The pEGFP-C1-PAPSS2 plasmid encoding human full-length PAPSS2b C-terminally fused to an EGFP fluorescent protein (Schröder et al., 2012), was used for DpnI-based site-directed mutagenesis and subsequently for cell counting experiments. To generate the APSK37 sensor, the APS kinase domain of PAPSS2 was PCR-subcloned into a modified pDream2.1 vector with an N-terminal AcGFP1 and a C-terminal mCherry (Ebbinghaus et al., 2010). The APSK enzyme was truncated between the two isoleucine residues I220 and I221, within the flexible linker that connects the kinase and the sulfurylase domains (Harjes et al., 2005). Furthermore, the flexible and disordered N-terminal region, which is known to assist in the dimerization of the protein (Sekulic et al., 2007; Grum et al., 2010), was truncated by 37 amino acids (Δ37). Further, DpnI-based site-directed mutagenesis was used to introduce different disease related point mutations (G78R, L76Q, C43Y and T48R). All constructs were verified by Sanger DNA sequencing.

HeLa cells were grown at 5% CO₂ at 37°C in DMEM supplemented with 10% FBS, 100 U/ml penicillin and 0.1 mg/ml streptomycin. Cells were passaged at a 1:4 or 1:6 ratios at 80–90% confluence, using trypsin digestion. For transfection, cells were seeded in six-well plates (Sarstedt). Using Lipofectamine 3000 (Thermo Fisher), cells were transfected according to the manufacturers protocol. Concisely, a mixture of 125 µl Opti-MEM (Thermo Fisher) with 2 µg of the respective plasmid DNA and 4 µl P3000 reagents was prepared. After 5 min of incubation, the mixture was transferred to another solution containing 125 µl Opti-MEM supplemented with 4 µl Lipofectamine3000 reagent. Cells were incubated for 6 h after the addition of transfection mixture to the cellular growth medium at 5% CO₂, 37°C. The cells were passaged using trypsin digestion and seeded in 35 mm glass bottom dishes (Fluorodish, World Precision Instruments). Cells were grown for 2 days at standard cell culture conditions before imaging.

Fast Relaxation Imaging (FReI) was performed with transfected cells grown on 35 mm glass bottom dishes (Fluorodish, World Precision Instruments). Cells were washed with Dulbecco’s Phosphate Buffered Saline (DPBS) (Sigma-Aldrich) after removing the growth medium. 30 µL Leibovitz’s L15 medium supplemented with 30% FBS were sealed between a glass cover slip (Menzel #1.0) with a 120 µm thick imaging spacer (Sigma-Aldrich) and a glass bottom dish with cells.

FReI is a combination of wide field fluorescence microscopy with millisecond temperature jumps induced by an IR diode laser (m2k-Laser, 2200 nm). The technique was previously described (Ebbinghaus et al., 2010; Gnutt et al., 2019a). Shortly, fluorescent light was split by a dichroic beam splitter into donor and acceptor signal that was recorded using CCD cameras while the sample is rapidly heated by an IR laser. The temperature sensitive dye Rhodamine B (Sigma Aldrich) was used for the calibration of temperature jumps (Vöpel et al., 2015; Gao et al., 2016; Büning et al., 2017). The heat profile used in this study showed an average temperature increase of 2.2°C per jump at intervals of 50 s, covering a range from 23.0 to 58.2°C in 16 steps. Image acquisition was performed at one frame per second (fps) with LED exposure times typically between 50 and 200 ns. Data was recorded using AxioVision software and the images were processed and analyzed using ImageJ (National Institute of Health, United States) and further evaluated using self-written MatLab (Mathworks) codes and GraphPad Prism 6 (GraphPad).

For data analysis, fluorescence intensities were averaged throughout the cytoplasmic region for each channel individually (Dhar et al., 2011). Further, background subtraction was performed for the individual channels and the ratio of the donor and acceptor channel (D/A) calculated. The changes of D/A ratio upon temperature jump yield information about the associated conformational change. An increase in D/A refers a decrease in FRET that may be attributed to protein unfolding. To analyze the kinetics of protein unfolding, the individual channel intensities were used as D-αA according to (Dhar et al., 2011). To determine the melting point (T_M) of the protein, the thermodynamic model introduced as Better thermodynamics from kinetics (Girdhar et al., 2011) was used: D − α A ( T ) = − δ g 1 Δ T ⋅ T m R ( T − Δ T / 2 ) 2 ⋅ ( A 0 + m A ( T − T m ) ) ⋅ exp ( − δ g 1 ( T − ( Δ T / 2 ) − T m ) ⋅ ( R ( T − Δ T / 2 ) ) − 1 ) ( 1 + exp ( − δ g 1 ( T − ( Δ T / 2 ) − T m ) ⋅ ( R ( T − Δ T / 2 ) ) − 1 ) ) 2 Where, ∂g1 is pre-factor of the linear Taylor approximation of the two-state populations. ΔT is the amplitude of the temperature jump (set to 2.2°C) and A₀ and m_A are fitting parameters of the underlying baseline (with m_A set to 0).

HEK293 cells were cultured in DMEM with high glucose (Gibco, United Kingdom), supplemented with 10% fetal FBS and penicillin/streptomycin at 1%. Cells were passaged at 80–90% confluence, using trypsin digestion. Regular checks ensured that all cells were mycoplasma-free. Cells were seeded 1:8 or 1:10 in culture flasks or maintaining stocks or at 200,000 cells per well into six-well plates with microscopic slides in them. Transfection of these HEK293 cells 24 h after seeding on cover slips was performed using the XtremeGENE HP DNA transfection reagent (Roche, United Kingdom), according to manufacturer’s instructions. Cells were left growing for 24 or 48 h, then washed with ice-cold PBS and incubated with ice-cold methanol, followed by three further washing steps using PBS. Finally, cover slips with cells were mounted on microscopic slides, using fluorescence mounting media, and fixed with nail varnish. The slides were anonymized to enable blind, non-biased analysis. The slides were then viewed under a wide-field fluorescent microscope and scored at least three different sections using a ×20 objective. Cells were imaged with a ×20 objective. The number of speckles per cell was scored in large numbers of cells, in a blinded fashion. As a control, fluorescence intensity was ranked as well (low/medium/high). No correlation was found between fluorescence intensity and number of speckles, suggesting that protein over-expression levels were not linked to the observed patterns of speckles. Significant changes of population of non-speckled cells have been determined using two-way ANOVA with a post-hoc Holm-Sidak’s test correcting for multiple comparisons.

We analyzed the effect of disease-related mutations on the folding stability and aggregation of the APS kinase domain of the bifunctional PAPSS2 protein, using our recently established APSK folding sensor (APSK37) (Brylski et al., 2021). The sensor reports intramolecular FRET between the N-terminal AcGFP1 and the C-terminal mCherry fusion proteins (see Materials and Methods for details).

We analyzed PAPS synthase disease point mutations located in the APS kinase domain to understand whether the phenotypes seen clinically correlate with misfolded, destabilized or inactive protein. Therefore, we created the variants G78R, L76Q, T48R and C43Y within APSK37 (Oostdijk et al., 2015), expressed them in HeLa cells and studied their in-cell protein stability in comparison to the wt protein using Fast Relaxation Imaging (FReI). In FReI, a rapid perturbation of temperature is applied by absorption of infrared light (IR) by the sample (Ebbinghaus et al., 2010; Vöpel et al., 2017; Gnutt et al., 2019b) (Figures 1A,B). Dual-color imaging allows to measure changes in donor-to-acceptor intensity ratio (D/A) (Figure 1C) that display unfolding kinetics and thermodynamics of the protein in the cell (see Materials and Methods for details) (Figure 1D).

A structural analysis of the surface exposure of the disease-related mutants revealed that C43Y, T48R and L76Q are deeply buried in the protein core (solvent accessible surface area (SASA) ≤ 1 Å²) compared to G78R (Kabsch and Sander, 1983; Brylski et al., 2019); all of them located in close proximity in the central beta-sheet of the APSK (Figure 1E).

For APSK37 wt, we observed an increase of the normalized D/A ratio upon IR-laser heating and decrease after returning to the starting temperature (Figure 2A). This behavior is also evident for G78R mutant (Figure 2B), which exhibits an apparent two-state folding behavior as the respective unfolding kinetics can be fitted by a single exponential function. Plotting and fitting the respective amplitudes against temperature (Figures 2C,D) allowed the determination of a T_M = 46.1 ± 2.1°C which is similar to wt (48.0 ± 1.7°C) (Figure 2E). Additionally, no significant differences were found with respect to the modified standard state free energies of folding ΔG_f ^0’, suggesting that this mutation does not affect the stability of the protein (Figure 2F).

For the mutations C43Y, T48R and L76Q, we did not detect any unfolding transitions, impeding the determination of T_M. We rather observed a strong decrease in the D/A ratio (Figures 3A–C) that can be attributed to an increase in FRET by intermolecular energy transfer due to self-association (Ebbinghaus et al., 2010; Büning et al., 2017).

We then investigated if the self-association events observed in FReI result in the formation of microscopically visible aggregates. We monitored their formation by wide-field fluorescence microscopy of EGFP-labeled full-length human PAPS synthase 2 carrying the disease mutants, expressed in HEK293 cells along with PAPSS 2 wt proteins. The cellular distribution pattern of the protein and degree of aggregate formation was scored by classifying individual cells according to the number of speckles that were visible inside each cell (Figure 4A). Figure 4B illustrates the cellular distribution pattern of the protein and the degree of aggregate formation for different mutants compared to PAPSS2 wt. The mutants C43Y, T48R and L76Q caused a higher number of aggregates compared to PAPSS2 wt. The level of aggregate formation of the G78R mutation is not significantly different from wt. Thus, the results show that the self-association of the C43Y, T48R and L76Q APSK37 proteins measured by FReI is in accordance with protein aggregation of the respective mutants in the full-length PAPSS2 protein. On the other hand, the G78R mutant is stable and does not lead to aggregation both in APSK37 and PAPSS2 proteins.