the same as the sequence obtained by Stowell et a/. (1991) but it is different to the sequence derived from the cDNA reported by Rado et al. (unpublished) which was subsequently reported by Anderson et al. (1989). In these instances only three contiguous arginines were found in the first five positions. The extra arginine in LfN and in recombinant hLf was not expected to pose any problems but for future reference it is important to clarify the amino acid numbering system used. As the numbering system proposed by Anderson et al. ( 1989) has been extensively used in previous work this system was retained and the additional arginine was referred to as 3a. The sequence and the numbering system used is shown in Table 2.
Table 2 Sequence alignment of the N-terminus of human lactoferrin
5 6 7 8 etc S uence found for LfN
A 3 . 2 . 2 . 3 G lycosylation
-etc -etc
The amino terminal half of human lactoferrin contains a single recognition site for the addition of an N-linked carbohydrate chain. The molecular weight of LfN estimated from the mobility on SDS gels was -40 000 Da suggesting that the fragment is indeed glycosylated since the amino acid composition of LfN would give a molecular weight of 37 084 Da. The presence of an N-linked carbohydrate moiety was investigated by treating LfN with the endoglycosidases Endo F and PNGase F (Section A 1 .5. 10) according to the procedures previously used for the deglycosylation of native lactoferrin (Norris et al., 1989). Treatment with these enzymes decreased the molecular weight of LfN by -3 kDa (Fig. 26, lane 4) whereas no change in the molecular weight was detected when LfN was treated with N-octyl J}-D-glycopyranosidase which is specific for 0-linked carbohydrate groups (Fig. 26, lane 3). This result clearly demonstrated that LfN did indeed have at least one N-linked carbohydrate chain.
When LfN was purified it should be recalled that a minor contaminant -37 kDa in size was found to copurify with LfN. The contaminating band appeared to be very similar in molecular weight to deglycosylated LfN and to that expected for
unglycosylated LfN. The possibility that this band was in fact unglycosylated LfN was investigated by treating a small amount of LfN from fraction 2 (Fig. 24) with PNGase F. This treatment converted the doublet into a single band with a
molecular weight of 37 kDa suggesting that the impurity seen was unglycosylated LfN. The nature of the lower molecular weight band was unequivocally identified
Fig. 26 Deglycosylation of LfN.
Samples of pure Lf'N were treated with deglycosylation enzymes (Section Al.5.10) and then the samples were subjected to gel electrophoresis on a 12% polyacrylamide gel. Lane (1) molecular weight markers (kilodaltons); (2) pure LfN; (3) Lf'N which has been treated with N-octyl �-D glycopyranosidase; (4) Lf'N treated with PNGase and Endo F.
1 2 3 4
97 kDa -
45 kDa -
3 1 kDa -
by western blotting. A gel equivalent to that shown in Fig. 27 a was transferred to nitrocellulose and reacted with antibodies to human lactoferrin (Section A l .5.6). It is clearly evident that in lane 3 (Fig. 27b) both the high molecular weight band and the less intense lower molecular weight band have reacted with the antibodies, confirming that they are both lactoferrin species. The difference in molecular weight is thus attributed to the complete absence of carbohydrate groups as the mobility of the lower band does not change when it is treated with PNGase F. The presence of the lower molecular weight band in preparations of LfN raises
several interesting possibilities: (1) Is a small amount of LfN synthesised and
secreted by the
BHK
cells without the addition of carbohydrate groups? (2) Is thelower molecular weight band due to the gradual loss of carbohydrate with time?
Fig. 27 Identification of the lower 37 kDa band found in preparations of LfN. In an attempt to positively identify the the lower molecular weight which copurified with Lf'N samples, equivalent to fraction 1 and fraction 2 (Fig. 24), were separated on a 1 2%
polyacrylamide gel (Fig. 27a), transferred to nitrocellulose (Fig. 27b) and then reacted with anti human lactoferrin antibodies (Section AI S.6). Samples for both gels were (I) Molecular weight markers (kilodaltons); (2) fraction 1; (3) fraction 2.
Fig. 27a 1 2 3 Fig. 27b 1 2 3
97 kDa -
45 kDa -
3 1 kDa -
likely that a small amount of LfN escapes glycosylation as it passes through the endoplasmic reticulum. This is supported by the observation that the lower band
is present in freshly prepared LfN and that the intensity of this band in samples of LfN which were up to six months old did not appear to increase. Initially the small amount of unglycosylated protein was not a cause for concern as it could be
separated from the higher molecular weight band and it did not represent more than
-2% of the protein synthesised. In later batches, however, the multiplicity of LfN species became more of a problem because often up to four bands were present at quite high levels. The variation in the extent of glycosylation between batches is shown in Fig. 28. In the later batches it was not possible to purify one form of LfN away from the others. This made crystallisation of this protein more difficult (
�
ection B3.2. 1 ). Reasons for the difference in the extent of glycosylation between batches are unclear as the cells used to produce the protein have always been derived from the same original batch of frozen cells. The problemsencountered here raise several issues regarding the capacity of BHK cells to
produce recombinant glycoproteins at high levels in a consistent manner. The time available did not allow an investigation into the conditions affecting this processing but it would make an interesting topic of study for the future.
Fig. 28 Analysis of the variation in protein purity between different batches. A small amount of protein was saved from several different batches and compared on a 12% polyacrylamide gel. Lane ( I) Molecular weight markers (kilodaltons); (2, 3 and 4) Protein samples from several of the earlier preparations.; (5, 6 and 7) Protein samples from several of the later batches of protein.
66 k.Da - 45 k.Da -
31 k.Da -
� 2 . 2 . 4 S pect roscopy
1 2 3 4 5 6 7
The shift in the visible absorption maximum from 466 nm in native human lactoferrin to 454 nm in
UN-
suggested that there existed small differences in the iron coordination between the two proteins. In an attempt to analyse thedifferences an ESR spectrum was recorded for iron saturated LfN (Fig. 29). The spectrum obtained was essentially indistinguishable from that obtained for native lactoferrin with both spectra characterised by a split peak at 1 500 Gauss. The similarity of these ESR spectra suggested that differences between the sites were likely to be minor.
In an attempt to understand the nature of the differences in the metal sites LfN was saturated with Cu2+ instead of Fe3+. Copper is a more informative metal as it gives rise to a more interpretable ESR spectrum (Aisen and Leibman, 1972). When Cu2+ was added to ApoLfN (prepared by lowering the pH to 4.0, section A3.2.2.5), as CuCI2, the protein became yellow in colour. The yellow colour is due to copper-tyrosine charge transfer absorption and clearly demonstrated that
Fig. 29 ESR spectra of iron saturated LfN and hLf. The ESR spectra were recorded as described in section A 1 .5. 11 . FeLfN
1500 magnetic field (Gauss)
1500 magnetic field (Gauss)
2300
binding had occurred. Copper saturated native lactoferrin (Cu2Lf) was prepared as described by Smith et al. (1991). Samples prepared in this way were used to record UV-visible and ESR spectra. Analysis of the UV-visible spectra (Fig. 30)
shows that, as for FeLfN, the absorption band has shifted to a lower wavelength in CuLfN, compared with Cu2Lf (absorption maxima at 424 nm instead of 434 nm). The same protein sample used for the UV -visible analysis were concentrated to 10
mg/ml and used to record ESR spectra (Section Al .4.10). The two copper ESR spectra (Fig. 3 1) are similar, although not identical, and the calculated splitting constants (An and grr and g!) agree very well (data not shown). The values obtained
are
characteristic of Cu(II) in a rhombic environment providing further evidence that no major changes have occurred at the metal binding site of LfN. The more complex pattern of peaks seen at -3250 Gauss in Cu2Lf compared to CuLfN may be due to the fact that Cu2Lf contains two metal binding sites, which may be slightly different, while CuLfN contains only a single metal binding site. Fig. 30 Comparison of the copper UV -visible spectra obtained from LfN and hLf.Spectra were recorded between 250 run and 700 nm for both LfN (solid line) and hLf (dashed line).
The inset contains an enlargement of the region where the shift in the maximum has occurred.
360 380 400 420 440 460 480 o.o
250 350
� . 2 . 5 Metal binding and release
450 550 650
Wavelength (run)
The pH dependence of iron release from LfN was measured and compared with the value obtained for native human lactoferrin. The extent of iron saturation,
estimated from the absorbance at 466 nm ( 454 nm for native LfN), was monitored over the pH range 8.0 to 2.0 (Section A 1 .5.12). For LfN iron removal began at pH -6.0 and was essentially complete at pH -4·0. When native lactoferrin is treated in the same way iron removal does not begin until pH - 4·0 and is not complete until pH -2·5 (Fig. 32).
Fig. 3 1 ESR spectra of copper saturated LfN and hLf. The ESR spectra were recorded as described in section A1.5. 11. CuLfN
2900 magnetic field (Gauss)
Fig. 32 Analysis of the pH dependent release of iron from LfN.
Samples of human lactoferrin ( • ) and LfN ( 0) were dialysed for 48 hours against buffers of different pH (Section A.5.12). The spectra of the dialysed samples were recorded and the percentage of iron bound detennined.
1 00 80 � =
_g
60 =.§
40 � 20 0 1 . 5 2 . 5 3 . 5 4 . 5 5 . 5 6 . 5 7 . 5 8 . 5 pHThe reversibility of iron binding to LfN was shown by binding studies carried out after complete removal of iron at pH 4·0. Titration of LfN with iron (added as ferric-nitrilotriacetate complex) at pH 7 ·8 showed that the iron free LfN binds one molar equivalent of iron when either fluoresence quenching or absorbance at 454 run were measured (Fig. 33). The curve obtained for fluoresence quenching does not flatten out as much as for other lactoferrins. The reason for this difference is unknown although several tyrosines at the back of the molecule are exposed to
solvent in the half molecule (see section B3.3) and may be free to interact with the iron in solution in a non-specific manner .
. 3 DIS C U S S I O N
Analysis of LfN has shown that the signal peptide has been correctly processed during transfer of the recombinant protein across the endoplasmic reticulum. The protein has also been shown to be glycosylated although the precise nature of the carbohydrate groups added has not been determined. In some preparations of LfN several bands of different mobility were obtained when analysed by SOS-PAGE. Treatment of this protein with glycosidases specific for N-linked groups converted
all species to a single discrete band indicating that the difference in molecular weight was due to variation in the nature of N-linked carbohydrate groups. These results
are
similar to those obtained using mutant forms of human lysozyme containing artifical glycosylation sites and a BHK cell based expression systemFig. 33 Titration of UN with iron at pH 7 .8.
Iron was removed from Lf'N by dialysis against pH 4.0 buffer (Section A1.5. 12) and then iron was added back to the protein after it had been equilibrated in 0.01 M HEPES, pH 7.7, 0.2 M NaCl. In Fig. 33a iron binding was measured by monitoring the absorbance at 454 nm while in Fig. 33b the extent of fluorescence quenching was measured.
Fig. 33a 0. 1 2 0. 1 0 -
�
� 0.08 :::!, 8 a 0.06 -e � .t:J < 0.04 0.02 0 . 0 0 . 2 0 . 4 0 . 6 0 . 8 1 . 0 1 . 2 1 . 4 1 . 6 1 .8r (metal ions I protein)
Fig. 33b 1 00 90 80
�
70�
60 0 50 0 0:: 40 � 30 20 1 0 0 0 . 0 0 . 5 1 . 0 1 . 5 2 . 0 2 . 5r (metal ions I protein)
(Horst et al., 199 1). In this study they showed that the amount of carbohydrate added to the recombinant lysozyme produced from a single batch of cells varied and that the level of glycosylation was dependent on the position of the
glycosylation site in the polypeptide chain. This apparent variation in the extent of protein glycosylation raises questions about the fidelity of the glycosylation
machinery in BHK cells. This has not been pursued in this study.
Spectral analysis of LfN has shown that the metal binding characteristics are very similar to those of the native protein. The similarity of the UV-visible and ESR spectra obtained from LfN and hLf indicates that the ligands involved in metal
binding have essentially the same geometry in both proteins. The small reduction in Amax to 454 nm implies a slight change in the ligand field around the iron atom.
A similar shift has been reported to occur in LfN30, the proteolytically derived
fragment of human lactoferrin, where a maxima at 440 nm was found (Le grand et
al.,
1990). In the case of transferrin no such consistent affect is seen with themaxima
of sTfN, the recombinant N-terminal half of human transferrin, reported tobe higher at -473 nm (Woodworth et
al.,
199 1 ). No shift has been reported forsTfN35, the proteolytically derived fragment of human transferrin (Lineback-Zins
and Brew, 1980). Presumably there
are
very small changes in the precise natureof the metal site, but the peaks
are
broad and the significance of these results is notclear.
The most striking difference between the properties of LfN and native hLf is the lesser pH stability of the half molecule with respect to iron release. The iron release curve for LfN is similar in form to that of native lactoferrin but is displaced 2 pH units higher and is in fact very similar to that of human serum transferrin which is 50% desaturated at pH 5.2. A similar effect has been noted for the proteolytically derived N-terminal fragment of lactoferrin, LfN30 (Legrand et
al.,
1990), although the later is even less stable, with iron release beginning at pH -6.6 and being complete at pH 6.0. If the pH stability is described in terms of pH50, the pH at which 50% of the iron has been released, the values are Lf 3.0, LfN 4.8, LfN30 6.3, compared with 5.2 for sTf and 5.4 for sTfN35. a
proteolytically-derived transferrin half-molecule (Lineback-Zins & Brew, 1980). These results suggest that the interaction of the two lobes of lactoferrin may contribute to the greater stability of iron binding at low pH in lactoferrin compared to transferrin. The possible reasons for these differences will be discussed more fully later (Section 2. 1 ).
Part A : Chapter Four