• Open Access

Comparative in vitro and experimental in vivo studies of the anti–epidermal growth factor receptor antibody nimotuzumab and its aglycosylated form produced in transgenic tobacco plants

Authors


Correspondence (Tel (53-7) 2504372; fax +(53-7) 2731779/2504494; email meilyn.rodriguez@cigb.edu.cu)

Summary

A broad variety of foreign genes can be expressed in transgenic plants, which offer the opportunity for large-scale production of pharmaceutical proteins, such as therapeutic antibodies. Nimotuzumab is a humanized anti–epidermal growth factor receptor (EGFR) recombinant IgG1 antibody approved in different countries for the treatment of head and neck squamous cell carcinoma, paediatric and adult glioma, and nasopharyngeal and oesophageal cancers. Because the antitumour mechanism of nimotuzumab is mainly attributed to its ability to interrupt the signal transduction cascade triggered by EGF/EGFR interaction, we have hypothesized that an aglycosylated form of this antibody, produced by mutating the N297 position in the IgG1 Fc region gene, would have similar biochemical and biological properties as the mammalian-cell-produced glycosylated counterpart. In this paper, we report the production and characterization of an aglycosylated form of nimotuzumab in transgenic tobacco plants. The comparison of the plantibody and nimotuzumab in terms of recognition of human EGFR, effect on tyrosine phosphorylation and proliferation in cells in response to EGF, competition with radiolabelled EGF for EGFR, affinity measurements of Fab fragments, pharmacokinetic and biodistribution behaviours in rats and antitumour effects in nude mice bearing human A431 tumours showed that both antibody forms have very similar in vitro and in vivo properties. Our results support the idea that the production of aglycosylated forms of some therapeutic antibodies in transgenic plants is a feasible approach when facing scaling strategies for anticancer immunoglobulins.

Introduction

Recombinant antibodies are without doubt the most successful therapeutic drug platform developed by modern biotechnology (PhRMA, 2011). Close to two dozen therapeutic antibodies have been approved worldwide for the treatment of a variety of human diseases and conditions, such as cancer, autoimmunity, inflammation, infections, organ transplant rejection and retinopathies, and many more are in advanced development stages (Beck et al., 2010; ElBakri et al., 2010).

In 1989, the academic world learned about the possibility of producing functional antibody molecules—plantibodies—in plants cells (Hiatt et al., 1989). Sustained research efforts in this theme in the years since have demonstrated that transgenic plants have the potential to be used as host systems for the production of therapeutic antibodies (Basaran and Rodríguez-Cerezo, 2008; De Muynck et al., 2010; Ma and Wang, 2012; Paul and Ma, 2011; Twyman et al., 2012).

When compared with high-scale mammalian cell culture technology and bacterial fermentation, the two current technologies used worldwide for the production of therapeutic antibodies and antibody fragments (ElBakri et al., 2010), transgenic plants offer new possibilities in terms of the storage of raw material, extensive scaling of production and final product safety due to the absence of viral, prion or bacterial endotoxin contamination (Boothe et al., 2010; Brereton et al., 2007; Fischer et al., 2012; Grohs et al., 2010; Peters and Stoger, 2011). On the negative side, production of therapeutic antibodies in plants has shown discrete yields due to low expression levels, needs a long development timescale to be implemented, and plantibodies can be immunogenic or allergenic in humans due to nonmammalian glycosylation patterns (Gomord et al., 2010; Loos et al., 2011; Rybicki, 2009).

To address some of the aforementioned disadvantages, many experimental efforts in course explore the development of production systems that speed up production time (Franconi et al., 2010; Komarova et al., 2010), the optimization of transcription and stability of expressed proteins (De Muynck et al., 2009; Streatfield, 2007) and the manipulation of the plantibody glycosylation patterns by altering the host plant cells, or the antibody-encoding genes (Gomord et al., 2010; Loos et al., 2011; Rademacher et al., 2008).

Nimotuzumab is a humanized anti–epidermal growth factor receptor (EGFR) recombinant IgG1 antibody produced in mammalian cell culture that has received approval in different countries for the treatment of head and neck squamous cell carcinoma, paediatric and adult glioma and nasopharyngeal and oesophageal cancers (Bebb et al., 2011; Bode et al., 2007, 2008; Boland and Bebb, 2010; Crombet et al., 2004, 2006; Osorio et al., 2010; Ramakrishnan et al., 2009; Ramos-Suzarte et al., 2012; Soriano et al., 2007; Strumberg et al., 2012; Takeda et al., 2011; You et al., 2010).

Because the antitumour mechanism of nimotuzumab is mainly attributed to its ability to bind to the extracellular domain of EGFR that interrupts the signal transduction cascade triggered by EGF (Berger et al., 2011; Crombet et al., 2002; Garrido et al., 2011; Talavera et al., 2009; Viloria-Petit et al., 2001), we have hypothesized that mutating the N297 position in the IgG1 Fc region gene of this antibody and using transgenic plants as hosts would produce an aglycosylated form of nimotuzumab with binding properties, ability to block the EGFR, biological half-life and experimental antitumour effect, similar to the mammalian-cell-produced glycosylated counterpart. To test this idea, we expressed the aglycosylated plantibody in stable transgenic tobacco plants and developed a bench-scale production and purification process, starting from leaf material. We have then compared the in vitro and experimental in vivo performance of the plantibody versus the original mammalian-cell-culture-produced nimotuzumab. Our results support the idea that production of aglycosylated forms of nimotuzumab and of other antibodies in transgenic plants is a feasible approach when facing scaling strategies for anticancer immunoglobulins that base their effect in the direct blockade of cell receptors or soluble ligands, essential for tumour growth and survival.

Results

Transgenic tobacco plants stably express the aglycosylated plantibody

Tobacco leaf tissues were infected with pDEGF-R-transformed Agrobacterium tumefaciens, and extracts from 96 tobacco kanamycin-resistant plants were studied for human IgG production by ELISA. The 20 best IgG-expressing lines were propagated in vitro and later grown in glasshouses for the collection of T1 generation samples.

Figure 1 compares human IgG expression in T1 lines. ELISA optical density variation between the lowest and highest values was around twofold. Tobacco line #86 was selected for progeny analysis and the identification of homozygosis. The estimated amount of the plantibody (hereafter denominated PhR3) in fresh leaves was around 0.25% of total soluble protein (TSP), equivalent to 30 μg/g of fresh leaf according to an ELISA with a nimotuzumab standard curve as reference (results not shown in detail).

Figure 1.

ELISA for human IgG expression in total soluble protein (TSP) of T1 transgenic tobacco plants. Ordinate: Optical density at 405 nm. Abscissa: Samples. B, extraction buffer (PBS–Tween); N, nontransgenic plants (negative control); #s, different transgenic plants (TSP diluted 1 : 250); P, 50 ng of nimotuzumab (positive control). Values are the average of three experiments. Bars indicate standard deviations.

Purification of the PhR3 plantibody

A homozygous tobacco line derived from #86 was propagated and grown in the glasshouse. Plant leaves were harvested and used as starting material for Protein A Sepharose IgG purification. At a 5 kg of raw material scale, the affinity chromatography bench-scale process exhibited 50% recovery, rendering antibody preparations with 96% purity as estimated by SDS-PAGE and gel densitometry. The final product was formulated in sodium phosphate, sodium chloride and Tween 80 and lyophilized. The preparation was found free of phenolics, alkaloids and pyrogen. Figure 2a depicts a reducing SDS-PAGE in which the expected 50- and 25-kDa protein bands of the plantibody individual heavy and light chains, respectively, are visible in the eluate lane. In Western blots under nonreducing conditions, PhR3 appears as a high molecular weight band (about 150 kDa), comparable to nimotuzumab (Figure 2b). A study of the purified plantibody and nimotuzumab by analytical HPLC showed that aggregated antibody forms comprehended only 0.4% and 0.6% of the preparations, respectively, being monomers the major species found (88.2% and 90.2%, respectively), over dimers.

Figure 2.

SDS-PAGE and Western blot of purified PhR3 plantibody. (a) Coomassie Blue-stained 12.5% SDS-PAGE under reducing conditions; (b) Western blot from samples transferred from 8% nonreducing SDS-PAGE, revealed with anti-human IgG-conjugated antibodies. Samples: hR3, 10 μg of nimotuzumab; N, total soluble protein (TSP) from nontransgenic plants, E1 and E2, eluate fractions from two plantibody purification processes. Arrows indicate molecular size.

Binding to Her1-ECD in ELISA

Recognition of the extracellular domain of human EGFR (Her1-ECD) was evaluated in ELISA (Figure 3a). Both antibodies reached the highest optical density (OD) values at approximately 1 μg/mL. The OD values at different antigen concentrations were always slightly smaller for PhR3.

Figure 3.

Recognition of human epidermal growth factor receptor (EGFR) by PhR3 and nimotuzumab in ELISA. Ordinate: Optical density at 405 nm. Abscissa: Protein concentration. Values are the average of three experiments. Bars indicate standard deviations. (a) Plates coated with the extracellular domain of human EGFR and tested with different concentrations of PhR3, nimotuzumab (hR3) or unrelated plantibody (PB) and developed with anti-human alkaline phosphatase-conjugated antibody. (b) Similarly, coated plates were tested with different concentrations of PhR3 or nimotuzumab, together with a fixed concentration of biotinylated nimotuzumab, and its binding was measured with streptavidin–alkaline phosphatase.

PhR3 and nimotuzumab showed similar Her1-ECD binding kinetics when competing with a biotinylated nimotuzumab preparation (Figure 3b). The calculated IC50 value favoured nimotuzumab (0.2843 nm) over PhR3 (0.6328 nm).

Flow cytometry

Fluorescence mean intensity (FMI) values for several human tumour lines with different levels of EGFR expression, incubated with increasing concentrations of nimotuzumab or PhR3, followed similar kinetics (Figure 4). Absolute FMI values at a given antibody concentration were slightly higher for nimotuzumab. The nonrelated plantibody used as negative control showed no reaction with any of the cell lines. Nimotuzumab or PhR3 did not recognize the EGFR-negative U1906 cells (results not shown).

Figure 4.

FACS analysis of PhR3 and nimotuzumab binding to human tumour cells with different epidermal growth factor receptor (EGFR) expression. Different human tumour cells were incubated with different concentrations of nimotuzumab (hR3), PhR3 plantibody (PhR3) or a nonrelated plantibody (PB), followed by FITC-conjugated goat anti-human IgG antibodies. Ordinate: fluorescence mean intensity (FMI) of stained cells. Abscissa: sample concentration.

In vitro thermal stability

The stabilities of the aglycosylated plantibody and a sample of nimotuzumab were studied in the same buffer at different times and temperature storage conditions, using SDS-PAGE, Western blot and EGFR recognition by FACS. Both antibodies were equally stable for 3 days at 37 °C and for 21 days at 4 °C. Beyond this time, a loss of EGFR recognition was seen for PhR3, associated with fragmentation visible by SDS-PAGE.

Glycosylation profiles

Figure 5 shows the HPLC 2-aminobenzamide (2AB) fluorophore-labelled N-glycans profiles of the heavy chains of nimotuzumab and PhR3, after enzymatic digestion with PNGase F and PNGase A, respectively. Nimotuzumab's heavy chain N-glycosylation is composed mainly of complex oligosaccharides, structured as monogalactosylated and a-galactosylated molecules typical of the NSO mouse myeloma cell where this antibody is produced (Figure 5a). These structures were not detected in the PhR3 heavy chain (Figure 5b). Light chains of both antibodies showed no signal in the chromatograms, as expected, and are not presented.

Figure 5.

N-glycosylation profiles of nimotuzumab and PhR3 heavy chains. (a) Nimotuzumab purified from the supernatant of myeloma cultures. (b) PhR3 purified from transgenic tobacco.

Displacement of the binding of 125I-labelled EGF to human placenta material

Figure 6 shows the ability of nimotuzumab and PhR3 to displace the binding of 125I-labelled EGF to its receptor in human placenta membrane preparations. IC50 values were 3.87 × 10−9 m for nimotuzumab and 2.04 × 10−9 m for PhR3. The nonrelated plantibody did not displace the binding of radiolabelled EGF.

Figure 6.

Displacement of 125I-EGF by nimotuzumab and PhR3 in human placenta material. Displacement curves were plotted as the inhibition percentage versus the logarithm of the concentration of ligand (nimotuzumab or PhR3 antibodies). IC50 values were calculated using Prism 5 for Windows, version 5.1. Bars indicate 95% confidence interval.

Affinity measurements by BIACore

Nimotuzumab and PhR3 Fab fragments were obtained by enzymatic digestion, and the equilibrium constants for Her1-ECD recognition were determined using surface plasmon resonance (SPR). The affinity (KD) values of the Fabs were 3.97 × 10−8 m and 6.10 × 10−8 m for nimotuzumab and PhR3, respectively.

EGFR tyrosine phosphorylation inhibition

Figure 7a shows that EGFR phosphorylation produced in A431 human cells by 100 ng/mL of EGF (lane 2) was similarly inhibited with 10 μg/mL of nimotuzumab or PhR3 (lanes 3 and 4, respectively), with respect to untreated A431 cells (lane 1). Inhibition by 10 μm of tyrphostin AG1478 (control) is seen in lane 5. Figure 7b shows that the amount of EGFR in all samples was the same, as estimated by densitometry.

Figure 7.

Inhibition of phosphorylation of human epidermal growth factor receptor (EGFR) in A431 cells by PhR3 and nimotuzumab. (a) Immunodetection of EGFR by an antiphospho-tyrosine antibody (P-EGFR); (b) Total EGFR expression determined with an anti-EGFR polyclonal antibody (EGFR). Lane 1: untreated cells. Lane 2: 100 ng/mL EGF. Lane 3: nimotuzumab (10 μg/mL) and 100 ng/mL EGF. Lane 4: PhR3 (10 μg/mL) and 100 ng/mL EGF. Lane 5: tyrphostin AG1478 (10 μm) and 100 ng/mL EGF.

H125 cell cycle progression inhibition

Treatments during 3 days with 100 μg/mL of nimotuzumab or PhR3 inhibit cell cycle progression and result in the accumulation of cells in the G0/G1 phase (67.13% and 66.84%, respectively), compared with the untreated control (54.97%). A simultaneous decrease in the percentage of cells in the G2/M phases was observed in antibody-treated cells (14.09% and 14.67%, respectively) relative to the untreated control (24.28%). The positive control tyrphostin AG1478 produces an accumulation of the cells in the G0/G1 phase (70.47%), at the expense of S-phase and G2/M phase depletions (14.17% and 11.13%, respectively). No differences in cell cycle progression phases with respect to the untreated H125 cells were seen for the nonrelated plantibody-treated samples.

Rat pharmacokinetics and biodistribution studies

In vitro immunoreactivities of 99mTc-labelled nimotuzumab and PhR3 for H125 cells were higher than 80%. Blood samples taken at 2, 25 and 50 min and 1, 4, 12 and 24 h after intravenous injection of the 99mTc-labelled antibodies showed that both immunoglobulins had similar pharmacokinetic profiles, characterized by a biphasic curve that could be adjusted to a two-compartment model (results not shown in detail).

There was a statistically significant difference between both antibodies with respect to the distribution phase half-life (T½-α) and the intercompartmental rate constant (K12). As seen in Table 1, the T½-α and K12 values for PhR3 were lower than those of nimotuzumab, indicating a larger and faster transfer to the peripheral compartment. However, no statistically significant differences were seen in the mean blood residence time (MRT), between the two antibodies.

Table 1. Serum pharmacokinetic parameters of 99mTc-labelled nimotuzumab and PhR3 in rats
Pharmacokinetic parameter99mTc-nimotuzumab (mean ± SE)99mTc-PhR3 (mean ± SE)
  1. T½-Ke, mean elimination time; T½-α, distribution half-life; Ke, elimination constant; K12, intercompartment rate constant; Cl/kg, systemic clearance; Vss, volume of distribution; MRT, mean blood residence time; AUC, area under the curve.

T½-Ke (h)5.08 ± 0.723.76 ± 0.37
T½-α (h)1.01 ± 0.790.31 ± 0.09
Ke (per hour)0.14 ± 0.020.19 ± 0.02
K12 (per hour)0.55 ± 0.441.29 ± 0.3
Cl/kg (mL/h * kg)6.69 ± 0.358.87 ± 0.47
Vss (mL)27.79 ± 5.4127.66 ± 4.88
MRT (h)14.75 ± 1.3812.68 ± 1.47
AUC (μg/mL * h)108.20 ± 14.9393.01 ± 13.41

From the radioactivity contents in urine samples (not shown in detail) and the pharmacokinetic area under the curve (AUC) parameters for both antibodies, the renal clearance/kg was calculated as 6.1% for 99mTc-PhR3 and 4.8% for 99mTc-nimotuzumab, confirming the slightly faster clearance of the plantibody-associated radioactivity. Low clearance parameters (below 2%) were found after studying faeces samples for both animal groups.

The results of the quantification of radioactivity in the rat tissue and organ samples at 24 h are shown in Table 2. We found no statistically significant differences in the biodistribution of both antibodies.

Table 2. Biodistribution of 99mTc-labelled nimotuzumab and PhR3 in rats
Tissue99mTc-nimotuzumab99mTc-PhR3
  1. All values are % ID/g (injected dose per gram of tissue), exception made of (*), where values are % of injected dose.

Femur0.08 ± 0.030.13 ± 0.03
Heart0.14 ± 0.030.14 ± 0.01
Kidneys0.38 ± 0.070.59 ± 0.07
Liver8.17 ± 1.799.42 ± 0.93
Lungs0.27 ± 0.030.31 ± 0.06
Muscle0.03 ± 0.000.03 ± 0.02
Spleen0.31 ± 0.040.48 ± 0.06
Small intestine*0.63 ± 0.250.50 ± 0.16
Large intestine*6.44 ± 0.364.97 ± 1.46
Stomach*0.20 ± 0.060.19 ± 0.02

Antitumour activity of the PhR3 plantibody

The antitumour activity of PhR3 was compared to that of nimotuzumab using two protocols where nude mice were treated either 24 h after the injection of A431 human cells (three doses levels of the antibodies) or after the tumour reached 85 mm3 (one antibody dose level). Phosphate–buffered saline (PBS) was used as negative control.

In the first protocol variant, the tumours in PBS-treated animals were measurable after a week of cell injection and continued to grow with fast kinetics. Tumours in antibody-injected mice were only detectable from day 16 and on, and the growth was characterized by a much slower constant slope. Figure 8 resumes the findings at day 36 of the experiment. There are statistically significant differences in tumour volume between PBS-treated mice and animals injected with any of the three dose levels of PhR3 or nimotuzumab. Mice treated with 1 mg/day for 10 days of either antibody had statistically significant differences in tumour volume with respect to the 0.1- and 0.5-mg doses. We found no differences per dose-level group, when comparing the two antibodies. All tumours were histologically confirmed as being derived from the same cells.

Figure 8.

Antitumour effect of PhR3 and nimotuzumab injected 24 h after A431 human tumour cell implantation in nude mice. Nimotuzumab (hR3) or PhR3 was injected at (a) 0.1, (b) 0.5 and (c) 1 mg/animal, for 10 days, 24 h after s.c. inoculation of A431 tumour cells to nude mice. Phosphate–buffered saline (PBS) was used as negative control. Both antibodies exhibited a strong negative effect on tumour growth kinetics with respect to PBS. No statistical differences in effect were found between the two antibodies on day 36 (d) of the experiment. Points/bars in graphics are volume mean values and standard deviations.

In the second protocol variant, the randomized tumour-bearing animals received eight 0.5-mg antibody doses or PBS, every other day. The experiment was evaluated on day 23 due to the first deaths in the PBS-treated control group. On day 29, all PBS-treated animals had died. Figure 9 shows the tumour growth (volume) kinetics of the three groups. PBS-treated mice exhibited fast and constant tumour volume kinetics, which was statistically different from those of the antibody-injected animals. The experiment was stopped on day 37, with animals still living in the nimotuzumab (1/4) and PhR3 (4/4) groups.

Figure 9.

Therapeutic antitumour effect of PhR3 and nimotuzumab in nude mice bearing A431 human tumours. Nimotuzumab (hR3) or PhR3 at 0.5 mg/animal was injected every other day, for 8 days, in nude mice with A431 tumours measuring approximately 85 mm3. Phosphate–buffered saline (PBS) was used as negative control. Both antibodies exhibited a strong negative effect on tumour growth kinetics with respect to PBS. Points in graphics are volume mean values and standard deviations.

Discussion

To produce the aglycosylated version of nimotuzumab, we used a vector with the CaMV35S promoter and the sweet potato sporamine signal peptide, A. tumefaciens for tobacco leaf tissue transformation, and the addition of a sequence encoding for KDEL tetrapeptide endoplasmic reticulum retention signal to the 3′-end of both antibody chain genes. This combination has been successfully used before by us to express full antibodies and antibody fragments in transgenic tobacco (Pujol et al., 2005; Ramírez et al., 2002, 2003).

In this work, we also went for the development of stable transgenic plants instead of relying on transient expression, which had been employed by us in a previous report with this same aglycosylated plantibody (Rodríguez et al., 2005). Transient expression is best suited for low-volume protein production such as personalized therapeutics, seasonal vaccines and other specialized markets (Komarova et al., 2011; Pogue et al., 2010). Stable transgenic plants are preferred for larger-scale productions and often result in better expression levels. As an example of the latter, the amount of plantibody obtained by us was 25 times higher in the stable transgenic tobacco leaf material than that found with transient expression, using the same genetic construction.

Comparing our expression values with those reported by other authors that also used a single T-DNA and stable transgenic tobacco plants, our yields were higher, with 30 μg/g of fresh leaf, versus 8.5, 3.0 and 0.9 mg/g of leaf described by Schillberg et al. (1999), Ko et al. (2003, 2005), respectively. KDEL can be a factor to consider in explaining these differences. The importance of targeting expression to the endoplasmic reticulum for better yields had been already shown by us with other full-size immunoglobulins and single-chain fragment variable (scFv) fragments (Ramírez et al., 2002, 2003). In general, KDEL has been reported to increase expression levels between 2 and 10 times, due to the longer stance of the molecule in the endoplasmic reticulum oxidizing environment abundant with chaperones and other factors important for the correct folding and assembly of complex proteins (Foresti and Denecke, 2008; Jamal et al., 2009; Sainsbury et al., 2010; Sharma and Sharma, 2009). When comparing our results with the work of the mentioned authors, it is noteworthy that Schillberg et al. (1999) and Ko et al. (2005) did not add a KDEL endoplasmic reticulum retention signal in their constructions, and Ko et al. (2003) only used the tetrapeptide in their heavy chain construction.

Another difference between our work and that of the mentioned authors is the use of the sweet potato sporamine signal peptide for PhR3. Schillberg et al. (1999) and Ko et al. (2005) used a murine signal peptide in their constructions, while Ko et al. (2003) employed native human immunoglobulin signal peptides. We could hypotesize that the use of a plant-derived signal peptide could benefit expression levels for some heterologous proteins produced in transgenic plants.

The leaf material purification process used by us was designed taking into account the potential of the aglycosylated plantibody to be further developed as a therapeutic product in its own, and followed the experience existing in our institution for other plant antibodies that are employed for human vaccine purification (Ferro et al., 2012; Valdés et al., 2003). Protein A chromatography has been shown by us and by others (De Muynck et al., 2010; Obembe et al., 2011; Wilken and Nikolov, 2012) to be effective in removing plant-related contaminants, and our PhR3 preparations tested negative for alkaloids and phenolics. Assuring Good Manufacturing Practices conditions for plantibody production also contributed with PhR3 preparations being negative for pyrogen.

Comparison of PhR3 and nimotuzumab demonstrated that aglycosylation did not produce important changes in EGFR recognition in ELISA or on live cells, nor the ability of the plant-produced antibody to inhibit receptor tyrosine phosphorylation signalling. PhR3 also effectively competed with radiolabelled EGF in the binding to human placental material and blocked cell cycle progression of human tumour cells in culture. These results are in line with those reported by Ludwig et al. (2004), which compared the anti-EGFR chimeric antibody cetuximab, produced in mammalian cells, with glycosylated and aglycosylated antibody forms expressed in transgenic corn. These authors found that plantibodies were functionally indistinguishable from the mammalian cell–derived antibody in terms of binding to the receptor, the blockade of ligand-dependent signalling and the inhibition of cell proliferation.

The discrete differences in activity that favoured nimotuzumab in several of aforementioned experiments, as well as the slightly better affinity value obtained for the nimotuzumab Fab fragment, which coincides with data produced by Talavera et al. (2009), are difficult to ascribe to aglycosylation. The role of carbohydrates in antibody structure and function (Arnold et al., 2007) is known to be crucial for the structural integrity of the heavy chain constant domains CH2 and CH3 that determine complement activation and ADCC in selected immunoglobulin subtypes, which include IgG1 (Beck et al., 2008; Nimmerjahn and Ravetch, 2006; Sibéril et al., 2007). There is no evidence, however, that aglycosylation can directly alter the antibody binding sites, if carbohydrates are not present in the original variable regions (Arnold et al., 2007), which is the case of nimotuzumab (Matamoros et al., 2001; Mateo et al., 1997; Montesino et al., 2012). Also, our in vitro thermal stability study did not provide information that could explain the differences found between the glycosylated and aglycosylated antibody forms, from a purely biochemical point of view.

In view of this, it is our belief that small differences in protein concentration and purity between nimotuzumab and PhR3, which can in turn lead to different final antibody specific activities, could be at the basis of the described phenomena. It should also be taken into account that PhR3 and nimotuzumab share 98% amino acid homology (Sanchez et al., 2007), with an extra methionine, which comes from the sporamine signal peptide, at the N-terminus of PhR3 and the KDEL tetrapeptides at the C-terminus of the light and heavy chain constant regions.

Once we demonstrated that PhR3 and nimotuzumab have very similar in vitro biological and biochemical properties, our subsequent studies centred in determining the possible influence of aglycosylation in pharmacokinetics and biodistribution of the plantibody, as well as in its antitumour effect.

The pharmacokinetics experiments performed in rats with radiolabelled PhR3 and nimotuzumab revealed discrete statistically significant differences between the two antibodies for some but not all of the examined parameters. T½-α and K12 obtained values suggested a faster transference of PhR3 to the periphery compartment, and those of T½-Ke, Ke and Cl/kg a faster elimination of the plantibody, with respect to nimotuzumab. No statistical differences were found in the mean residence times (MRT) of both molecules (14.75 and 12.78 h for nimotuzumab and PhR3, respectively), these values being similar to the 13.95 h reported by Iznaga-Escobar et al. (1998) for nimotuzumab in another rat study using 99mTc labelling. We also did not find statistically significant differences in the organ biodistribution of radiolabelled nimotuzumab and PhR3. Biodistribution patterns indicated that hepatobiliary and urinary elimination of the antibodies were comparable to those previously reported for 99mTc-nimotuzumab (Iznaga-Escobar et al., 1998).

Stability results and the lack of high molecular weight aggregates in the plantibody preparation make it difficult to ascribe the origin of the discrete differences found between PhR3 and nimotuzumab to these factors. Theoretically, the absence of Fc carbohydrates is known not to affect the interaction of IgG antibodies with the neonatal FcRn receptor (Baker et al., 2009; Kuo et al., 2010; Roopenian and Akilesh, 2007), which is one of the main responsible mechanisms for the biological half-life of immunoglobulins, so aglycosylation by itself is not a convincing explanation.

We could speculate that the reduction in molecular mass (between 2.8 and 3.4 kDa) resulting from carbohydrate absence and amino acid changes in the plantibody with respect to nimotuzumab could be at the basis of these discrete differences. Finally, small experimental errors, always present in complex animal experiments like the ones described, should not be ruled out.

Overall, pharmacokinetics and biodistribution behaviours are very close for PhR3 and nimotuzumab and support the work published by Ludwig et al. (2004) who found that the radiolabelled aglycosylated plantibody version cetuximab exhibited in vivo kinetics similar to the mammalian-cell-produced antibody.

Also, the small in vitro and pharmacokinetic differences between PhR3 and nimotuzumab had no apparent impact on the antitumour activity of the former, when used to treat A431 tumours growing in nude mice. In the study where the two antibodies were administered daily, for 10 days, starting 24 h after tumour cell injection, tumour volumes were statistically reduced with respect to controls injected with PBS, at the three employed dosages, with no significant differences between the two antibodies. In the second study, where PhR3 and nimotuzumab were administered every other day at 0.5 mg/animal after the tumours reached a measurable volume, a similar picture emerged. An interesting additional note was that in the second antitumour protocol, all PhR3-treated mice were alive at day 37 (when mice were sacrificed), versus only 25% of the animals treated with nimotuzumab.

The higher antibody therapeutic effectiveness in the protocol where treatment started after tumours were measurable can probably be explained considering that early treatment after tumour cell injection may select antibody-resistant cells (i.e. less EGFR sites/per cell or dependence of EGFR for growth; Viloria-Petit et al., 2001). The greater relative presence of these cells in the tumour mass with time would make less effective the effect of further antibody injections. If instead of this, treatment starts after the EGFR-rich cells have already preferentially populated the tumour mass, then the effect of therapeutic antibodies could be stronger in the established study time, as happened in the second scenario.

In conclusion, the work described in this paper shows that in the case of therapeutic antibodies designed mainly to block receptors on cells, or to prevent the interaction of soluble ligands with cell receptors, the aglycosylated versions of these immunoglobulins, expressed in transgenic plants, should be considered within the scope of production strategies to evaluate. Aglycosylated plantibody forms would be equally reactive with their targets, most probably exhibit very similar pharmacokinetics and biodistribution with respect to mammalian-cell-produced counterparts, and at the same time would be devoid of plant-related carbohydrates that like the β (1,2)-linked xylose, and α (1,3)-linked fucose moieties, have been considered as a potential source of antigenicity and allergic reactions for the patients (Gomord et al., 2010; Karg and Kallio, 2009; Loos et al., 2011).

Taking into consideration that nimotuzumab has already received several approvals as a therapeutic antibody for specific types of cancer, the tobacco aglycosylated PhR3 plantibody is a potential candidate for further clinical development as a related, albeit different, therapeutic product for cancer immunotherapy.

Experimental procedures

Plant transformation

Nicotiana tabacum L. cv. Habana 2.1.1 leaf discs were transformed by A. tumefaciens-mediated gene transfer method (Zambryski et al., 1983). Recombinant Agrobacterium bearing the binary vector pDEGF-R (Rodríguez et al., 2005) was used for plant cell infection. Briefly, the vector contains two cassettes composed of a cauliflower mosaic virus (CaMV) promoter, the sweet potato sporamine signal peptide, the nopaline synthase terminator and the heavy and light chain genes of nimotuzumab (Mateo et al., 1997). A mutation at the position 297 (N for Q) was introduced in nimotuzumab's heavy chain Fc region to produce an aglycosylated version of the antibody. The KDEL tetrapeptide endoplasmic reticulum retention signal–encoding sequence was also added to each chain gene 3′ end. Transgenic plants were selected with 100 mg/L kanamycin-containing medium.

Nimotuzumab

Nimotuzumab (TheraCIM®, hR3) was supplied by the manufacturer (Center for Molecular Immunology, Havana, Cuba) in 5 mg/mL vials for human use, with the antibody formulated in 3.2 mm Na2HPO4, 11.5 mm NaH2PO4, 0.15 m NaCl, 0.15 mm Tween 80, pH 7, and a reported purity of 99.7%.

ELISA for the detection of total human IgG in plant samples

Total soluble protein extracts were made by grinding fresh leaves to a fine power in liquid nitrogen and mixing 1 : 2 with PBS–0.1% Tween 20 as described by Rodríguez et al. (2005). Protein concentration was determined by the Bradford's method using a human antibody as reference (Bradford, 1976).

As described (Rodríguez et al., 2005), 96-well EIA plates (Costar; Corning Incorporated, New York, NY) were coated with 3 μg/mL of anti-human kappa light chain antibodies (Sigma-Aldrich, St. Louis, MO), and 50 μL of the test or control samples was added per well. After 16 h at 4 °C, the reaction was developed using anti-human IgG gamma chain antibodies, conjugated with alkaline phosphatase (Sigma-Aldrich) diluted 1 : 5000 for 1 h at 37 °C, and p-nitrophenyl phosphate (Boehringer Mannheim GmbH, Mannheim, Germany). A nimotuzumab standard curve was used as positive control. The contents of a commercial vial of the humanized antibody were diluted conveniently to assess the amount of antibodies present in the samples.

Antibody purification

Leaves from 8- to 10-week-old transgenic tobacco plants were ground in liquid nitrogen to fine powder, and samples were mixed 1 : 5 (w/v) with 10 mm Tris–HCl, pH 8.0, 0.5 m NaCl, 0.1 m Na2HPO4, polyvinylpolypyrrolidone (PVP), ascorbic acid and 1% Tween 20. The extracts were prepared to avoid the introduction of exogenous endotoxin, and solutions were prepared using sterile, pyrogen-free water (Sartorius arium® 611UF and 611VF water purification systems; Sartorius AG, Goettingen, Germany). All purification procedures were developed in GMP areas. Plant protease activities in homogenates and clarified extracts were minimized by controlling extraction temperature, pH, time and by adding protease inhibitors (Protease Inhibitor Cocktail 2 MultiPurpose; AppliChem GmbH, Darmstadt, Germany). Insoluble material was pelleted for 20 min at 9000 g and filtered. Immunoglobulins were purified using ProSep-vA High Capacity media (Millipore Corporation, Billerica, MA) following the instructions of the manufacturer, using a linear flow rate of 22.6 cm/h and 0.32 h as retention time. To ensure the removal of all contaminating solutes from the protein A column, a series of washings were performed with citric acid buffer pH 5.0 and antibodies eluted in the same buffer at pH 3.0. The elution fraction was submitted to a buffer change (3.2 mm Na2HPO4, 11.5 mm NaH2PO4, 0.15 m NaCl, 0.15 mm Tween 80, pH 7) using Sephadex™ G-25 (GE Healthcare Bio-Sciences-AB, Uppsala, Sweden) with a linear flow rate of 150 cm/h and was concentrated to 5 mg/mL in 50-kDa Centricon YM-50 Centrifugal Filter Devices (Millipore Corporation, Bedford, MA). The purified plantibody was lyophilized using a freeze dryer system (Ilshin Lab. Co., Ltd, Seoul, Korea) and stored at −20 °C until use. During all purification steps, quality controls were performed on supernatants, column elution fractions. ELISA, SDS-PAGE, Western blot and FACS analysis (described below) were used to analyse the purity, structural integrity and activity of plant-produced antibody. High-performance liquid chromatography (HPLC) system was used to analyse the presence of aggregated, dimers and monomer forms in plantibody-purified preparations and nimotuzumab using a Superdex™ 200 column (XK 10/300GL; GE Healthcare Bio-Sciences AB, Uppsala, Sweden) at a flow rate of 0.5 mL/min.

SDS-PAGE and Western blot

Purified antibody samples were studied in 12% and 8% polyacrylamide SDS-PAGE, under reducing or not reducing conditions, respectively. Gels were stained using Coomassie Brilliant Blue R250 (Sigma-Aldrich) or transferred to nitrocellulose membranes (Hybond-N+; Amersham Biosciences, UK). The latter were incubated for 1 h at 37 °C with anti-human IgG antibodies, conjugated to alkaline phosphatase (Sigma-Aldrich) and developed with 0.1 mg/mL nitroblue tetrazolium and 0.06 mg/mL 5-bromo-4-chloro-3-indolyl phosphate (Promega Corporation, Madison, WI). Apparent molecular weights of proteins were estimated with prestained protein molecular weight markers (Amersham Pharmacia Biotech UK Limited, Amersham Place, UK).

EGFR recognition by ELISA

The binding capacity to EGFR of antibodies was detected by ELISA as described (Sánchez et al., 2006). 96-well EIA plates (Costar; Corning Incorporated) were coated with 5 μg/mL of extracellular domain of human EGFR (Her1-ECD; Center for Molecular Immunology). The plates were blocked with PBS–0.1% Tween 20 and 1% bovine serum albumin for 1 h at 37 °C. Test samples were added in the same buffer for 1 h at 37 °C. The reaction was developed by adding anti-human conjugated alkaline phosphatase antibodies (Sigma-Aldrich) diluted 1 : 5000 for 1 h at 37 °C and p-nitrophenyl phosphate. Optical density was measured at 405 nm in a Microwell System reader (Organon Teknika Corporation, Durham, NC).

Competition ELISA assay

Plates were coated as above, and pure antibody and plantibody preparations were added to the plates at different concentration ratios in blocking buffer, together with biotinylated nimotuzumab, for 1 h at 37 °C. The reactions were developed and read as above.

Human tumour cell lines

Human tumour cell lines used were A431 (epidermoid carcinoma; CRL-1555, ATCC), H125 (lung adenocarcinoma), U1810 (non-small-cell lung cancer; Carney et al., 1985), U1906 (small-cell lung cancer; Bergh et al., 1985) and PC3 (prostate cancer; CRL-1435, ATCC). All were cultured in DMEM : F12 (Gibco Invitrogen Corporation, Canada) supplemented with 10% foetal bovine serum (FBS; Gibco Invitrogen Corporation, Burlington, ON, Canada) and 2 mm glutamine at 37 °C in 10% CO2.

Flow cytometry analysis

A431, H125, PC3, U1810 and U1906 cells (0.25 × 106) were incubated for 30 min at 4 °C with the antibody samples, washed twice with ice-cold PBS containing 0.5% BSA and incubated at 4 °C in the dark for 30 min with FITC-conjugated goat anti-human IgG antibodies (Sigma-Aldrich) diluted 1 : 60 in PBS–BSA. After two washing steps with ice-cold PBS–BSA, cells were suspended in 0.5 mL of ice-cold PBS–BSA and analysed in a PARTEC FlowMax FACS machine using WinMDI software version 2.8 (The Scripps Research Institute, La Jolla, CA). The data were expressed as FMI of stained cells. A nonrelated purified anti-hepatitis B surface antigen plantibody (CIGB, Havana, Cuba) was used as control for nonspecific binding to EGFR.

Thermal stability

Stability in time of plantibody and Nimotuzumab was performed at 4 and 37 °C, adjusted to 1 mg/mL in buffer 3.2 mm Na2HPO4, 11.5 mm NaH2PO4, 0.15 m NaCl, 0.15 mm Tween 80, pH 7. Samples at 37 °C were taken every 24 h for 7 days and every 72 h for 21 days in the case of 4 °C. The integrity and biological activity of antibodies was evaluated by ELISA, SDS-PAGE, Western blot and FACS. The assay was performed in triplicate.

Analysis of antibody glycosylation

Light and heavy chains from nimotuzumab and anti-EGFR plantibody were isolated by separation on 12.5% SDS–PAGE under reducing conditions. For nimotuzumab, enzymatic deglycosylation with PNGase F enzyme (New England Biolabs, Ipswich, MA) was carried out in-gel on Coomassie-stained bands (Kuster et al., 1997). For PhR3, a trypsin (Promega Corporation) digestion of extracted bands was first carried out (Wilm et al., 1996) and subsequently deglycosylated with PNGase A (Roche Diagnostics GmbH, Mannheim, Germany; Triguero et al., 2005). Glycans were fluorescently labelled with 2-aminobenzamide (2AB) by reductive amination (Bigge et al., 1995). Normal-phase HPLC of the 2AB N-glycan derivatives was carried out using a TSK-GEL Amide-80 column (4.6 × 250 mm) on a separation module (Merck-Hitachi, Tokyo, Japan) equipped with a fluorescence detector. Fluorescence was measured at λex = 330 nm and λem = 420 nm (Guile et al., 1996). Retention times of N-glycans were calculated in GU and compared with the reported values in the GlycoBase from Dublin-Oxford Glycobiology Laboratory (http://glycobase.ucd.i.e./cgi-bin/profile_upload.cgi).

EGF–EGFR radioligand competition assay

EGF–EGFR binding assays were performed according to Ledón et al. (2011), with modifications, using 100 μg of crude membranes from Homo sapiens (human) placenta (Macías et al., 1985), human recombinant EGF (CIGB) and 28.4 mCi/mg of 125I-EGF (CENTIS, Havana, Cuba). Displacements of 125I-EGF binding by different concentrations of unlabelled nimotuzumab, purified anti-EGFR plantibody and a nonrelated plantibody were determined in a gamma counter. Data were processed and IC50 values were calculated using Prism 5 for Windows, version 5.1 (GraphPad Software, Inc., La Jolla, CA).

Affinity measurements for Fab antibody fragments

Fab fragments of nimotuzumab and plantibody were prepared by papain digestion (Garrido et al., 2011) and prepared in 10 mm NaAc (pH 5.2) at 5 μg/mL (Talavera et al., 2009). Binding kinetics of Fabs to EGFR were determined by SPR using a BIACORE 3000 (GE Healthcare Bio-Sciences AB) at 25 °C, according to the manufacturer's instructions. A research-grade CM5 sensor chip was used to immobilize Her1-ECD (Center for Molecular Immunology) by amine coupling. Affinity constant (KD) was calculated from the ratio of the Koff/Kon rate constants using BIA Evaluation 3.2 software.

Phosphorylation inhibition assay

A431 cells (0.25 × 106/well) were grown in 6-well plates (Costar; Corning Incorporated) and incubated for 24 h at 37 °C and 5% CO2. After incubated in fresh serum-free medium for 16 h at 37 °C and 5% CO2, 10 μm of tyrphostin AG1478 (LC laboratories, Woburn, MA), 10 μg/mL of nimotuzumab, plantibody samples or medium alone was added for 2 h. Cells were then spiked with 100 ng/mL of EGF (CIGB) for 10 min and lysates prepared in RIPA lysis buffer (Cell Signaling Technology, Inc., Danvers, MA) supplemented with 50 mm NaF, 1 mm Na3VO4, 5 mm EDTA and 1 mm phenyl-methyl-sulphonyl-fluoride (PMSF). One hundred μg of total protein lysate was separated by 10% SDS-PAGE, electroblotted to PVDF membranes (Hybond-N+; Amersham Life Sciences, Arlington Heights, IL) and blocked overnight at 4 °C in buffer NETG (150 mm NaCl, 5 mm EDTA, 50 mm Tris–HCl, 0.05% Tween 20 and 0.4% gelatin) while rocking. The membranes were probed with antiphospho-tyrosine antibody (BD Transduction Laboratories™; BD Biosciences, Ontario, CA). The primary antibody was detected with its corresponding horseradish peroxidase-conjugated secondary antibody (Cell Signaling Technology, Inc.) followed by enhanced chemiluminescence development (GE Healthcare UK Limited, Amersham Place, UK). Membranes were stripped using the glycine/HCl buffer (0.2 m glycine, 0.1% (v/w) SDS and 1% (v/v) Tween 20), pH 2.2, before reprobing with anti-EGFR polyclonal antibodies (BD Transduction Laboratories™; BD Biosciences) for total EGFR expression.

Cell cycle arrest

Cell cycle phase arrest was assessed by flow cytometry using propidium iodide (PI) according to Nicoletti et al. (1991) with modifications. A total of 0.25 × 106 H125 cells/well were grown in 6-well plates (Costar; Corning Incorporated) in culture medium containing 10% FBS for 24 h, at 37 °C and 5% CO2. The medium was changed and cells were incubated with 100 μg/mL of anti-EGFR antibodies, isotype-matched control antibodies or tyrphostin AG1478 (Levitzki and Gazit, 1995; Shushan et al., 2004) for 72 h. Cells were trypsinized, washed in ice-cold PBS, fixed in methanol–acetone and stained using 100 μg/mL of PI (Sigma-Aldrich) with 100 μg/mL of RNAse A (Sigma-Aldrich) for 20 min at 37 °C. Data were collected using a FACScan Flow Cytometer (Becton Dickinson, Mansfield, MA) and analysed using WinMDI 2.8 software. Samples were assessed in triplicate and mean values calculated.

Antibody radiolabelling

Antibodies were labelled with 99mTc as proposed by Schwarz and Steinstraber (1987), and done before for nimotuzumab (Iznaga-Escobar et al., 1998). Briefly, 5 mg of antibody was reduced with 2-mercaptoethanol at a 1 : 2000 ratio (w/v) at 23 °C during 30 min and purified using PD-10 columns (GE Healthcare Bio-Sciences-AB), using phosphate buffer 0.1 m pH 8.2 for elution. One milligram antibody fractions was vacuum-dispensed and frozen at −20 °C. For labelling, a pyrophosphate kit (CENTIS) was dissolved in 5 mL of nitrogen-gased saline solution, 50 μL was added to 1 mg of antibody, and labelling was carried out with 740–925 MBq (20–25 mCi) of 99mTc at 25 °C for 10 min. The radiolabelled products were analysed in thin-layer chromatography and purified through Sephadex G-25 (GE Healthcare Bio-Sciences-AB) using 0.9% NaCl. Immunoreactivity of the radiolabelled antibodies was determined by the Lindmo method, as reported (Morales et al., 1999), using different concentrations of H125 cells.

Pharmacokinetic and biodistribution studies in healthy rats

Female Wistar rats (CENPALAB, Cuba) weighing approximately 250-g weight were injected by the tail vein with 10 mg/kg of radiolabelled antibody. Blood samples were collected at 2, 25 and 50 min and 1, 4, 8, 12 and 24 h. The radioactivity was counted in a CliniGamma counter (Wizard 1470; Wallac, Uppsala, Sweden) and expressed in cpm/mg. The reference standard was 100 μL of a 1 : 100 dilution of the injected dose. Rats were sacrificed after 24 h, and spleen, liver, kidneys, stomach, lungs, heart, intestines and muscle were surgically removed. Organ and tissues were carefully washed, dried, weighed and counted for radioactivity in a Capintec counter (CRC 35R; Capintec, Inc., Ramsey, NJ). Accumulated radioactivity was expressed as percentage of the injected dose per gram of tissue (% ID/g). The calibration was made with 100 μL of a 1 : 100 dilution of the injected dose, and pharmacokinetic analysis with Winnonlin version 2.1, using a bicompartmental model. For biodistribution analyses, we used the Statistical System StatSoft version 6. The differences between groups were assessed by the Mann–Whitney U-test. The study was carried out according to protocols approved by the Institutional Animal Care and Use Committee of the Center for Genetic Engineering and Biotechnology (CIGB).

Antitumour assays

Two types of assays were carried out using 6- and 15-week-old female athymic mice (BALB/c-nu/nu; Center for Genetic Engineering and Biotechnology, Havana, Cuba) weighing between 22 and 23 g. All experimental protocols were institutionally approved, as described above.

Protocol A

A431 cells (2.5 × 106) were subcutaneously (s.c.) injected into the right dorsal flank of 6-week-old mice. The day after tumour cell inoculation, mice were randomized into three groups of 6 or 7 animals each for intraperitoneal (i.p.) injection of antibodies at 0.1, 0.5 and 1 mg per dose or saline solution, every day for 10 days. Tumours were measured twice each week and volumes calculated by the formula (a × b2)/2, where ‘a’ and ‘b’ represent the largest and smallest tumour diameter, respectively. Five weeks (day 36) after the tumour cell injection, mice were euthanized and the tumours removed and subjected to histological examination. Statistical differences in tumour volumes were assessed using the Mann–Whitney U-test and computed using GraphPad Prism v5.0.

Protocol B

A431 cells (2.5 × 106) were injected s.c. into the right dorsal flank of 15-week-old mice. Tumours were allowed to reach approximately 85 mm3 in size, and mice were randomized into groups of four animals each. Mice were treated with a total of 8 i.p. injections of 0.5 mg of antibodies or saline solution every 2 days. Tumour volumes were calculated as described above. The animals were euthanized on day 37. Statistical differences in tumour volumes were determined as described above.

Acknowledgements

The authors wish to thank Osvaldo Oliva's staff for glasshouse plant care assistance at the CIGB. We would also like to acknowledge the contributions and help of Ariel Talavera, Ailem Rabasa, Diana Hernández and Ariana Iglesias from CIM and Yanelys Morera, Javier Sánchez and Humberto Lamdán from the CIGB. No potential conflicts of interest were disclosed in this work.

Ancillary