficulty, and without evidence of aberrant products or loss of the phenoxy/ligand group (data not shown). From these data, we conclude that for conjugation of IRDye78-NHS to peptides, histidine should be substituted with a non-nucleophile, and cysteine should only be used if it is in an oxidized (i.e., disulfide) state. Peptides should also be of the highest possible purity, especially thiol-free, and possibly with trifluoroacetic acid exchanged with HCl.

A third caveat is that certain unusual conjugates may be more easily purified using TLC than HPLC. An example of such a conjugate is pamidronate-78. Although we were able to complete the purification using the above strategy with HPLC, we noticed that pamidronate-78 had a high propensity to adhere to metal surfaces, including HPLC tubing, despite its extremely high solubility in TEAA, pH 7.0. We suspect that this is either related to pamidronate-78′s marginal solubility in MeOH, or to the previously described binding of bisphosphonates to ferrous iron [14] and similar metals. In the case of pamidronate and molecules like it, we have previously published a TLC method for purification of IRDye78 conjugates [7], and we feel that this is the preferred method for this type of compound. Although simple and inexpensive, TLC is not the preferred method for most IRDye78 purifications, however, since it does not permit identification of products and reactants lacking chromophores, does not work well for purification of peptide conjugates, and is often difficult to scale upwards.

A fourth caveat is that certain charged molecules, such as peptides, will tend to elute with multiple peaks when using ion-pairing HPLC at neutral pH. This is usually due to different ionization states of the molecule. Provided that all peaks are separable from unconjugated ligand and free IRDye78, one can simply collect all conjugate peaks, perform mass spectroscopy, and pool the desired peaks prior to concentration. For molecules that overlap one or more other reactants and products, different chromatographic gradients can be tested, provided that IRDye78 is not subjected to destabilizing acidic or basic conditions.

The fifth, and final, caveat is that the method of final sample concentration and storage can have profound effects on the stability of IRDye 78 conjugates. In particular, slow evaporation under nitrogen is highly preferred over lyophilization. For example, the blue form of IRDye78 that occurs during lyophilization in the presence of TEAA, pH 7.0 has an emission curve distinct from the starting material, when 575 nm excitation light is used (Figure 4). Although this suggests that the blue compound is chemically distinct, it is sometimes possible to reconvert this material back to green simply with dilution into H₂O or with gentle heating. Indeed, MALDI-TOF mass spectroscopic analysis revealed the same mass for all three compounds, suggesting that this reconversion can occur during the MALDI-TOF procedure itself. In the original description of IRDye78-like molecules [4], it was suggested that the acetate ion could attack carbon 4′, and in some cases, displace an attached group. Such reactions resulted in major bathochromic and hypsochromic shifts [4]. Our data suggest that acetate, or even the conjugate bases of H₂O and alcohols, may be able to form a metastable intermediate at carbon 4′ [4], carbon 2 [13], or elsewhere, which causes a noticeable change in color, but which can sometimes be reversed. From the standpoint of working with IRDye78 and its conjugates, the most important point is that a change in color from green to blue can either reflect a harmless intermediate or a catastrophic loss of ligand. Careful structural and functional analysis of the final product will always be required to distinguish these possibilities.

Care should be taken with the storage and handling of IRDye78 conjugates. Our preferred method of storage for IRDye78 conjugates is as a dry powder, in a desiccator, at −80°C, in the dark. In light of a report that a similar heptamethine, indocyanine green, is unstable in water [15], we prefer an aprotic solvent such as DMSO when storing concentrated stock solutions in liquid form. However, we have stored pamidronate-78, which is insoluble in all nonaqueous buffers, for over a year at −80°C as a concentrated stock solution in H₂O without any evident change in photoproperties or loss of bioactivity in vivo [7].

For in vivo imaging applications, care must also be taken in the formulation used for intravenous injection. Specifically, it must be confirmed that conjugate self-aggregation does not occur at the concentration injected, and at 37°C, since this will result in a significant first-pass extraction from the lung and reticuloendothelial system. If micro-aggregation is present, it may be necessary to add solubilizing agents to the formulation. With respect to the stability of IRDye78/ligand conjugates after injection in vivo, results to date with pamidronate-78 [7] are promising. For instance, sites of bone turnover provide an inhospitable micro-environment, where local pH can be as low as 2.0. However, there does not appear to be any loss of NIR fluorescence or HA targeting of pamidronate-78, for periods up to 12 [7] or 72 hr (A. Zaheer and J. V. Frangioni, unpublished observation) after intravenous injection. This suggests that both the heptamethine backbone and the amide linkage to the ligand have the potential for prolonged stability in vivo under certain circumstances.

In summary, IRDye78-NHS reacts reliably with amine-containing low-molecular-weight ligands and has the potential for widespread use in cancer detection. This article has addressed the issue of rendering the purification of IRDye78 conjugates both reliable and convenient. Working with this fluorophore requires attention to detail with respect to the choice of ligand, the conjugation reaction, the purification strategy, and the formulation for in vivo use. We presented the conjugation and purification of seven chemically distinct ligands, and discussed the problems that were encountered during development of these methods. These results should decrease the development time for NIR fluorophore/ligand conjugates, and therefore, hasten translation of NIR fluorescence imaging to the clinic.