`
`MPI EXHIBIT 1038 PAGE 1
`
`MPI EXHIBIT 1038 PAGE 1
`
`
`
`282 Full B-chain NH 2 I-~ A-chain NH 2 B-chain COO" Lyse2gThrB3O.co O- I (OH2) 4 I NH I CO I R des-(B30) A-chain NH O- NH 2 LysB29-CO0 B-chain I (OH2) 4 I NH 1 CO I R Fig. L Schematic representation of insulin and des-(B30) insu- lin acylated in the e-amino group of the side-chain of Lys rag. The A- and B-chains have 21 and 30(29) amino acids, respec- tively, each having an N-terminal c~-amino group. R designates the aliphatic carbon-chain of the fatty acid used for acylation of Lys B29, resulting in an amide bond connecting the fatty acid and insulin moieties. By removal of Thr B3~ the negative charge of the C-terminal gets closer to the aliphatic moiety R, where- by the modified site may mimic a non-esterified fatty acid bet- ter than in the presence of residue B30 Previously, we tried to develop preparations of in- sulin analogues, soluble at pH below 4, which would crystallize spontaneously upon neutralization in the tissue [6-9]. This was achieved by shifting the isoelec- tric point of the analogues from that of human insu- lin, which is about pH 5.4, to pH 7.2-7.3 by adding two positive charges. However, the blood glucose lowering effect of the selected insulin, OPID 174, proved unsatisfactory in the clinic situation [10]. Consequently, we had to find another mechanism that would slow down the absorption rate of a soluble insulin from the subcutaneous tissue, aiming at a longer and more reproducible action than that of NPH. In particular, intra-patient variability in bio- availability and absorption rate should be reduced as compared to conventional crystalline products. One such mechanism might be binding of a soluble insulin to a large protein that is a constituent of the subcuta- neous tissue fluid. Albumin, 67, 000 Mr, is present in peripheral tissue fluids in a concentration of about J. Markussen et al.: Soluble, fatty acid acylated insulins 0.3 mmol/1. Due to its size and negative charge we ex- pected its half-life in subcutis to be sufficiently long to be useful for insulin retardation. Albumin has multi- ple binding sites for non-esterified fatty acids, the binding constants ranging from 108 1/mol for the first to 1061/mol for the fifth fatty acid [11]. In human plas- ma the normal loading of albumin is about i mole of fatty acid per mole of albumin [12], leaving a huge surplus of sites for interaction. Considering the slow disappearance rate of albumin, the abundance of al- bumin in subcutis and the surplus of fatty acid bind- ing sites on albumin, we chose to modify insulin by fatty acid acylation, hoping that the analogues would mimic fatty acids and bind to albumin, be biologically active, stay in solution at neutral pH in concentra- tions of 600 nmol/ml and, consequently, show a retar- ded absorption and action. There are three free amino groups in insulin avail- able for acylation, the two N-terminal s-amino groups of the A- and B-chains, Gly At and Phe m, and the e-amino group of Lys B29 (Fig. 1), We excluded acy- lations of Gly A~ because the potency of insulin drops markedly when groups are introduced in the A-chain N-terminus [13]. However, B-chain acylation by small groups such as acetyl and succinyl has only a minor effect on the potency, both when introduced in the N-terminal Phe m and in the side-chain of Lys B29 [13]. The properties of Phem-octadecanoyl and Phem-hexadecanoyl insulins have been reported earlier [14, 15]. When tested in rabbits and rats, re- spectively, the potencies were reported to be 0 and 22 %. Neither of these investigations attempted to measure the rate of absorption or the albumin bind- ing of the fatty acid acylated insulins. Furthermore, we knew that modifications of Phe m interfere with the formation of the insulin hexamer unit, which is the most desirable state due to its inherent stability [16]. Consequently, acylation of the e-amino group in the side-chain of Lys B29 became the preferred target for modification with fatty acids. In a series of analo- gues we deleted Thr B3~ which places the acylated Lys B29 residue in the C-terminal position of the B- chain (Fig. 1). Materials and methods Insulins. Insulins and des-(B30) insulins acylated in N~-Phe B1 and N~-Lys B29 were prepared from porcine insulin or single- chain, biosynthetic precursors using conventional peptide chemistry and recombinant technology [17]. The correspond- ing 125I-tracers, labelled in Tyr Al4, were prepared as described earlier [18]. Solutions containing 600 nmol of acylated insulin per ml, 2 Zn 2+/hexamer, 1.5 % glycerol and 0.3 % phenol were used in the testing of pharmacokinetic and pharmacody- namic properties. TyrA14(125I)-human insulin was used for la- belling of human NPH preparations. Binding studies. Binding constants were determined using im- mobilized human serum albumin (HSA) [19]. Fatty acid free
`
`MPI EXHIBIT 1038 PAGE 2
`
`MPI EXHIBIT 1038 PAGE 2
`
`
`
`J. Markussen et al.: Soluble, fatty acid acylated insulins Table 1. Binding of fatty acid acylated insulins to HSA and siR (soluble, pearance after subcutaneous injections in pigs 283 extracellular domain of the insulin receptor) and disap- Insulin AcyI-chain Binding Binding length to HSA to siR Number Relative Relative of C-atoms affinity a affinity Disappearance in pigs Tsoo/o (h) c.v. b (%) n Soluble human insulin 0 N~m-tetradecanoyl des-(B 30) HI 14 N~mg-decanoyl HI 10 NSmg-tetradecanoyl HI 14 N~B:9-decanoyl des-(B 30) HI 10 N~mg-undecanoyl des-(B 30) HI 11 N~mg-dodecanoyl des-(B 30) HI 12 N~B29-tridecanoyl des-(B 30) HI 13 N~B29-tetradecanoyl des-(B 30) HI 14 N~m9-hexadecanoyl des-(B 30) HI 16 NPH, human insulin 0 0 1.00 2 0.26 0.48 n.d. 0.063 0.76 5.1 10 5 0.58 0.41 11.9 7 4 0.12 0.76 5.6 21 6 0.27 n.d. 6.9 23 6 0.42 0.54 10.5 21 12 0.71 n.d. 12.9 12 6 1.00 0.46 14.3 15 32 0.69 0.19 12.4 23 5 10.5 41 44 The relative binding affinities for HSA and siR varied less than 10 % in the two assays. N*S29-tetradecanoyl des-(B30) HI represents the relative affinity of 1.00. The K, for binding of this analogue to HSA is 2.4 _+ 0.7 x 10 s 1/mol at 23 ~ (n = 5) and 1.0 +- 0.3 x 105 mol/1 at 37 ~ (n = 4) b Coefficients of variation (c. v.) refer to variations between pigs human albumin was immobilized on MiniLeak (Kem-En Tec, Copenhagen, Denmark), which is divinylsulphone activated Sepharose 6B, to a concentration of 0.2 mmol/1 in the gel. The immobilized HSA was suspended into buffer at seven different concentrations ranging from 0 to 10 gmol/1. The buffer con- tained 0.1 mol/1 Tris, adjusted to pH 7.4, and 0.025 % Triton X-100 for prevention of non-specific adhesion. After incuba- tion with tracer amounts of TyrA~4(~25I)-labelled insulin analo- gues for 2 h at 23 ~ free and albumin-bound insulin was sepa- rated by centrifugation. Plots of bound/free insulin vs albumin concentration were linear. The apparent association binding constant, K,, is found from the slope of the plots. The relative binding affinities of the insulins in Table 1 are the average of two assays, each of which included all the analogues. Receptor binding affinities relative to human insulin, using a high affinity variety of the soluble insulin receptor siR [20, 21], were determined in a similar assay. The siR was immobilized on MiniLeak to a concentration of 1 nmol/ml in the gel. The im- mobilized siR was diluted into the buffer to concentrations ran- ging from 0 to 20 pmol/ml. This buffer contained 0.1 mol/1 Hepes, adjusted to pH 7.8, plus 0.1 mol/l NaC1, 0.01mol/1 MgCI> and 0.025 % Triton X-100. Incubation with tracer amounts of labelled insulins and calculation of binding con- stants were performed as in the albumin binding assay, and the binding constants expressed relative to that of human insulin. The receptor binding affinities of the insulins in Table I are the average of two assays, each of which included all the analogues. Studies in pigs. The principles of laboratory animal care were followed. Specific pathogen-free LYYD, non-diabetic female pigs, cross-breed of Danish Landrace, Yorkshire and Duroc, were used throughout (Holmenlund, Haarloev, Denmark). Pharmacokinetic and pharmacodynamic studies were carried out in conscious pigs 4-5 months of age and weighing 70-95 kg after being fasted overnight for 18 h. For tissue reaction stu- dies pigs weighing 50-60 kg were used. Disappearance studies. The disappearance from the subcuta- neous injection site of a series of 125I-labelled fatty acid acyla- ted insulins, NPH and porcine albumin was measured using a modification [22] of the traditional external gamma-counting method [23, 24]. With this modified method it was possible to measure continuously the disappearance of radioactivity from a subcutaneous depot for several days using portable detectors. The measurements were performed at 1-min intervals, and the counting values were corrected for background activity. An in- sulin dose of 60 nmol (equal to 10 units of human insulin) was used and each pig received both one of the insulin derivatives and human NPH simultaneously in separate depots. The disap- pearance T50 o/0 values from 44 experiments using NPH did not follow the normal distribution, whereas the values from 32 ex- periments with NN-304 were normally distributed. Hence, the means and the variation in Ts0o/o for NPH and NN-304 were compared by the t-test and the two-sided variance ratio F-test of the log transformed values, respectively. The absorption of albumin was studied by injection of 5 nmol nSI-labelled albu- min in two separate subcutaneous depots. Glucose lowering effect after i. v. injection. LysB29-tetradecanoyl des-(B30) insulin (NN-304) or human insulin was injected in- travenously through a catheter. The dose was 0.24 nmol/kg, and five pigs received both preparations in random order with a 1-week interval. Blood samples of 1 ml were drawn into he- parinized glass tubes 15 and 10 rain before administration and between 3 and 150 rain after injection. The plasma glucose concentration for each time point was determined by a hexoki- nase method using a Technicon II analyser (Garges-les-Go- hesse, France). Since no assay was available capable of measur- ing NN-304 specifically in the presence of porcine insulin, no attempts were made to assess plasma insulin concentrations. Euglycaemic glucose clamp. Five pigs received subcutaneous injections of NN-304 or human NPH in random order with an interval of 10 days. The dose was 216 nmol (equal to 36 units of human insulin), it was injected in three equal depots of 72 nmol, in order to administer a clinically relevant dose at each site and in order to counteract the large variation of the ab- sorption of NPH. The pigs were kept euglycaemic at their indi- vidual fasting glucose levels (4.4 mmol/1) for 24 h by a variable rate intravenous infusion of a glucose solution, 1.5 mmol/1 [25]. The infusion was given through a catheter inserted in the jugu- lar vein and a Braun Infusumat Secura pump (Melsungen, Germany) was used. Depending on changes in plasma glucose concentrations observed during frequent plasma glucose mon- itoring, the necessary adjustments of the glucose infusion were made empirically. Blood samples were collected in heparinized
`
`MPI EXHIBIT 1038 PAGE 3
`
`MPI EXHIBIT 1038 PAGE 3
`
`
`
`284 > *5 o "1 '13 rr 100 80 60 40 20 16 t.-- v 12 o I-- c) o " 8 (cid:14)9 ID_ 0~ 4 CO ?5 a 0 10 20 30 40 50 Time (h) . b 7" ;k, ;k0 , , , , I .... r .... ~ .... 1 2 3 10 .5 x K a (l/tool) Fig.2. (a) Disappearance of radioactivity (mean_+SEM, n = 6) following subcutaneous injection of NPH insulin (thick line), LysB29-decanoyl des-(B30) insulin (---), LysB2%dodeca - noyl des-(B30) insulin ( .... ) and NN-304 (-). (b) Correlation between the Ts0o/o for disappearance and the association con- stant, Ka, for binding of the acylated insulin to HSA. The curve is fitted assuming that the fraction of free insulin, (cid:127)f, disap- pears with a first order rate constant kf = 0.69 h 1, correspond- ing to Ts0o/o = i h for the insulin monomer, and the fraction of albumin-bound insulin, (1-af), disappears with a rate constant k b. The overall rate constant, k, for disappearance of the insulin can be expressed as: k = c~f (cid:141) k~ + (]-0tf) (cid:141) k b. C~ depends on K~ and on the concentration of albumin in subcutis, [A] = 0.3 mmol/1, and can be expressed as 1/clf = K, x [A] + 1. The curve fits the experimental data points with a coefficient of correlation of 0.97 using the fitted k b of 0.043 h -], which cor- responds to a Ts0 ~/o of the albumin-bound fraction of about 16 h glass tubes every 15 rain, plasma was separated and glucose was determined within 1.5 min of blood sampling with a YSI (Yellow Springs Instrument, Yellow Springs, Ohio, USA) glu- cose analyser (glucose oxidase method). During the experi- ment the pigs were free to move in their pens. The paired t- test was used in the statistical comparison of NN-304 with NPH, assuming that the parameters were normally distributed. This was not possible to test with n = 5. J. Markussen et al.: Soluble, fatty acid acy~ated insulins Tissue reactions. Four pigs were given subcutaneous injections of 200 ~1 NPH (20 IU), NN-304 (120 nmol) and medium (1.5% glycerol, 0.3 % phenol, 200 nmol Zn2+/ml) subcuta- neously at distinct sites on the back 9, 5, 2, 1 day(s) and 12 h prior to killing by an overdose of pentobarbital. Human NPH and 0.9 % NaC1 were injected at adjacent sites as reference and control, respectively. All injections were given in a depth of 3 mm using a G28-cannula equipped with a stopper. Sam- ples of skin (20 x 20 ram) including underlying subcutis were fixed in Bouin's fixative. Light microscopy procedures. All tissues were embedded in paraffin wax according to standard histological procedures. Sections 6-8 ~m thick were stained by haematoxylin eosin and iron haematoxylin Picro-Acid Fuchsin. Adjacent sections were stained immunohistochemically for insulin by subsequent incul ations of the deparaffinized sections with 0.05 tool/1 tris buffered saline (TBS) (5 rain), 10 % rabbit serum in 0.05 mol/1 TBS (35 min), monoclonal mouse anti-human insulin (HUI.018; Novo Nordisk, Bagsvaerd, Denmark), diluted 1:50 in 10 % rabbit serum in 0.05 mol/1 TBS (40 rain), 0.05 mol/1 TBS containing 0.1% Triton X-100 (2 x 5 min), 0.05 mol/1 TBS (1 rain), alkaline phosphatase conjugated to rabbit anti- mouse immunoglobulin (D314; Dako, Copenhagen, Den- mark), diluted 1:50 in 10 % rabbit serum in 0.05 mol/l TBS (30 rain), and finally with 0.05 mol/1 TBS (3 x 5 rain). As chro- mogen New Fuchsin was prepared from stock solutions of: (a) 1.52 % New Fuchsin in 2 tool/1 HC1; (b) 2.24 % sodium nitrite; (c) 0.1 mol/1 Tris/HC1, pH 8.2; (d) Tris 12.33 g, naphthol AS- MX-phosphatase 9.5 g in 440 ml 0.1 mol/1 HC1, distilled water to a total volume of 1.0 litre, (BioGenex GmbH, Mainz, Ger- many) and (e) Levamisole i mol/l (Sigma, St. Louis, USA). Working solutions were prepared from 50 F1 (a), 50 ~1 (b), 5000 ~1 (c), 400 ~1 (d) and 10 ~d of (e). Incubation for 11 min was followed by three washings for 5 rain with 0.05 mol/l TBS, containing 0.1% Triton X-100, and after additional washing in running tap water for 10 min the sections were counterstained in Mayer's haematoxylin for 45 s. Finally, after 10 min of wash- ing in running tap water the sections were dehydrated in etha- nol, cleared in xylene and coverslips were mounted with Eukitt (O. Kindler & Co GmbH, Freiburg, Germany). All washings and incubations were carried out at room temperature. Three control sections without (a) primary antibody (b) secondary antibody or (c) chromogen, respectively, were treated similar- ly. All controls turned out negative, meaning that endogenous alkaline phosphatase had been sufficiently inhibited. Results Disappearance and binding studies. The Ts0 % for the des-(B30) insulins was found to increase with the number of C-atoms in the fatty acid up to 14 (Table 1). The fatty acid acylated insulins bound to HSA with binding constants in the order of 104 to 105 l/tool. The relative affinities for HSA varied less than 10 % between the two assays. The maximal binding affinity was obtained for Lys~29-tetradecanoyl des- (B30) insulin, NN-304. The absolute values of the binding constants were different from those valid in vivo, because the K a was determined in aqueous buf- fer at 23 ~ The Ts0o/o times were strongly correlated to the albumin binding constants, the coefficient of the correlation being 0.97. NN-304 had the longest
`
`MPI EXHIBIT 1038 PAGE 4
`
`MPI EXHIBIT 1038 PAGE 4
`
`
`
`J. Markussen et al.: Soluble, fatty acid acylated insulins E E v o o (9 -1 -2 , ~_ :r. q _] T -- -'[--7"T ...... 1 .... .3 , i . . , ; , , r , i r i i i i i [ i r r i i. I r r , i i O 30 60 90 ~20 "fSO Time (min) Fig. 3. The plasma glucose lowering effect in five pigs after i. v. administration of 0.24 nmol/kg of human insulin (--) and NN- 304 ( .... ), respectively. Bars represent SEM. Initial glucose concentrations were 4.6 +_ 0,1 mmol/1 for human insulin and 4.5 +_ 0.4 mmol/1 for NN-304 .'c 10 E s 0) 6 4 o c1. 2 co o c9 = 0 (9 0 6 12 18 24 Time (hours) Fig.4. Euglycaemic glucose clamp in five pigs after subcuta- neous injection of human NPH (--) and NN-304 ( .... ), re- spectively. Bars represent SEM. The glucose target levels were set to the initial values, which were 4.4 retool/1 in both groups. The achieved controls were 4.4 _+ 0.1 mmol/1 for NPH and 4.4 _+ 0.2 for NN-304 (mean _+ SD). The coefficients of var- iation of glucose concentrations within animal were 13.6+3.7% for NPH and 11.2-+2.9% for NN-304 (mean _+ SD) Ts0 % time (14.3 + 2.2 h), significantly longer than that of NPH (10.5 _+ 4.3 h) (p < 0.001). The coefficients of variation of the Ts0 % were 15% for NN-304 and 41% for NPH. Using the two-sided F-test on the log transformed data the F-value was calculated to 8.0, highly significantly different at p < 0.001. The disappearance Ts0 % of nSI-labelled porcine al- bumin after subcutaneous injection was found to be 44 h, substantially longer than that of any insulin. The removal of the C-terminal residue of the B- chain (B30) resulted in an increase in the affinity for albumin by a factor of 1.7-1.8, as demonstrated for decanoylation and tetradecanoylation (Table 1). Tet- radecanoylation of N%Lys B29 in des-(B30) insulin re- sulted in a four times greater affinity for albumin than tetradecanoylation in position N<Phe m. Within 285 the des-(B30) series an increase in binding constant was correlated with fatty acid chain length from 10 to 14 carbon atoms, whereas a chain lengthening to 16 apparently failed to improve binding any further. The affinity for the insulin receptor decreased somewhat in the des-(B30) series, from 76 to 19 % re- lative to human insulin, as the fatty acid substituent of Lys B29 increased from 10 to 16 C-atoms (Table 1). Both Phe m and Lys Bz9 tetradecanoylations resulted in the same slight decrease in affinity, indicating that neither type of substitution made close contact to the active site of insulin. bm'avenous bolus. The blood glucose lowering effect (Fig.3) of NN-304 was clearly protracted as com- pared to the effect of human insulin, and the action appeared to continue towards the end of the study, 150 rain after injection. Euglycaemic glucose clamp (Fig. 4). The glucose dis- posal rate after a subcutaneous bolus of the soluble NN-304 was compared to that of human NPH in a 24-h euglycaemic glucose clamp. The areas under the curves, i.e. the total amounts of glucose infused, 454 + 226 g and 469 _+ 114 g, respectively, are not sig- nificantly different (p = 0.88). The times when 50 % of the glucose was infused, 7.9 + 1.0 h and 6.2 _+ 1.3 h were significantly different (p = 0.04), as were the maximal infusion rates, 6.6_+ 2.2 and 9.9_+ 0.8 rag. kg -1. min -1) (p = 0.01) and the times to reach the maximal infusion rate, 6.4 _+ 2.2 h and 3.4 _+ 0.2 h (p = 0.03), respectively for the two insulins, provided these last three parameters distributed normally. Histology (Fig. 5). NPH gave a reaction in subcutis with numerous macrophages surrounding condensed heaps of NPH-crystals left in the interstitium and be- tween fat cells in subcutis. The reaction was strong on day 1 (Fig. 5 a) and turned to granulation tissue with mild formation of collagen without macrophages on day 5 and diminished on day 9. Insulin reaction was positive in NPH-crystals, macrophages with uptake and digestion of NPH-crystals, fibroblasts, reticular fi- bres, surface of collagen fibres, fat cell cytoplasm (but not in fat vacuoles) and vascular walls. Insulin reaction almost disappeared after 5 days. NN-304 showed a mild oedematous cellular infiltrate with granulocytes after 12 h, 1 day (Fig.5b) and 2 days. The findings were uniform in all four pigs. After 5 and 9 days, only a mild formation of granulation tissue was seen. No macrophages were involved. Insulin reaction was po- sitive in granulocytes and otherwise as above, bearing in mind the absence of crystals and macrophages. At all time points investigated, the medium itself showed a weak cellular infiltrate and oedema. In all pigs some injection sites, except those for the saline injections, had loci of foreign body reaction with few epitheloid and giant cells in the cellular infiltrates.
`
`MPI EXHIBIT 1038 PAGE 5
`
`MPI EXHIBIT 1038 PAGE 5
`
`
`
`286 J. Markussen et al.: Soluble, fatty acid acylated insulins Fig. 5a, b. Histological sections of subcuta- neous tissue from pigs 24 h after injection of 200 ~1 of two pharmaceutical preparations, both in concentrations of 600 nmol/ml. a Crystalline human NPH. Accumulation of NPH crystals is seen in area (A), surrounded by invading macrophages (M). (F) is fat cells, (I) interstitium with cellular infiltrates and C is collagen. Staining by haematoxylin eosin. Bar 100 ~tm. Inset Enlargement of adjacent section immunohistochemically stained for insulin (red). Densely packed NPH crystals (A), di- gested by macrophages (arrow). Bar 10 9m. b Soluble NN-304. Staining and magnification as above. Cellular infiltrations are seen in in- terstitia (I) and between fat cells (F). Inset Moderate staining of insulin in granulocytes (arrow), collagen (C) and fat cell cytoplasmic contour (F) Discussion Fatty acid acylated insulins which bind to albumin have been synthesized. The highest affinity for albu- min was found for N~-LysB29-tetradecanoyl des- (B30) insulin, NN-304. Given an albumin concentra- tion of 0.3 mmol/1 in subcutaneous tissue, this insulin is more than 96 % albumin-bound at all times, which causes retardation of its disappearance from subcutis. It cannot be excluded that binding to tissue proteins or cell membranes contributes to the retarded disap- pearance, but the observed effect can be accounted for by albumin binding alone. In the circulation NN- 304 gives rise to a protracted effect, which can be ac- counted for by albumin binding. Chemical crosslink- ing of 125I-labelled NN-304 to serum proteins, fol- lowed by electrophoresis, showed that only the albu- min band had been distinctly labelled (data not shown). NN-304 has a more retarded disappearance and a longer duration of action than NPH. The ad- vantages of administration of long-acting insulin as a solution rather than as a suspension of crystals are several. Firstly, the insulin concentration in the pre- paration remains constant through its time of use, in- dependent of resuspension of crystals by shaking of the vial or pen. Secondly, solutions will be distributed
`
`MPI EXHIBIT 1038 PAGE 6
`
`MPI EXHIBIT 1038 PAGE 6
`
`
`
`J. Markussen et al.: Soluble, fatty acid acylated insulins in a larger volume than crystals, which are filtered by the tissue and appear as clusters near the tip of the needle (Fig. 5 a). Hence, local variations in the subcu- taneous tissue are likely to influence the variations in disappearance less for solutions than for crystals. Thirdly, albumin binding might decrease the forma- tion of antibodies. Last, but not least, the attraction of macrophages after injection of crystalline insulin suspensions may contribute to the day-to-day varia- tions in bioavailability, if endocytosis of insulin crys- tals occurs to a substantial and varying extent. Injec- tion of NN-304 leads to a reaction in the subcutis characterized by oedema and granulocytes. This is clearly different from the reaction that follows injec- tion of NPH, where a large number of macrophages dominates. The time courses of the two reactions are also different. The consequence of the use of phenol and m-cresol for preservation, and of glycerol for iso- tonicity, for the occurrence of epitheloid and giant cells remains to be elucidated. This foreign body reac- tion may develop after local exposure to xenobiotics, lipids or debris of epithelial cells and keratin dis- placed by the injection cannula. The degree to which endocytosis of crystals impedes the proper manage- ment of insulin-dependent diabetes mellitus remains to be evaluated. The poor effect in the clinic situation [10] of the subcutaneously formed crystals of the long-acting, acid-soluble insulin, OPID 174, might have been due to endocytosis by macrophages. The disappearance Ts0o/o of NN-304 (14.3 h) does not approach that of porcine albumin (44 h). This ap- parent discrepancy most likely has a pharmacokinetic explanation. If the 44 h represents the sum of a diffu- sion term in the tissue and a transcapillary term, we primarily exploit the diffusion term for protraction of insulins because the on-off kinetics are rapid. Thus, when the insulin-albumin complex dissociates near the capillary wall the free insulin is transferred more rapidly over the capillary than albumin. If the fit in Figure 2 b is extrapolated to total albumin bind- ing, the Ts0 % for the diffusion term of albumin can be estimated to about 16 h (k b = 0.043 (cid:12)9 h-Z). With the fit- ted value of k b approximately one third of the trans- port of NN-304 in the tissue takes place in the free state and about two thirds in the albumin-bound state. We found no reports in the literature on tissue enzymes capable of deacylation of side-chain acyla- ted lysine residues in proteins. Although the protrac- ted disappearance is accounted for by albumin bind- ing we cannot exclude interactions with tissue pro- teins or cell membranes. The elimination of NN-304 from the circulation is most likely mediated by insulin receptor endocytosis, since a substantial receptor affi- nity (46 %) has been retained in NN-304. Hence, the results do not suggest that deacylation of NN-304 oc- curs prior to removal from the circulation. If a glucose disposal rate curve like that obtained with NN-304 in pigs (Fig. 4), can be obtained in hu- 287 mans, NN-304 appears to be suitable for providing basal insulin in the treatment of diabetes. The action profile obtained with NN-304 might permit a tighter control of blood glucose and a safer regulation at a lower level. Due to the absence of a distinct peak in the glucose disposal the dose can possibly be raised without increasing the risk of hypoglycaemia. The bars on the glucose disposal rate curves in Figure 4 re- present inter-pig variation, most likely due to vari- ations in insulin sensitivity between pigs. Previous results have shown that insulin analo- gues, having receptor affinities of 20-300 %, are equi- potent on a molar basis in vivo [25]. This is apparently also true for insulin analogues having affinity for al- bumin, but the binding to albumin renders the analo- gue long-acting after intravenous administration. Gi- ven that the binding constant of NN-304 to albumin is 1.0 x 105 l/tool at 37 ~ and the albumin concentration in the blood is 0.6 mmol/1, the albumin-bound frac- tion of NN-304 can be calculated to more than 98 %. It cannot be excluded that the reduced receptor affi- nity (46 % of human insulin) contributes slightly to the protraction in the circulation due to a decreased receptor mediated clearance rate. Since NN-304, having a receptor affinity of 46 %, is likely to be equipotent to human insulin in vivo [25] and is found to be equipotent to NPH in the subcuta- neous clamp, the bioavailability cannot be very differ- ent from that of NPH. The discrepancy between the bioavailability of NN-304 in pigs and those reported earlier for PheBl-octadecanoylated insulin in rabbits [14] and Phem-hexadecanoylated insulin in rats [15] may be due to differences in fatty acid chain length, pharmaceutical formulation, site of modification or the species used for the bioassay. The spatial position of the negative charge of the C-terminus of the B-chain is of importance for the binding to albumin. When the charge is moved closer to the fatty acid substitution in Lys B29 by deletion of residue Thr B3~ the affinity for albumin increases. Pos- sibly, the charge mimics the carboxylate group of a fatty acid and thereby enhances binding to albumin. When the fatty acid is substituted in position Phe m there is no negative