Albumin (20% (w/v)) from clinical grade product was exhaustively dialysed into high purity deionised water (NIBSC media services) using 8 kDa dialysis tubing (Spectrum, Thermo Fisher, Hemel Hempstead, UK) at 2–8 °C over a period of 44 h to remove all existing excipients. After dialysis, the final protein concentration was approximately 11.1% (w/v) based on OD280 nm absorbance.

The protein solution prepared above was formulated to a final concentration of 5% (w/v) protein, 5% (w/v) sucrose (USP/Ph Eur grade, Merck Life Science Uk Ltd, Gillingham, UK), and either 15 mM histidine, 15 mM sodium phosphate or 15 mM sodium citrate (pH 7). Another formulation with 5% (w/v) trehalose (highest analytical grade, Merck Life Science Uk Ltd, Gillingham, UK) in place of sucrose was also prepared for the sodium phosphate formulation. 0.25 mg/mL of 1-13C labeled malonic acid (Sigma Aldrich Ltd., Poole, UK) was added as a probe molecule to measure microenvironment pH changes between formulations (21,22,23). 3 mL of each formulation were dispensed into eight 10 mL vials (VC005-20C 10 mL, Schott, Adelphi Tubes, Haywards Heath, UK), and vials were semi-stoppered with 20 mm diameter halobutyl igloo closures.

Freeze drying microscopy was performed on a Linkam FDCS 196 stage with liquid nitrogen cooling and programmable heating using Linkam control software and an Olympus BK51 microscope with plane polarised lighting (Linkam Scientific Ltd, Epsom, UK). A solution with known collapse temperature (5% trehalose in water, - 30 °C) was used to confirm instrument calibration. 3—5 µL of each of the sucrose formulations were tested by pipetting into a quartz crucible and trapping the droplet under a 13 mm coverslip. The droplet was frozen to -50 °C at 10 °C/min. Pressure was reduced to < 100 µbar, and then the temperature was increased at 5 °C/min until the collapse was observed. Collapse temperature can be found in the Supporting Information Table S1.

Modulated Differential Scanning Calorimetry (mDSC) was performed on a TA Instruments Q2000 DSC with autosampler and cooling accessory. (TA Instruments, Wilmslow, UK) 80 µL of each sample was added to a high-volume hermetic steel pan. Freezing was at maximum rate (nominally 10°C/min) to -90 °C and the samples were held isothermally for 10 min. The sample was then ramped to 25 °C at 3 °C/min with a modulation of 1 °C/min. Glass transitions/eutectics were calculated using Universal Analysis Software (TA Instruments, Wilmslow, UK). Glass transition temperature (midpoint) of maximally freeze concentrated solution (Tg’) are tabulated in the Supporting Information Table S1.

Glass transition temperature of the maximally freeze concentrated solution (Tg’) and the collapse temperature (Tc) by freeze dry microscopy were measured for each of the sucrose formulations to determine the shelf temperature for primary drying which would not result in cake collapse. Tg’ and Tc can be found in the Supporting Information in Table S1.

Vials were loaded into a single freeze dryer run on a Telstar LyoBeta 15 freeze dryer (Telstar Azbil spA, Terrassa, Spain). Samples were frozen by ramping the shelf to -40 °C at 1 °C/min. Samples were held for 2 h at -40 °C to ensure the solution in all vials was frozen and the ice crystallization was complete. Primary drying was started by reducing the pressure to 200 µbar and raising the shelf temperature to -20 °C. Samples were held at these conditions for 30 h. By this point, two thermocouples placed in vials near the front of the vial pack had shown inflection, although comparative vapour pressure measurement still indicated a difference between Pirani gauge and baratron capacitance manometer measurements. Secondary drying was begun by ramping the shelf temperature to 25 °C at 0.125 °C/min, and the pressure was reduced to 50 µbar. Samples were held at these conditions for 8 h. The thermocouple profiles indicated only a small lag during the ramping phase and equilibration at the shelf temperature during most of secondary drying. The vials were then backfilled with dry nitrogen to atmospheric pressure. The vials were stoppered in situ and sealed with aluminium crimp closures.

All lyophilized samples were pharmaceutically elegant well-formed cakes with no apparent collapse. Photographs of the lyophilized formulations are shown in Fig. S1 in the Supporting Information. Residual moisture contents ranged from 0.27 to 0.38% (w/w) after lyophilization. Headspace oxygen levels were < 0.15% and suggest that headspace gases should not adversely impact storage.

Residual moisture in the lyophilised cakes at time zero was measured by coulometric Karl Fischer (KF) titration (A1-Envirosciences, Blyth, UK.). Residual moisture was not measured at extended time points because the vial headspace was backfilled with dry nitrogen after the lyophilization cycle, and vials were sealed with aluminum crimp caps. In a drybag (Pyramid, Cole Parmar, London UK) with RH < 10%, a single vial of each product was distributed between three 4 mL HPLC autosampler vials (Thermo Fisher Ltd, Loughborough, UK). Vials were capped with screw caps with a polytetrafluoroethylene (PTFE) membrane seal. Samples were then analyzed on an automated KF system based upon the AquaFast (A1 Envirosciences Ltd, Blyth, UK) with a CA-200 coulometer (Mitsubishi, supplied by A1 Envirosciences) and a GX-270 robotic sampler (Gilson, supplied by A1 Envirosciences). Mitsubishi cathode and anode reagents were used (Aquamicron, Mitsubishi, supplied by A1 Envirosciences) and a check solution of known water content was used to check performance on each assay (Aquamicron P, Mitsubishi). The moisture content of empty vials was measured and used as a blank subtraction for the freeze-dried materials. The mean of three determinations with CV was reported for each sample. Residual moisture content can be found in Table S2 in the Supporting Information.

Frequency-modulated spectroscopy at 760 nm (FMS 760, Lighthouse Instruments, Charlottesville, VA, USA) was used as a non-destructive method to measure residual oxygen in the headspace gas on three vials of each freeze dried formulation. Calibration was performed with NIST traceable oxygen standards (0% and 20%) in identical 10 mL vials. Oxygen headspace data can be found in Table S2 in the Supporting Information.

Samples were stored at -20 °C, 4 °C, 37 °C, and 56 °C for up to 12 months. At 6 and 12 months, samples were removed for analysis either by SEC to assess aggregation, or by ssNMR to determine 1H T1 relaxation times.

HPLC SEC was performed according to the Ph Eur SEC HPLC method (2.2.30, Ph Eur 9th Ed., Strasbourg, France; Council of Europe 2019). Samples were reconstituted to 3 mL with deionised water and then assayed by loading approximately 100 µL of 0.2 µm filtered sample injection volume onto a Thermo Ultimate 3000 HPLC (Thermo Fisher, Hemel Hempstead, UK) with a disodium hydrogen phosphate, sodium dihydrogen phosphate, sodium chloride, sodium azide buffer as mobile phase and with isocratic elution on a TSKgel G3000SWXL column (7.8 × 300 mm, I.D. x length) at 0.5 mL/min. Protein molecular size distribution was assessed by OD280nm and compared to a non-lyophilized albumin control preparation.

ssNMR data at 0 and 6 months were acquired using a Tecmag Apollo spectrometer operating at 100.6 MHz for 13C (9.4 T static magnetic field). 12 month data was acquired using a Bruker Avance spectrometer operating at 100.505 MHz for 13C. In a drybox (RH < 3%), lyophilized powders were packed into zirconia rotors and sealed with Kel-F endcaps (Revolution NMR, LLC, Fort Collins, CO). All experiments were performed at room temperature using a double resonance magic angle spinning (MAS) probe. All 13C spectra were acquired under MAS (24) at 4 kHz, using ramped-amplitude cross-polarization (CP) (25), total sideband suppression (TOSS) (26), and SPINAL-64 1H decoupling (27). An approximately 4 µs 1H 90º pulse, and a 2 ms contact time were used in all experiments. 3-Methylglutaric acid was used as an external chemical shift standard and to optimize spectrometer settings (28).

1H T1 relaxation times were measured using a saturation recovery experiment through 13C observation. A saturation recovery experiment is a pseudo 2D NMR experiment where the 13C spectrum is plotted against a variably delay time. 90º pulses were used to saturate the magnetization, followed by a variable pulse delay to allow the magnetization to recover back to equilibrium. After Fourier transform, a peak from each sample component was identified (25 ppm HSA, 75 ppm disaccharide, 176 ppm malonic acid) based on high intensity and minimal overlap with other system components. The chosen peaks were then integrated over the full width at half maximum, and the areas were plotted against the delay time. Equation 1 was fit to the data using KaleidaGraph (Synergy Software, Reading, PA)

where M is the area of the chosen peak, t is the variable delay time, M0 is an amplitude parameter from the fit, and T1 is the spin-lattice relaxation time in the laboratory frame. 1H T1 relaxation times are reported with error values that reflect the goodness-of-fit of Eq. 1 to the data.