Sodium glucose co-transporter 2 inhibitor prevents nephrolithiasis in non-diabetes by restoring impaired autophagic flux

Summary

Background

Sodium-glucose cotransporter 2 inhibitors (SGLT2i) offer significant cardiovascular and kidney protection, independent of diabetes mellitus (DM). Recent cohort studies also suggest that SGLT2i can decrease the risk of nephrolithiasis in patients with DM. We aimed to use both animal models and human data to investigate whether SGLT2i can prevent nephrolithiasis and explored autophagy as a possible mechanism.

Methods

We utilised SGLT2i, dapagliflozin (DAPA), on a glyoxylate (GOX)-induced calcium oxalate (CaOx) nephrolithiasis non-DM mouse model to test whether SGLT2i inhibited CaOx stone formation through modulating autophagy. Moreover, the clinical data retrieved from the National Health Insurance Research Database was analysed to confirm the findings from animal models.

Findings

DAPA increased urine citrate, magnesium, pH, and decreased oxalate, effectively inhibiting CaOx stones in GOX mice. While autophagy was increased in the kidneys of GOX mice, as demonstrated by upregulated AMP-activated protein kinase (AMPK) and increased LC3B conversion; impaired autophagic flux was indicated by p62 accumulation. DAPA improved autophagy by downregulating mammalian target of rapamycin (mTOR), AMPK, and restoring autophagic flux. Rapamycin co-treatment preserved DAPA's nephrolithiasis inhibition, while hydroxychloroquine (HCQ) co-treatment abolished it. Finally, cohort data confirmed that SGLT2i reduced nephrolithiasis risk, but this protective effect disappeared if HCQ had been used within the prior year, suggesting that HCQ may compromise SGLT2i's protection against nephrolithiasis.

Interpretation

SGLT2i, DAPA, inhibits nephrolithiasis by restoring impaired autophagic flux, and co-administration with autophagy inhibitor, HCQ, compromises SGLT2i's protection.

Funding

This research was funded by grants from the National Science and Technology Council, Taiwan (110-2314-B-006-023, 110-2320-B-006-017MY3, and 112-2314-B-006-058) and the research grants (NCKUH-11202005, -11210020) from the National Cheng Kung University Hospital, Tainan, Taiwan.

Research in context.

Evidence before this study

Sodium-glucose cotransporter 2 inhibitors (SGLT2i) are renowned for their glucose-lowering effects in type 2 diabetes mellitus (DM) management, alongside significant benefits in cardiovascular and kidney health. Recent cohort studies have extended these benefits to reducing the risk of nephrolithiasis in patients with DM, suggesting a potentially pivotal role in kidney stone prevention.

Existing evidence has shown the therapeutic effects of SGLT2i on nephrolithiasis maybe related to the downregulation of osteopontin expression and the suppression of inflammation. However, recent studies have suggested that modulation of autophagy has a profound impact on kidney stone formation, the mechanisms by which SGLT2i influences autophagy to inhibit nephrolithiasis are still unclear. Therefore, it is desirable to identify the primary molecular mechanism underlying SGLT2i's therapeutic effects on nephrolithiasis.

Added value of this study

We found in this study that autophagy was significantly increased in the kidneys of non-DM mice with calcium oxalate (CaOx) nephrolithiasis, but autophagic flux appeared to be impaired. Dapagliflozin (DAPA) completely inhibited CaOx nephrolithiasis formation. Although increased urine citrate, magnesium, pH, and decreased urine oxalate might explain the beneficial outcome of DAPA, it paradoxically increased urinary sodium and uric acid. DAPA restored impaired autophagic flux and enhanced lysosome biogenesis, leading to inhibit nephrolithiasis formation.

Most importantly, we further used clinical data from Taiwan's National Health Insurance Research Database (NHIRD) to confirm our hypothesis and we found that patients receiving SGLT2i were associated with a reduced risk of incident and recurrent nephrolithiasis events. Interestingly, we found the SGLT2i-mediated reduced risk disappeared when patients were concomitantly exposed to autophagy inhibitor, HCQ, suggesting that the reduction of nephrolithiasis from SGLT2i was by modulating autophagy.

Implications of all the available evidence

Our findings suggest that SGLT2 inhibitors (SGLT2i) could be a promising option for preventing nephrolithiasis in clinical settings. Additionally, this study uncovers a mechanism by which SGLT2i prevents kidney stone formation, providing a solid evidence base for future randomised clinical trials.

Introduction

Nephrolithiasis is increasingly prevalent globally, with calcium oxalate (CaOx) as a common composition, often idiopathic without definite mechanisms.¹ Surgery remains the mainstay of treatment for CaOx stone disease²; while few pharmacotherapies have been applied on nephrolithiasis with unsatisfactory efficacy and negligible side effects, resulting in relatively low patient adherence.³ Considering that metabolic syndrome and diabetes mellitus (DM) are all strongly associated with an increased risk of nephrolithiasis,4, 5, 6 antidiabetic medications are thought to be potentially therapeutic for nephrolithiasis.

Sodium-glucose cotransporter 2 inhibitors (SGLT2i) are the newest class of antidiabetic agents and provide promising cardiorenal protection effects in patients with and without DM.7, 8, 9 SGLT2i-mediated renal protection, isolated from its glucose-lowering properties, has been linked to reductions of excessive plasma volume and oxidative stress, as well as an improvement in tubular oxygenation.¹⁰ Moreover, SGLT2i increases urinary citrate, a key nephrolithiasis inhibitor,¹¹ but, paradoxically, increases urinary sodium and uric acid excretion,¹² two nephrolithiasis promoters.¹ Thus, SGLT2i could be beneficial or detrimental to the development of nephrolithiasis.

Multiple sub-cellular mechanisms, particularly autophagy activation, are implicated in the renoprotection of SGLT2i.¹³ SGLT2i enhances autophagy,¹⁴ through regulation of AMP-activated protein kinase (AMPK) and mechanistic target of rapamycin complex (mTOR).¹⁵ Enhanced autophagy safeguards cellular integrity, preserving renal function.¹³ Meanwhile, recent studies have suggested that modulation of autophagy may have a profound impact on kidney stone formation.¹⁶ Because both activation and inhibition of autophagy may contribute to stone formation, whether SGLT2i administration by affecting autophagy can inhibit nephrolithiasis needs to be investigated urgently.

Two recent large cohort studies have demonstrated that SGLT2i therapy was associated with an approximate 30%–50% reduced risk of urinary tract stone events.^17,18 In the past year, several large-scale studies have also demonstrated that SGLT2 inhibitors reduce the risk of nephrolithiasis.19, 20, 21 However, studies attempting to explain the underlying molecular mechanisms of SGLT2i against kidney stone formation remain lacking. One recent study applied phlorizin, a class of dual SGLT1 and 2 inhibitor, on hyperoxaluric CaOx rats, and found that SGLT1/2 inhibition suppressed kidney stone formation possibly through reducing osteopontin (OPN) expression and renal inflammation.²² Additionally, they used male SGLT2-knockout mice and discovered that SGLT2-specific inhibition could attenuate the formation of kidney CaOx stones. These results suggest the beneficial effect of an experimentally used nonselective SGLT inhibitor phlorizin on kidney stone formation. Therefore, our study aimed to investigate the effect of SGLT inhibitors on nephrolithiasis in non-DM mice. We hypothesised its efficacy in suppressing CaOx kidney stones through autophagy modulation. To be more practical in the real world, we examined the effect of dapagliflozin (DAPA), a classic SGLT2i, in non-diabetic hyperoxaluric CaOx stone mice to test the hypothesis that SGLT2i directly inhibit CaOx stone formation through activation of autophagy and AMPK signalling pathway. Additionally, we retrieved the clinical data from a nationwide population-based database to investigate the association between SGLT2i use and nephrolithiasis risk. We also evaluated the risk of nephrolithiasis in patients concomitantly exposed to an autophagy inhibitor, hydroxychloroquine (HCQ), allowing verification of in vivo findings.

Methods

Study protocol

Eight-week-old male C57BL/6JNarl mice, from National Laboratory Animal Centre, Taiwan, were randomly divided into 3 groups (n = 10 in each): control, glyoxylate (GOX), and GOX + DAPA. All mice had free access to water and regular chow. DAPA (10 mg/kg/day, Forxiga, AstraZeneca, Cambridge, UK, catalogue #4895160300680) was given via oral gavage to the GOX + DAPA group for six consecutive days (Fig. 1a). After the second day, each mouse in the GOX and GOX + DAPA groups received an intraperitoneal (IP) injection of GOX (100 mg/kg/day, Santa Cruz Biotechnology, Santa Cruz, CA, USA, catalogue #sc-251025) for five consecutive days. DAPA was given 30 min before GOX injection. On day 7, all mice were sacrificed with isoflurane (Attane 074416, Panion & BF Biotech, Taipei, Taiwan, catalogue #1349003), and serum, urine, and organ samples were then harvested and stored for further analysis. To investigate the role of autophagy in kidney stone formation, we co-treated DAPA with various autophagy-modulating compounds to assess their therapeutic effects on kidney stones (Fig. 1). Due to the differing pharmacokinetics of these compounds, we extended the experimental period by an additional 6 days, resulting in a total of 11 consecutive days of GOX IP injections at 100 mg/kg/day. This extension was designed to allow for effective co-treatment without exceeding the optimal period for stone formation as established by the Okada et al. study.²³ To examine autophagy flux, hydroxychloroquine (HCQ, 100 mg/kg/day, Plaquenil®, Sanofi-Aventis, Spain, catalogue # HY-B1370) was given via oral gavage for 12 consecutive days as co-treatment, 30 min before each DAPA administration or GOX injection. Rapamycin (Rapa, 8 mg/kg/day, Pfizer, USA, catalogue #13346) was given via oral gavage for 12 consecutive days as co-treatment, 30 min before each DAPA administration or GOX injection.

Fig. 1.

The experimental protocols of co-treatment with autophagy-modulating compounds. (a) Experimental design of cotreatments with Rapamycin (Rapa) and dapagliflozin (DAPA) in Glyoxylate (GOX)-induced calcium oxalate (CaOx) mice. (b) Experimental design of cotreatments with hydroxychloroquine (HCQ) and DAPA in GOX-induced CaOx mice.

Urine and serum chemistries

Urine Ca, Na, K, Mg, and uric acid levels, as well as serum BUN, creatinine (Cr), and glucose levels, were measured using a Dri-Chem 4000i analyser (Fuji, Tokyo, Japan) and urine pH value was analysed using an RT-4010 analyser (Arkray Inc., Kyoto, Japan). Urine Citrate (Cit) and oxalate (Ox) were measured by a commercially available colourimetric assay kit (Elabscience, Biotechnology, Wuhan, China, catalogue # E-BC-K351-S, E-BC-K892-M). Urine supersaturation with respect to CaOx, which simplifies the estimation of ionic active products of calcium oxalate (AP(CaOx)), was calculated using the index proposed by Tiselius et al.²⁴

Kidney stone formation

Calcium crystals deposited in renal sections were visualised under the NIKON Ci-L Upright polarised light microscope (NIKON, Tokyo, Japan). Two observers blinded to treatment scored all kidney sections on a scale ranging from 0 (no stones) to 4 (extensive stones), and the mean score was reported for each mouse, as described previously.³

RNA isolation and real-time PCR

For RNA isolation, RNA was extracted from kidney samples with REzol (Protech Technology, London, UK, catalogue # PT-KP200CT). RNA was dissolved in nuclease-free water and quantified using a Nanodrop ND 2000 (ThermoFisher Scientific, Waltham, MA, USA). Levels of mRNA were analysed with SYBR green-based real-time quantitative PCR assays (Applied Biosystems, Darmstadt, Germany, catalogue #4368706). Samples of mRNA were analysed with glyceraldehyde-3-phosphate dehydrogenase (Gapdh) and hypoxanthine guanine phosphoribosyl transferase (Hprt) as the reference genes in each reaction. Primer sequences are listed in Table 1.

Table 1.

Primer sequences designed for PCR.

Genes	Species	Forward primer	Reverse primer
KIM-1	Mouse	5′-AAACCAGAGATTCCCACACGTC-3′	5′-GTCGTGGGTCTTCCTGTAGCTG-3′
NLRP3	Mouse	5′- CCTTGGACCAGGTTCAGTGTG -3′	5′- AGAAGAGACCACGGCAGAAGC-3′
IL-1b	Mouse	5′-CAACCAACAAGTGATATTCTCCATG-3′	5′-GATCCACACTCTCCAGCTGCA-3′
IL-6	Mouse	5′-AGGATACCACTCCCAACAGAC-3′	5′-GTGCATCATCGTTGTTCATAC-3′
MCP-1-2	Mouse	5′-CCCACTCACCTGCTGCTACT-3′	5′-TCTGGACCCATTCCTTCTTG-3′
TNF-α	Mouse	5′-CATCTT CTCAAAATTCGAGTGACA A-3′	5′-TGGGAGTAGACAAGGTACAACCC-3′
F4/80 (Emr1)	Mouse	5′-CTTTGGCTATGGGCTTCCAGTC-3′	5′-GCAAGGAGGACAGAGTTTATCGTG-3′
CD68	Mouse	5′-AGCTGCCTGACAAGGGACACT-3′	5′-AGGAGGACCAGGCCAATGAT-3′
mSpp1	Mouse	5′-GGATGAATCTGACGAATCTC-3′	5′-GCATCAGGATACTGTTCATC-3′
TGF-beta1	Mouse	5′-GCAACATGTGGAACTCTACCA-3′	5′-ACGTCAAAAGACAGCCACTCA-3′
Col1a1	Mouse	5′-TCAGAGGCGAAGGCAACAGTC-3′	5′-GCAGGCGGGAGGTCTTGG-3′
Col3a1	Mouse	5′-GACAGATTCTGGTGCAGAGA-3′	5′-CATCAACGACATCTTCAGGAAT-3′
Tfeb	Rat/Mouse	5′-ACAACTTAATTGAGAGAAGACG-3′	5′-AGGTCTTTCTGCATCCTC-3′
lamp1	Mouse	5′-ATTGCAGTTTGGGATGAATG-3′	5′-TTGCACTTGTATGAGTTTCC-3′
Ctsb	Mouse	5′-AGGTGGAGTCTACAATTCTC-3′ -3′	5′-GTGTACCCAAAGTGCTTATC-3′
BECLIN 1	Mouse	5′-TCAGAAGAGTCTCTACAAGG-3′	5′-GAATAAACTCAGAGGGTTGC-3′
CTSF	Mouse	5′-GCCATAAAGAATTTGGGAGG-3′	5′-CCAGGCTGCTATCTTATTTTC-3′
ATG9B	Mouse	5′-CTATTTCACGAATGGTCCAG-3′	5′-GTTGATGAAGCTGATGTAGG-3′
GAPDH	Mouse	5′-TGACGTGCCGCCTGGAGAAA-3′	5′-AGTGTAGCCCAAGATGCCCTT-3′
HPRT	Mouse	5′-AGTGTTGGATACAGGCCAGAC-3′	5′-CGTGATTCAAATCCCTGAAGT-3′

Western blot analysis

The kidney specimens were placed in RIPA buffer containing a protease inhibitor cocktail and homogenised. The tissue lysate was centrifuged, and the protein concentration of the resulting supernatant was determined using a protein assay kit (Bio-Rad Laboratories, Hercules, CA, USA, catalogue #500-0006). Samples were mixed with SDS loading buffer, boiled, electrophoresed in SDS-PAGE gels, and then transferred to PVDF membranes. Membranes were blocked with blocking buffer for 1 h at room temperature and incubated with primary antibodies, listed in Table 2. After washing, the membranes were incubated with horseradish peroxidase-conjugated secondary antibodies. Immunoreactive protein detection was performed with an enhanced chemiluminescence detection system (PerkinElmer, Waltham, MA, USA).

Table 2.

List of antibiotics used in experiments.

Antibody	Source	Brand, cat#
NFkB (C-20)	Rabbit	Santa Cruz, sc-372
PARP	Rabbit	Cell Signalling, #9532
NLRP3	Mouse	Adipogen, AG-20B-0014
Caspase-3	Rabbit	Cell Signalling, #9662
Cleaved-caspase 3	Rabbit	Cell signalling, #9661
Cathepsin B	Mouse	Santa Cruz, sc-365558
Cathepsin D	Mouse	ThermoFisher, MA5-17236
Cathepsin S	Goat	Abcam, ab18822
ATG5	Rabbit	GeneTex, GTX 113309
ATG7	Rabbit	GeneTex, GTX 32459
ACC	Rabbit	Cell signalling, #3676
p-ACC(S79)	Rabbit	Cell signalling, #3661
mTOR	Rabbit	Cell Signalling, #2972
Phospho-mTOR (Ser2448)	Rabbit	Cell Signalling, #2971
P62/SQSTM	Rabbit	MBL, PM045
ASC	Rabbit	Adipogen, AG-25B-0006
LAMP1	Mouse	Santa Cruz, sc-20011
LC3B	Rabbit	Cell Signalling, #3868
BECN1	Rabbit	Cell Signalling, #3738
Total-AMPKα	Rabbit	Cell Signalling, #2603
p-AMPKα(T172)	Rabbit	Cell Signalling, #4188
S6K	Rabbit	Cell Signalling, #9202
p-S6k	Mouse	Cell Signalling, #9206
TFEB	Rabbit	Cell Signalling, #4240S
Phospho-TFEB (Ser211)	Rabbit	Cell Signalling, #37681S
α-tubulin	Mouse	Sigma, SI-T5168
β-actin	Mouse	Sigma, A2228
phospho-ULK1 (Ser555)	Rabbit	Sigma, ABC124
SGLT2	Rabbit	ProteinTech, 24654-1-AP

Immunofluorescent staining and confocal microscopy

The kidney samples were fixed in 4% paraformaldehyde embedded in paraffin and sectioned. After deparaffination, samples were retrieved with citrate buffer (pH 6.0) for 15 min at 80 °C. Then, samples were blocked with 3% BSA blocking buffer for 1 h and incubated with primary antibodies (Table 2) in blocking buffer for overnight at 4 °C, and then incubated with the secondary antibody for 1 h. Images were observed with an Olympus FluoView FV3000 laser scanning confocal microscope (LSCM) (Tokyo, Japan) using FV10-ASW software.

Analysis of a national cohort database

We conducted a retrospective cohort study that employed Taiwan's National Health Insurance Research Database (NHIRD) as a data source. NHIRD was a nation-wide population-based health claims database that contained more than 99% of citizen's healthcare records covered under the single-payer national health insurance program in Taiwan. These medical records included outpatient and inpatient medical diagnoses, procedures and prescriptions that were reimbursed by the government. NHIRD was considered representative in conducting epidemiological studies^25,26 in Taiwan and more details can be found in the review of Hsieh et al.²⁷

In this study, we applied new-user with active comparator design^28,29 and included patients with type 2 diabetes (type 2DM) aged greater than 40 years who initiated SGLT2i or Dipeptidyl peptidase-4 inhibitor (DPP4i) between 2016 and 2020 (i.e., the study period) as the study cohort. A 1-year look back period was used to select new-users. We chose DPP4i as our active comparator because during the study period and as recommended by treatment guidelines³⁰ of type 2 DM, both classes of drugs were considered second-line oral anti-diabetic drugs after metformin monotherapy. We further excluded those prescribed with SGLT2i or DPP4i on the same date, as well as patients with end stage renal disease. In this study, we included patients without prior history of urolithiasis as the first study cohort, while patients with history of urolithiasis as the second study cohort. Urolithiasis was defined by The International Classification of Diseases, Ninth and Tenth Revision, Clinical Modification (ICD-9-CM and ICD-10-CM) diagnosis codes or procedure of shock wave lithotripsy. All relevant codes were listed in Table 3. The first prescription date of SGLT2i or DPP4i during the study period was defined as the index date.

Table 3.

Definitions of exposure, outcome and other covariates.

Item	Codes	Note
Exposure
SGLT2 inhibitor	ATC: A10BK, A10BD15, A10BD16, A10BD19, A10BD20, A10BD21, A10BD23, A10BD24, A10BD25, A10BD27
DPP-4 inhibitor	ATC: A10BH, A10BD07, A10BD08, A10BD09, A10BD10, A10BD11, A10BD12, A10BD13, A10BD18, A10BD22
Outcome
Urolithiasis	ICD-9-CM: 274.11, 592.0, 592.1, 592.9, 594.0, 594.1, 594.2, 594.8, 594.9, 788.0 ICD-10-CM: N200, N201, N202, N209, N210, N211, N218, N219, N22, N23 Domestic procedure codes (including shock wave lithotripsy and other surgery): 50023B, 50024B, 77026B, 77027B, 77028B, 77034B, 76011B, 76012B, 76016B, 76017B, 76023B, 76032B, 77001B, 77002B, 77009B, 78005B, 78024C, 78026C, 78027C, 78201C	Identified from any outpatient and inpatient diagnosis or procedure filed.
Effect modifier
Hydroxychloroquine	ATC: P01BA02	Time spam: 1 year prior to index date.
Exclusion criteria
SGLT2 and DPP4 combinations	ATC: A10BD19, A10BD21, A10BD24, A10BD25, A10BD27
End stage renal disease	ICD-9-CM: 585 ICD-10-CM: N185, N186	Time spam: 1 year prior to index date.
Comorbidities
Dyslipidemia	ICD-9-CM: 272.0–272.4, 272.9 ICD-10-CM: E78.0, E78.1-E78.5	Identified from any outpatient and inpatient diagnosis filed. Time spam: 1 year prior to index date.
Chronic kidney disease	ICD-9-CM: 593.9, 585 ICD-10-CM: N18.1-N18.5
Hyperparathyroidism	ICD-9-CM: 252.0, 252.8, 252.9 ICD-10-CM: E21
Gout	ICD-9-CM: 274.0 ICD-10-CM: M10, M1A
Hypertension	ICD-9-CM: 401-405 ICD-10-CM: I10–I15
Diabetic nephropathy	ICD-9-CM: 250.40, 250.42 ICD-10-CM: E11.2
Diabetic retinopathy	ICD-9-CM: 250.50, 250.52, 362.01, 362.02, 366.41 ICD-10-CM: E11.3
Diabetic neuropathy	ICD-9-CM: 357.2, 250.60, 250.62 ICD-10-CM: E11.4
Systemic lupus erythematosus	ICD-9-CM: 710.0 ICD-10-CM: M32
Sjögren syndrome	ICD-9-CM: 710.2 ICD-10-CM: M35.0
Rheumatoid arthritis	ICD-9-CM: 714.0–714.2 ICD-10-CM: M05
Sarcoidosis	ICD-9-CM: 135 ICD-10-CM: D86
Concurrent medications
Insulin	ATC: A10A	Time spam: 1 year prior to index date.
GLP1RA	ATC: A10BJ, A10AE54, A10AE56
Metformin	ATC: A10BA, A10BD02, A10BD03, A10BD05, A10BD07, A10BD08, A10BD10, A10BD11, A10BD13-18, A10BD20, A10BD22, A10BD23, A10BD25-27
Sulfonylurea	ATC: A10BB, A10BD01, A10BD02, A10BD04, A10BD06
Thiazolidinediones	ATC: A10BG, A10BD03, A10BD04, A10BD05, A10BD06, A10BD09, A10BD12, A10BD26
Other antidiabetic drugs	ATC: A10 (excluding other classes)
Thiazide	ATC: C03A, C07BA, C07BB, C07BG, C07DA, C07DB, C09XA52, C09XA54, C03EA13, C09DX06, C03EA07, C03EA01, C09DX07, C09DX03, C09BX03, C09DX08, C03EA02, C09DX01
Loop diuresis	ATC: C03CA, C03CB
Antigout drug	ATC: M04A
Potassium citrate	Domestic drug codes: A031113121, A036657121, A041577100, A049218121, A052540121, AC31113121, AC36657121, AC41577100, AC49218121, AC52540121, AC58069100, B017881100, B022477100, B024570100, BC24570100

Note. Confounders include variables that are causes of the exposure (dyslipidemia, chronic kidney disease, hypertension, diabetic nephropathy, diabetic retinopathy, diabetic neuropathy, insulin, GLP1RA, metformin, sulfonylurea, thiazolidinediones, and other antidiabetic drugs), the outcome (chronic kidney disease, hyperparathyroidism, gout, hypertension, thiazide, loop diuresis, antigout drugs, and potassium citrate), or both (chronic kidney disease, hypertension, gout, and diabetic nephropathy). We excluded any known instrumental variables (none identified) and included variables that, while not meeting the criterion, serve as good proxies for unmeasured common causes of the exposure and outcome (e.g., systemic lupus erythematosus, Sjögren syndrome, rheumatoid arthritis, and sarcoidosis, reflecting systemic inflammation as a shared risk factor). Furthermore, we controlled for the effect modifier of interest (hydroxychloroquine) and its important causes (systemic lupus erythematosus, Sjögren syndrome, rheumatoid arthritis, and sarcoidosis).

Exposure, outcome, effect modifier and follow-up

Exposure of interest included SGLT2i and DPP4i, which were captured through prescription records by The Anatomical Therapeutic Chemical (ATC) codes; while outcome of interest was incidence of urolithiasis, defined as the same as the exclusion criteria. In addition, the effect modifier was the exposure of HCQ, defined as any prescription within 1 year before the index date. All operational definitions can be found in Table 3. To mimic the intention-to-treat analysis in clinical trial, each patient was followed up from the first prescription date of SGLT2i or DPP4i (i.e., the index date), until the occurrence of outcome, death, or disenrollment of the national health insurance, whichever came first.

Statistics

Numerical data are presented as mean ± SEM in in vivo experiments and mean ± SD in human database analysis. Means between two groups were compared with nonparametric Mann–Whitney U tests. Comparisons among three or more groups were performed by one-way ANOVA followed by the corresponding post-hoc test (Fisher's LSD test or Bonferroni's test). The significance level was set at p value < 0.05 for all tests. Statistical analyses were performed using GraphPad PRISM 9 (version 9.5.1 for macOS).

In this human study using database analysis, we presented patient characteristics through descriptive analysis and compared the SGLT2i and DPP4i groups using standardised mean differences (SMD), with absolute values over 10% indicating significant differences. We calculated the incidence rates of nephrolithiasis for both groups, presented as the number of incidence events per thousand patient-years, and depicted these findings through cumulative incidence plots. Additionally, incidence rate ratios along with their 95% confidence intervals were determined using Poisson regression. To adjust for confounders and enhance comparability, we utilised propensity score (PS) analysis with standardised mortality ratio weighting (SMRW) in Cox proportional hazard regression, estimating the hazard ratio (HR) for nephrolithiasis in the SGLT2i group vs the DPP4i group. The SMRW weighting estimated average treatment effect in the treated (i.e., the SGLT2i group) and assigned was 1 for the SGLT2i group and PS/(1 – PS) for the DPP4i group. Descriptive characteristics of the weights, including mean, standard deviation, and distribution, are provided in Supplementary Table S1. We identified potential confounders and risk factors, such as demographics, diabetes complications, relevant comorbidities, and concurrent medications (detailed in Table 3), through literature reviews and discussions with clinical nephrologists, based on the modified disjunctive cause criterion.³¹ These covariates were incorporated into the logistic regression model for propensity scoring, and the details of the confounder selection were provided in Table 3. Additionally, to address potential bias due to informative censoring from competing risks, such as death, we implemented inverse-probability-of-censoring weighting (IPCW) in our analysis. The censoring proportion for each study cohort was calculated and reported in Supplementary Table S2.

To investigate the effect modification of HCQ, we categorised all patients according to their exposure status of SGLT2i/DPP4i as well as HCQ. We defined the group of DPP4i without prior HCQ exposure as the reference group and calculated the HR of urolithiasis of all the other three groups using Cox regression model with PS SMRW. We also test the interaction term of SGLT2i/DPP4i and HCQ exposure in the regression to see if significant effect modification exists on a multiplicative scale. All analyses were conducted in study cohorts without and with history of urolithiasis, respectively.

To test the robustness of the study results, sensitivity analysis using inverse probability of treatment weighting (IPTW) was conducted to verify the robustness of our results and to estimate the average treatment effect across the total population. Further, to mimic the per-protocol analysis in additional to intention-to-treat, we applied as-treated analysis by adding drug discontinuation and switching to the other drug class as additional censoring points during the follow-up. For all analyses, the statistical significance level was set to be 0.05. All analyses were performed using SAS v9.4 (Cary, NC).

Ethics

The animal care and experimental protocols followed the Guide for the Care and Use of Laboratory Animals, and was reviewed in detail and approved by the Institutional Animal Care and Use Committee (IACUC) of Taiwan National Cheng Kung University (Approval No: 110325, 111106, 112017). We followed the ARRIVE guidelines in every experiment. The national cohort database study was approved by National Cheng Kung University Hospital Institutional Review Board (A-ER-109-567). Written informed consents from patients were waived owing to the retrospective study design, non-invasive nature of the study.

Role of funders

No funder had any role in study design, data collection, analysis, interpretation of data, writing of the report or in the decision to submit for publication.

Results

DAPA suppressed CaOx kidney stone formation

GOX treatment significantly induced weight loss (p = 0.0294, Fisher's LSD test), and DAPA co-treatment attenuated GOX-induced weight loss (Fig. 2a and b and Supplementary Figure S1). Kidney weight was significantly increased in the GOX group (p = 0.0120, Fisher's LSD test) but recovered after DAPA co-treatment (p = 0.0002, Fisher's LSD test) (Fig. 2c). GOX showed the presence of stones with a mean score of 2.375 and an incidence of 100% (Fig. 2d). Surprisingly, co-treatment with DAPA completely reduced GOX-induced kidney stones: severity score: 0 and incidence: 0%. The H&E staining image revealed dilated tubules with crystal deposits along the corticomedullary junction (Fig. 2e, yellow arrow). All of these pathological findings were not seen in the DAPA co-treatment group.

Fig. 2.

Effects of DAPA on kidney stone and biochemistry of urine and serum in GOX-induced CaOx mice (n = 10 in each group). (a) Experimental design of DAPA treatment in GOX-induced CaOx mice. (b) Body weight changes (%). (c) Kidney/body weight ratio. (d) Calcification score. (e) Polarised light microscopic view and H&E of kidney. Scale bar of the upper two panels = 100 μm. Scale bar of the upper two panels = 400 μm. (f) Urine Ca. (g) Urine oxalate. (h) Urine citrate. (i) Urine Mg. (j) Urine K. (k) Urine pH. (l) Urine Na. (m) Urine uric acid. (n) AP (CaOx) score. (o) Plasms BUN. (p) Plasma BUN/Cr. (q) Plasma glucose. Data were expressed as mean ± SEM with statistics analysed by Fisher's LSD test. ∗p < 0.05; ∗∗p < 0.01; and ∗∗∗p < 0.001.

DAPA attenuated CaOx stone formation and impaired renal function

We examined the effects of GOX treatment and DAPA co-treatment on urine and plasma biochemistry (Fig. 2f–m). GOX treatment significantly decreased urine Ca²⁺ (p = 0.0127, Fisher's LSD test) but increased urine Ox by over threefold (p = 0.0049, Fisher's LSD test). It also decreased urine citrate and Mg²⁺, key inhibitors of nephrolithiasis. Additionally, GOX treatment significantly decreased urine K⁺ (p = 0.0290, Fisher's LSD test), likely associated with urinary citrate.³² Acidified urine is also a risk factor of CaOx stone formation, and we also found GOX significantly lowered urine pH (p = 0.0011, Fisher's LSD test). However, GOX treatment did not significantly increase urine Na⁺ and uric acid, which are the risk factors of kidney stones.²⁴ These changes significantly increased the risk of CaOx stone formation, as indicated by the ion activity product for CaOx stone (AP(CaOx)) (p = 0.0227, Fisher's LSD test) (Fig. 2n). DAPA administration had no significant impact on urine Ca²⁺ but effectively reduced urine Ox (p = 0.0021, Fisher's LSD test) and increased urine citrate (p = 0.0084, Fisher's LSD test), thereby lowering CaOx supersaturation, as reflected by decreased AP(CaOx) (p = 0.0232, Fisher's LSD test). DAPA also reversed acidified urine pH (p = 0.0033, Fisher's LSD test), enhancing Ca²⁺ complexation by citrate and further reducing CaOx supersaturation.³³ While DAPA paradoxically increased urinary risk factors such as Na⁺ (p = 0.0033, Fisher's LSD test) and uric acid (p = 0.0130, Fisher's LSD test), this is consistent with the well-known pharmacological effects of SGLT2 inhibitors.¹² Thus, DAPA not only reversed CaOx stone formation histologically but also significantly altered key biochemical parameters related to CaOx stone formation.

We then asked whether these changes are associated with alteration in renal function. The GOX group had the worst renal function, reflected by the highest plasma BUN levels and ratio of plasma BUN to creatinine (CRE). Nevertheless, the administration of DAPA reversed the decline in renal function (Fig. 2o and p). The results of the renal function over different time periods are provided in Supplementary Figure S2. These changes, along with the concurrent change in urine K⁺, indicate that DAPA reversed CaOx stone-impaired renal function. However, all the above changes were not associated with changes in plasma glucose levels (Fig. 2q).

DAPA ameliorated kidney stone-induced inflammation and apoptosis

We examined the effects of DAPA on inflammation and apoptosis caused by nephrolithiasis (Fig. 3 and Supplementary Figure S3). Kim1 gene expression increased in the GOX group but decreased with GOX + DAPA (Fig. 3a). Inflammation markers (Tnfα, Il6, Ccl2), macrophage markers (Emr1, Cd68), and inflammasome mediators (Nlrp3, Il1b) were increased in the GOX group but ameliorated with DAPA (Fig. 3b–h). Spp1, associated with nephrolithiasis risk, increased with GOX and decreased with DAPA (Fig. 3i). Fibrosis-related genes (Tgfb1, Col1a1, Col3a1) increased with GOX and decreased with DAPA (Fig. 3j–l). Immunoblotting also showed increased NFκB, NLRP3, and ASC in the GOX group, reversed by DAPA (Fig. 3m and Supplementary Figure S4a). Apoptosis-related proteins (cleaved Caspase 3 and PARP) were also significantly up-regulated by GOX and down-regulated after DAPA. These findings suggested that inflammation and apoptosis, caused by nephrolithiasis, can be resolved after DAPA co-treatment.

Fig. 3.

Effects of DAPA on kidney injury, inflammation, macrophage, inflammasome, kidney fibrosis pathways, and apoptosis signal in kidney (n = 10 in each group, with glyceraldehyde-3-phosphate dehydrogenase (Gapdh) as the reference gene). Expression of (a) Kim1, (b) Tnfα, (c) Il6, (d) Ccl2, (e) Emr1, (f) Cd68, (g) Nlrp3, (h) Il1b, (i) Spp1, (j) Tgfb1, (k) Col1a1, (l) Col3a1. The results using Hprt as the reference gene were provided in Supplemental Figure S3. (m) Immunoblot analysis on inflammasome- and apoptosis-related proteins. For quantification of immunoblot analysis, the intensities of bands quantified densitometrically relative to the control were provided in Supplementary Figure S5. Data were expressed as mean ± SEM and statistics were calculated by Fisher's LSD test. ∗p < 0.05; ∗∗p < 0.01; and ∗∗∗p < 0.001.

Kidney stone impaired autophagosome degradation, which is reversed by DAPA

We then tested whether CaOx kidney stone formation is related to autophagy (Fig. 4a–c and Supplementary Figure S4b–d). GOX treatment significantly upregulated LC3B-II level and LC3B-II/I ratio, as well as SQSTM1/p62 level (Fig. 4a and Supplementary Figure S4b). These results suggest that GOX induced autophagosome synthesis but did not clear p62. Moreover, downregulated LC3B-II and p62 levels as well as the LC3B-II/I ratio after DAPA administration demonstrate that DAPA can resolve both accumulated LC3B-II and p62 proteins.

Fig. 4.

Effect of DAPA on kidney autophagy (n = 10 in each group). (a) Immunoblot analysis on LC3B and p62, (b) p-ULK1/BECN1/downstream ATG, (c) AMPK/mTOR related proteins. For quantification of immunoblot analysis, the intensities of bands quantified densitometrically relative to the control were provided in Supplementary Figure S5. (d) Immunofluorescence staining for LC3B (red), NLRP3 (green), and SGLT2 (white) in kidney. The DAPI nuclear counterstain appears blue. The original magnification of upper panel was ×20 and lower panel was ×400. The scale bar is 10 μm. Representative images for SGLT2 staining of mouse kidney are provided in Supplementary Figure S4. (e) Immunoblot analysis on lysosome biogenesis-related proteins. (f) Immunoblot analysis on p-TFEB and TFEB. (g) Ubiquitin smear. Expressions of TFEB downstream genes: (h) Tfeb, (i) Lamp1, (j) Lc3b, (k) Ctsf. Data were expressed as mean ± SEM and statistics were calculated by Fisher's LSD test. ∗p < 0.05; ∗∗p < 0.01; and ∗∗∗p < 0.001.

We then examined the upstream pathway for autophagy and found that GOX significantly upregulated phosphorylated (p-) Unc-51-like kinase 1 (ULK1) and Beclin-1 (BECN1), which is essential to induce the formation of autophagosomes from pre-autophagic structures (Fig. 4b and Supplementary Figure S4c). GOX tended to increase the protein level of ATG5, but it did not change ATG7. While DAPA did not significantly alter the levels of ATG5-ATG12 and ATG7, it significantly downregulated p-ULK1 and BECN1 levels. Thus, the downregulations of BECN1, p-ULK1, LC3B-II, and p62 after DAPA administration strongly suggest DAPA improves stone-induced impairment in autophagic flux.

We then examined two predominant upstream signalling kinases, AMPK, a positive driver, and mTOR, a negative driver, for the autophagy pathway (Fig. 4c and Supplementary Figure S4d). GOX upregulated both phosphorylated (p-) AMPK (T172) and AMPK levels. While DAPA co-treatment did not alter p-AMPK level, it significantly increased p-AMPK/AMPK ratio. These results suggest that DAPA further activated AMPK, consequently leading to further induction of autophagy.

For the mTOR pathway, GOX induced both phospho-mTOR (S2448) and mTOR levels, resulting in increased p-mTOR/mTOR ratio, suggestive of mTOR activation by GOX treatment (Fig. 4c and Supplementary Figure S4d). We also found that phosphorylated (p-) p70S6K/S6K (S6 kinase) was activated by GOX, indicating that kidney stone increased mTOR activity. Interestingly, DAPA co-treatment attenuated both increased p-mTOR and p70S6K levels, demonstrating that DAPA co-treatment effectively ameliorated GOX-induced mTOR activation. These results suggest that kidney stone formation is associated with upregulated autophagy synthesis. However, it also upregulated mTOR, which impaired autophagic flux. DAPA co-treatment further activated AMPK and downregulated mTOR protein level, consequently causing induction of autophagy, and resolution of damaged autophagic flux.

The inflammatory site of kidney stone colocalised with autophagy positive area

We applied LC3B to locate autophagosomes (red) and NLRP3 for the stone-associated inflammasome (green) (Fig. 4d and Supplementary Figure S5). Firstly, the levels of SGLT2 (white), a target of DAPA predominantly found in the cortex, did not differ among the three groups. Secondly, NLRP3 levels, activated by crystalline material,³⁴ were significantly increased in the GOX group and reversed with DAPA co-treatment. The NLRP3 signal was concentrated in the medulla, encircled by dilated tubules, closely associated with CaOx crystal deposits. Thirdly, the number of LC3B puncta was elevated in the GOX group, particularly near NLRP3-positive tubules, indicating an accumulation of autophagosomes around deposited CaOx crystals (Fig. 4d). These results revealed that stocked autophagic flux is heightened by GOX and is closely related to CaOx stone formation, and resolved by DAPA administration.

DAPA induced lysosomal biogenesis and resolved lysosomal dysfunction

Because reduced lysosomal function is one of the causes of impaired autophagic flux, we then examined the lysosomal integrity and markers. GOX treatment dramatically decreased lysosomal associated membrane protein 1 (LAMP1) and cathepsins, including cathepsin B and cathepsin D, levels, but DAPA co-treatment reversed their levels (Fig. 4e and Supplementary Figure S4e). Therefore, our findings suggest that kidney stones impede lysosomal integrity and function. We then examined the involvement of the transcription factor EB (TFEB), a master transcription factor that induces the expression of genes involved in autophagy and lysosomal biogenesis. Phosphorylation of TFEB by mTOR inhibits TFEB activity, and p-TFEB is then degraded through a ubiquitin-proteasome system.³⁵ We found that GOX-induced mTOR activation, evidenced by increased p-mTOR and p-S6k levels (Fig. 4c and Supplementary Figure S4d), along with an increase in p-TFEB levels (Fig. 4f and Supplementary Figure S4f). Sha et al. have elucidated that p-TFEB forms inactive heterodimers with non-phosphorylated TFEB and accumulates if ubiquitinated p-TFEB is not effectively degraded by proteasome degradation.³⁶ Therefore, we examined the ubiquitin-proteasome system, and found that GOX led to a general increase in the level of protein ubiquitination (Fig. 4g), indicating the impairment of the proteasome degradation system. In contrast, DAPA treatment inhibited mTOR activity, which was associated with reversal of p-TFEB and p-S6k levels; meanwhile, the nephrolithiasis-mediated increase of ubiquitination was restored following DAPA use. Finally, DAPA significantly upregulated Tfeb, lamp1, Lc3b, and Ctsf genes transcription (p = 0.0025, 0.0002, 0.0003, 0.0468, Fisher's LSD test) (Fig. 4h–k and Supplementary Figure S6), suggesting that DAPA upregulated TFEB activity. These findings suggest that kidney stones impaired lysosomal function/integrity by inactivating TFEB, and DAPA possibly corrected lysosomal function and induced TFEB activity, leading to resolution of stone formation.

Rapamycin exhibited limited inhibitory effects on kidney stones

To determine whether mTOR inhibition is the key to suppress kidney stone formation, we applied the mTOR inhibitor rapamycin (Rapa) (Fig. 1a). Rapa co-treatment reduced body and kidney weight, which indicates an inhibitory effect on stone formation (Fig. 5a and b and Supplementary Figure S7). However, Rapa did not significantly affect CaOx stone deposits by GOX, indicating its limited ability to reverse the GOX-induced crystallisation (Fig. 5c and d). Biochemically, Rapa significantly decreased urine Ox (p = 0.0245, Bonferroni multiple comparison test) (Fig. 5e) but did not reverse acidified urine pH (Fig. 5f), showing limited efficacy in attenuating nephrolithiasis formation. Combining Rapa with DAPA did not enhance DAPA's effects on inhibiting nephrolithiasis formation, reflected by calcification score, urine citrate level, urine pH, and crystal deposits. These results suggest that Rapa possibly exhibits a similar pathway with DAPA. Molecularly, Rapa alone exhibited weaker effects on reducing NLRP3, ASC, and mTOR levels, compared to DAPA alone (Fig. 5g and Supplementary Figure S4g). However, combined treatment of Rapa with DAPA completely reduce NLRP3, ASC, and mTOR. Thus, Rapa exhibits a partial beneficial effect on kidney stone formation, likely through a pathway similar to DAPA.

Fig. 5.

Influence of autophagy modulators, Rapa and HCQ, on the therapeutic effects of DAPA in GOX-induced CaOx mice (In the GOX+Rapa group, n = 7, and in the other groups, n = 13). (a) Body weight changes (%). (b) Kidney/body weight ratio. (c) Calcification score. (d) Polarised light microscopic views of kidney. Scale bar = 100 μm. (e) Urine Ox. (f) Urine pH. (g) Immunoblot analysis on NLRP3, ASC, p-mTOR, and mTOR. For quantification of immunoblot analysis, the intensities of bands quantified densitometrically relative to the control were provided in Supplementary Figure S6. (h) Body weight changes (%). (i) Kidney/body weight ratio. (j) Calcification score. (k) Polarised light microscopic views of kidney. Scale bar = 100 μm. Immunoblot analysis on NLRP3, ASC, p-, mTOR, LC3B I, and LC3B II on (l) GOX with and without HCQ, (m) DAPA with and without HCQ. For quantification of immunoblot analysis, the intensities of bands quantified densitometrically relative to the control were provided in Supplementary Figure S5. Data were expressed as mean ± SEM and statistics were calculated by one-way ANOVA followed by Tukey's post-hoc test. ∗p < 0.05; ∗∗p < 0.01; ∗∗∗p < 0.001; and ∗∗∗∗p < 0.0001.

Inhibition of autophagy with HCQ reversed the inhibitory effect of DAPA on stone formation

To test whether DAPA's benefit is attributed by resolving impaired autophagic flux, we applied HCQ, which blocks autophagosome-lysosome fusion, in the DAPA group (Fig. 1b). HCQ treatment exaggerated GOX-induced stone formation, reflected by body weight reduction and CaOx crystal deposits (Fig. 5h–k and Supplementary Figure S7). Molecularly, HCQ alone exaggerated GOX-induced NLRP3, mTOR, and LC3-II levels (Fig. 5l and Supplementary Figure S4h).

Interestingly, HCQ co-treatment attenuated several beneficial effects of DAPA on kidney stone reversal. It mitigated DAPA's reductions in body and kidney weight (Fig. 5h and i) and abolished the improvements in calcification score and CaOx crystal deposits (Fig. 5j and k). Consistently, HCQ co-treatment attenuated the beneficial effects on reduced NLRP3, ASC, mTOR, and LC3-II levels (Fig. 5m and Supplementary Figure S4i). Thus, HCQ completely wiped out the effectiveness of DAPA.

While HCQ co-treatment tended to increase expression of Kim1, Il1b, Cd68, and Col3a1 it significantly increased expression of Nlrp3, Il6, Ccl2, Tnfα, Emr1, Tgfb1, Col1a1, and Spp1 (p = 0.0317, 0.0382, 0.0159, 0.0317, 0.0159, 0.0317, 0.0159, 0.0159, Mann–Whitney U test) (Fig. 6a–i and Supplementary Figure S8). Therefore, HCQ co-treatment counterbalanced many aspects of DAPA-mediated beneficial effects on kidney stone formation.

Fig. 6.

Impact of HCQ on DAPA on kidney injury, inflammation, macrophage, inflammasome, kidney fibrosis pathways and apoptosis signal in kidney (n = 5 in each group; with Gapdh as the reference gene). Expression of (a) Kim1, (b) Nlrp3, (c) Il1b, (d) Il6, (e) Ccl2, (f) Tnfα, (g) Emr1, (h) Cd68, (i) Spp1, (j) Tgfb1, (k) Col1a1, (l) Col3a1. The results using Hprt as the reference gene were provided in Supplementary Figure S8. Data were expressed as mean ± SEM and statistics were calculated by unpaired two-tailed Mann–Whitney U test. ∗p < 0.05.

SGLT2i reduces incident and recurrent kidney stone events: Evidence from human cohort database

To translate our findings into humans, we utilised nationwide cohort data to investigate nephrolithiasis risk in patients with diabetes using SGLT2i vs DPP4i, a second-line diabetic treatment. We included a total of 535,404 patients, of which 284,295 were male. Baseline characteristics before and after propensity score (PS) weighting are shown in Table 4, while the distribution of PS (SMRW) is presented in Supplementary Figure S9. All covariates were well balanced between groups after PS weighting (absolute SMD < 0.1). After a median (IQR) follow-up of 2.1 (1.0–3.5) years, patients without a history of nephrolithiasis who used SGLT2i had a lower incidence rate of urolithiasis (13.26 per 1000 person-years) compared to DPP4i users (15.41 per 1000 person-years), with an incidence rate ratio (IRR) of 0.86 (95% CI 0.82–0.90) (Fig. 7a and b). As-treated (AT) analysis showed a more pronounced IRR of 0.79 (95% CI 0.75, 0.84) (Fig. 7a and c). The PS SMRW weighted HRs were 0.83 (95% CI 0.78, 0.88) and 0.77 (95% CI 0.71, 0.82) for ITT and AT analysis, respectively. Among those with nephrolithiasis history, incidence rates were 127.24 and 132.41 per 1000 person-years for SGLT2i and DPP4i users, respectively, resulting in an IRR of 0.96 (95% CI 0.93, 0.99) (Fig. 7a and d). IRR in AT analysis was 0.90 (95% CI 0.87, 0.93) (Fig. 7a and e). The PS SMRW weighted HRs were 0.88 (95% CI 0.84, 0.92) and 0.80 (95% CI 0.76, 0.85) for ITT and AT analysis, respectively. Results from PS inverse probability treatment weighting (IPTW) were consistent, detailed in Supplementary Figures S9–S12 and Supplementary Table S3. Additionally, Supplementary Table S4 presents the results of a direct comparison between DAPA and DPP4i, given that DAPA was used as the representative SGLT2i in our in vivo experiments. The findings remained consistent with the primary analysis.

Table 4.

Baseline characteristics for the patients using SGLT2is and DPP4is before and after standardised mortality ratio weighting (SMRW).

Characteristic	Before weighting			After weighting
	SGLT2i (n = 86,922)	DPP4i (n = 448,482)	Absolute SMD	SGLT2i (n = 84,749a)	DPP4i (n = 79,360a)	Absolute SMD
Age, years (Mean ± SD)	59.4 ± 10.7	64.0 ± 12.3	0.40	59.4 ± 10.7	59.3 ± 5.0	0.01
Sex
Male, n (%)	49,631 (57.1)	234,664 (52.3)	0.10	48725 (57.5)	44,261 (55.8)	<0.01
Female, n (%)	37,291 (42.9)	213,818 (47.7)	0.10	36,024 (42.5)	35,099 (44.2)	<0.01
Medical history
Dyslipidemia, n (%)	58,349 (67.1)	267,637 (59.7)	0.16	57,402 (67.7)	51,943 (65.5)	<0.01
CKD, n (%)	3563 (4.1)	25,282 (5.6)	0.07	3438 (4.1)	3292 (4.1)	<0.01
Hyperparathyroidism, n (%)	24 (0.0)	155 (0.0)	0.00	23 (0.0)	22 (0.0)	<0.01
Gouty arthritis, n (%)	6207 (7.1)	33,417 (7.5)	0.01	6044 (7.1)	5645 (7.1)	<0.01
Hypertension, n (%)	54,181 (62.3)	277,363 (61.8)	0.01	52,898 (62.4)	48,581 (61.2)	<0.01
Diabetic nephropathy, n (%)	14,569 (16.8)	75,067 (16.7)	<0.01	14,205 (16.8)	13,170 (16.6)	<0.01
Diabetic retinopathy, n (%)	6280 (7.2)	30,942 (6.9)	0.01	6123 (7.2)	5756 (7.3)	<0.01
Diabetic neuropathy, n (%)	4215 (4.8)	24,080 (5.4)	0.02	4110 (4.8)	3887 (4.9)	<0.01
Systemic lupus erythematosus, n (%)	60 (0.1)	253 (0.1)	0.02	59 (0.1)	55 (0.1)	<0.01
Sjögren syndrome, n (%)	560 (0.6)	3812 (0.8)	0.02	546 (0.6)	513 (0.6)	<0.01
Rheumatoid arthritis, n (%)	296 (0.3)	1947 (0.4)	0.02	289 (0.3)	268 (0.3)	<0.01
Sarcoidosis, n (%)	9 (<0.1)	58 (<0.1)	<0.01	9 (<0.1)	7 (<0.1)	<0.01
Diabetes drugs
Insulin, n (%)	13,732 (15.8)	68,898 (15.4)	0.01	13,265 (15.7)	11,851 (14.9)	<0.01
Metformin, n (%)	72,160 (83.0)	337,498 (75.3)	0.19	70,879 (83.6)	64,791 (81.6)	<0.01
GLP1RA, n (%)	701 (0.8)	669 (0.1)	0.10	701 (0.8)	59 (0.1)	0.01
Sulfonylurea, n (%)	43,086 (49.6)	214,942 (47.9)	0.03	42,026 (49.6)	38,511 (48.5)	<0.01
Thiazolidinedione, n (%)	13,868 (16.0)	44,436 (9.9)	0.18	13,825 (16.3)	9895 (12.5)	0.01
Other OAD, n (%)	10,583 (12.2)	61,064 (13.6)	0.04	10,117 (11.9)	9532 (12.0)	<0.01
Other drugs
Thiazide, n (%)	6058 (7.0)	32,032 (7.1)	0.01	5901 (7.0)	5380 (6.8)	<0.01
Loop diuretic, n (%)	6250 (7.2)	48,074 (10.7)	0.12	5802 (6.8)	5929 (7.5)	<0.01
Antigout drugs, n (%)	8097 (9.3)	45,924 (10.2)	0.03	7845 (9.3)	7308 (9.2)	<0.01
Potassium citrate, n (%)	66 (0.1)	584 (0.1)	0.02	57 (0.1)	65 (0.1)	<0.01
Hydroxychloroquine, n (%)	320 (0.4)	2168 (0.5)	0.02	299 (0.4)	314 (0.4)	<0.01

Abbreviations: CKD, chronic kidney disease; GLP1RA, glucagon-like peptide-1 receptor agonists; OAD, oral anti-diabetic drugs; SD, standard deviation; SMD, standardised mean difference.

^aWeighted N.

Fig. 7.

National database analysis of SGLT2i and nephrolithiasis risk in association with HCQ. (a) Incidence rate ratios of SGLT2i vs DPP4i against nephrolithiasis in patients with/without a previous nephrolithiasis history. Cumulative incidence plots for the comparative risk of nephrolithiasis between cohorts of patients using SGLT2i and DPP4i: (b) patients without a history of nephrolithiasis using ITT analysis, (c) patients with a history of nephrolithiasis using ITT analysis, (d) patients without a history of nephrolithiasis using AT analysis, and (e) patients with a history of nephrolithiasis using AT analysis. (f) Hazard ratios of nephrolithiasis outcomes in patients using SGLT2i and DPP4i, stratified by prior HCQ exposure in patients with or without a history of nephrolithiasis, after propensity score weighting. The reference group for comparison was DPP4i users without prior HCQ exposure.

Exposure to HCQ blocks the effect of SGLT2i in preventing kidney stone events in humans

We further used this population-based cohort to test whether SGLT2i reduce nephrolithiasis risk by restoring autophagy flux. We selected HCQ as an effect modifier due to its common use in inflammatory diseases. Among enrolled patients, 320 SGLT2i initiators and 2168 DPP4i initiators had been prescribed with HCQ within one year before the index date. Baseline characteristics for the patients using SGLT2is and DPP4is after stratification by HCQ exposure are listed in Supplementary Table S5. Adjusting for baseline covariates using PS SMRW and IPCW, we compared DPP4i users never exposed to HCQ as a reference to determine the relative hazard ratio (HR) of nephrolithiasis. Served as a negative control, risk of event was not affected in DPP4i users with prior exposure to HCQ (HR 0.99, 95% CI 0.75–1.30; Table 5 and Fig. 7f). In intent-to-treat analysis (ITT) analysis, SGLT2i use without HCQ was associated with significantly lower incident nephrolithiasis risk (HR 0.83, 95% CI 0.79–0.88); while this protective effect was not significant when HCQ was prescribed with SGLT2i (HR 0.56, 95% CI 0.22–1.43). Although a trend was observed, the modifying effect of HCQ on the association between SGLT2i and DPP4i with nephrolithiasis is not statistically significant (p for interaction in multiplicative scale = 0.440). Similar results were found in AT analysis.

Table 5.

Modification of the effect of SGLT2i and DPP4i on nephrolithiasis by HCQ exposure.

	SGLT2i		DPP4i		HRs (95% CI) for SGLT2i and DPP4i use within strata of HCQ exposure
	N with/without nephrolithiasis	HR (95% CI)	N with/without nephrolithiasis	HR (95% CI)
Incident event (Cohort 1) - ITT analysis
Exposed to HCQ	8/312	0.56 (0.22, 1.43) p = 0.225	81/2087	0.99 (0.75, 1.30) p = 0.930	0.56 (0.21, 1.51) p = 0.253
Not exposed to HCQ	2142/84,460	0.83 (0.79, 0.88) p = <0.0001	15,965/430,349	1.0 (Reference)	0.83 (0.79, 0.88) p = <0.0001
Measure of effect modification on additive scale: RERI (95% CI) = −0.26 (−0.85, 0.33); p = 0.386. Measure of effect modification on multiplicative scale: ratio of HRs (95% CI) = 0.68 (0.25, 1.81); p = 0.440.
Recurrent event (Cohort 2) - ITT analysis
Exposed to HCQ	25/65	1.57 (0.99, 2.51) p = 0.058	149/481	1.06 (0.85, 1.32) p = 0.620	1.49 (0.89, 2.49) p = 0.134
Not exposed to HCQ	41,46/17,942	0.88 (0.84, 0.92) p = <0.0001	24,796/79,087	1.0 (Reference)	0.88 (0.84, 0.92) p = <0.0001
Measure of effect modification on additive scale: RERI (95% CI) = 0.64 (−0.13, 1.41); p = 0.106. Measure of effect modification on multiplicative scale: ratio of HRs (95% CI) = 1.69 (1.01, 2.85); p = 0.047.
Incident event (Cohort 1) - AT analysis
Exposed to HCQ	7/313	0.60 (0.24, 1.52) p = 0.282	56/2112	1.05 (0.75, 1.46) p = 0.770	0.57 (0.21, 1.53) p = 0.266
Not exposed to HCQ	1395/85,207	0.77 (0.72, 0.82) p = <0.0001	10,835/435,479	1.0 (Reference)	0.77 (0.72, 0.82) p = <0.0001
Measure of effect modification on additive scale: RERI (95% CI) = −0.22 (−0.88, 0.44); p = 0.516. Measure of effect modification on multiplicative scale: ratio of HRs (95% CI) = 0.74 (0.28, 2.00); p = 0.557.
Recurrent event (Cohort 2) - AT analysis
Exposed to HCQ	20/70	1.54 (0.86, 2.76) p = 0.145	118/512	0.97 (0.75, 1.25) p = 0.805	1.59 (0.84, 3.01) p = 0.151
Not exposed to HCQ	3266/18,822	0.8 (0.76, 0.84) p = <0.0001	20,086/83,797	1.0 (Reference)	0.8 (0.76, 0.84) p = <0.0001
Measure of effect modification on additive scale: RERI (95% CI) = 0.77 (−0.16, 1.70); p = 0.104. Measure of effect modification on multiplicative scale: ratio of HRs (95% CI) = 1.99 (1.05, 3.76); p = 0.035.

HRs are adjusted using PS SMRW and inverse probability censoring weight. Abbreviations: HCQ, hydroxychloroquine; HR, hazard ratio; PS, propensity score; SMRW, standardised mortality ratio weighting.

Analysis of the second study cohort to those with a history of nephrolithiasis also yielded consistent findings (Fig. 7f), however, with statistical significance for interaction in multiplicative scale (p = 0.470 and 0.035 for ITT and AT analysis, respectively). SGLT2i use reduced recurrent nephrolithiasis risk without HCQ in both ITT (HR 0.88, 95% CI 0.84–0.92) and AT analysis (HR 0.80, 95% CI 0.76–0.84). Conversely, HCQ use with SGLT2i abolished this protective effect (HR 1.57, 95% CI 0.99–2.51 and HR 1.54, 95% CI 0.86–2.76 in ITT and AT analysis, respectively).

Discussion

Our study, using GOX to induce CaOx kidney stones in non-DM mice, showed DAPA completely suppressed stone formation. Despite increasing urine Na⁺ and uric acid, DAPA significantly decreased urine Ox, increased citrate and pH, ultimately inhibiting CaOx stone formation. DAPA also resolved nephrolithiasis-induced renal inflammation, NLRP3 inflammasome, tubular injury, and fibrosis. While CaOx stone increased autophagosome synthesis, it impaired autophagosome degradation, leading to stocked autophagy flux and p62 accumulation. DAPA upregulated autophagy, restored the impaired autophagic flux, and increased lysosomal biogenesis. Furthermore, the clinical data retrieved from the nationwide cohort data confirmed that the use of SGLT2i was associated with a reduced risk of nephrolithiasis than DPP4i. Clamping the autophagosome degradation by HCQ deteriorated the therapeutic effect of DAPA in stone formation both in our animal model and in real-world human participants. Thus, these results suggested that exposure to HCQ reduced protection against incident and recurrent nephrolithiasis benefited from SGLT2i, which warrants further prescription between SGLT2i and autophagy blockers.

Although the pathophysiological mechanism of kidney stone formation is so far unclear,¹ recent evidence suggests that autophagy regulation may play an important role in kidney stone formation. However, different outcomes of autophagic responses to nephrolithiasis have been found.¹⁶ Most studies showed the significant increase of LC3B-II in the kidney of CaOx rat and mouse models,^37,38 suggesting the up-regulation of autophagy in the kidney stone formation. This phenomenon was also supported by in vitro kidney stone study.^37,39 For example, LC3B-II and BECN1 were both significantly increased in a dose and time-dependent manner in HK-2 cells treated by CaOx monohydrate crystals.³⁹ In the present study, the conversion of LC3B-I to LC3B-II was also significantly increased in the kidneys of CaOx mice. Therefore, although autophagy is clearly involved, the causal relationship of autophagy in the kidney stone formation remains to be established.

We noticed that the autophagy state is dynamic, evidenced by LC3B-II and p62 accumulation varying with treatment duration. Short-term (5 days) treatment showed prominent LC3B-II conversion and p62 accumulation (Fig. 1); whereas these increases gradually disappeared in the long-term treatment (11 days, Supplementary Figure S13). Consistent with the existing literature, most in vivo studies have found significantly decreased p62 in the CaOx animal model kidney.^37,38,40 However, Unno et al.,⁴¹ using short-term induction CaOx kidney stone mice, found that p62 gradually accumulated along with a rapid increase in LC3 puncta but a gradually decline over time. These completely opposite results from existing evidence support the dynamic nature of autophagy. Moreover, none of these autophagy modulators could successfully apply to human kidney stone treatment.¹⁶ It is worth mentioning that the conclusion of autophagy status could be also dependent on the animal models. For example, some studies had used hyperoxaluria mice to study autophagy in the pathogenesis of CaOx stone^38,41,42; whereas others used hyperoxaluria rats. The induction of hyperoxaluria in mice takes much shorter time (5–14 days) than in rats (>28 days).^43,44 Thus, the interpretation of autophagy in the mouse model is considered a relatively acute response to nephrolithiasis in ours and the study by Unno et al.,⁴¹ whereas the findings in the rat model—an increase in LC3B-II conversion and a decrease in p62—can be regarded as a relatively chronic response to nephrolithiasis in other studies.^37,42 Thus, the initial autophagic response to kidney stone formation is to enhance autophagosome synthesis but follows a failure of autophagic flux. Whether DAPA could restore autophagic flux and further inhibit the formation of nephrolithiasis prompted us to directly test its causal relationship.

To confirm DAPA's antilithic effects through autophagy, we co-treated hyperoxaluric mice with Rapa and HCQ to assess their impact on kidney stone formation. Rapa shows partial benefits similar to DAPA, likely through shared pathways. In contrast to the findings of Unno et al., who reported that Rapa suppressed renal stone formation in glyoxylate-induced nephrolithiasis model mice,⁴¹ our present study found only partial protective benefits, evidenced by a decrease in urine oxalate. The difference in rapamycin's impact on nephrolithiasis may be related to the dynamic autophagy process during the induction of nephrolithiasis. Notably, other studies have found that rapamycin instead aggravated stone formation.^37,45 HCQ co-treatment significantly increased CaOx deposition, confirmed in NHIRD data, diminishing SGLT2i's protective effects against stones. DAPA's therapeutic effect on human kidney stone reduction was also attenuated by HCQ. This study demonstrating DAPA's antilithic effect in non-diabetic animal models, highlighting autophagy role in DAPA's efficacy, an area beyond DM-focused studies.^17,18

The lysosome is the terminal component of autophagy, and impaired lysosomal function results in blocked autophagosome-lysosome fusion and impeded autophagic flux. We investigated the autophagy lysosomal machinery in the kidney by focussing on the components of lysosomal biogenesis, including LAMP1, CTSB, CTSD, and TFEB, and found that kidney stones possibly impaired lysosomal biogenesis to stock autophagosome degradation. Interestingly, DAPA rescued the damaged lysosome, thereby restoring autophagic flux. Consistently, Unno et al. showed that damaged lysosomes and minimal autolysosomes were observed in the mucosa with plaques from CaOx stone formers.⁴¹ SGLT2i was also found to enhance kidney LAMP1 expression in DM in vivo study.⁴⁶ HCQ mainly blocks autophagy by inhibiting the autophagosome-lysosome fusion and attenuating lysosomal acidification. Thus, lysosome function in nephrolithiasis is restored by DAPA, whereas it is blocked again by HCQ co-administration, finally diminishing the protective effects of DAPA on kidney stone.

The pharmacologic mechanism of SGLT2i on inhibiting stone growth remains unclear till now. Besides modulating autophagy, SGLT2i may alter urinary electrolytes to deter stone formation. Anan et al. used phlorizin, a SGLT1/2 dual inhibitor, on rats with ethylene glycol (EG)-induced renal CaOx stone formation and proposed phlorizin's efficacy by downregulating KIM1 and OPN mRNA and protein expressions.²² First, their findings revealed increased urine Na⁺ and uric acid levels after SGLT1/2 inhibition, consistent with SGLT2i pharmacology.⁴⁷ Secondly, differences emerged in the two main nephrolithiasis promoter Ox; Anan et al. reported no significant change in urine Ox with phlorizin, whereas DAPA significantly decreased urine Ox in our study. Thirdly, decreased urine pH and increased urine citrate have been thought to be the mechanism underlying SGLT2i's effect on kidney stone.⁴⁸ Both studies found decreased urine pH in the kidney stone group, which increased after SGLT inhibition, possibly due to SGLT2i's suppressive effects on Na⁺-H⁺ exchanger 3 (NHE3), leading to increased urinary bicarbonate excretion.⁴⁹ Most importantly, the main urine inhibitors of kidney stone, citrate and Mg²⁺, were significantly increased after DAPA use in our study, but urine Mg²⁺ level was not significantly changed in their study. It is noteworthy to mention that SGLT2i has been shown in earlier studies to increase urine citrate^11,50 and Mg²⁺.^11,51 Our study also utilised AP(CaOx), which combined urine Ca²⁺, Ox, and citrate, to extensively demonstrate that DAPA creates an unfavourable microenvironment to prevent kidney stone formation. Finally, it is important to note a significant increase in uric acid excretion after DAPA use, similar to contrast medium excretion. It is well-known that increased uric acid excretion, acidic urine, and low urine volume are the three major risk factors for uric acid nephrolithiasis. Therefore, whether the use of DAPA will increase the risk of uric acid nephrolithiasis requires further studies to confirm.

One prominent feature of our study is the combination of evidence from animal and human evidence to support a potential biological mechanism for the protective effect of SGLT2i. To the best of our knowledge, there is no other study utilising human cohort data to validate the involvement of autophagy in the protective effect of SGLT2i on nephrolithiasis. From our human study, we found that patients receiving SGLT2 inhibitors were associated with a reduced risk of nephrolithiasis. However, this risk reduction disappeared in those who concomitantly used SGLT2i and HCQ, an autophagy inhibitor. These findings may underscore the role of autophagy in the prevention of nephrolithiasis by SGLT2i, aligning with findings from animal experiments. Moreover, we conducted additional analysis on patients who received DPP4i, but did not find association among DPP4i, HCQ, and the risk of nephrolithiasis. This falsification analysis using DPP4i as negative controls strengthens the causality of the observed association between SGLT2i and the risk of nephrolithiasis, and the potential biological explanation of autophagy in the pathway of nephrolithiasis reduction by SGLT2i. Finally, a notable limitation in our human cohort analysis is the built-in selection bias of hazard ratios in all HR estimations. In order to address this, we tested the proportional hazards assumption using log-negative-log plots (Supplementary Figure S10), and the results indicated that the built-in selection bias is minimal. However, the use of hazard ratios in human studies to report risk estimates may involve a built-in selection bias due to the depletion of susceptibles. The reported hazard ratios can be interpreted as a summary of the treatment effect over the follow-up period.⁵² Nonetheless, we believe this potential bias is unlikely to affect our conclusions, given the use of an active comparator design and the absence of violations of the proportional hazards assumption.

In conclusion, we provide evidence that DAPA can inhibit kidney stone formation in nondiabetic hyperoxaluric mice. The inhibitory effects originate from activation of AMPK, inhibition of mTOR, and restoration of autophagic flux. These results could represent the basis for an important repositioning of the SGLT2i, DAPA, for the treatment of kidney stone disease. Our study underscores SGLT2i as an optimal pharmacotherapy to decrease the risk of nephrolithiasis.

Contributors

CJL conceived the study concept and idea. CJL, YST, and ECCL designed the studies. KTH performed most of the experiments. HSH performed all polarised microscopy analysis. MHCH acquired, cleaned and analysed the data. ZHL provided critical review of the manuscript for important clinical implications. WHW repeated the original experiments and performed the experiments of Fig. 3. YSC reviewed and supervised the initial version of the manuscript. All authors were involved with acquisition, analysis and interpretation of data. CJL analysed and interpreted the data, supervised the project and wrote the entire manuscript. All authors contributed to manuscript revisions and approved the final version of the article. YST is the guarantor of this work.

Data sharing statement

The corresponding authors, Edward Chia-Cheng Lai (edward_lai@mail.ncku.edu.tw) and Yau-Sheng Tsai (yaustsai@mail.ncku.edu.tw), had full access to all of the data in the study and has final responsibility for the decision to submit for publication.

In vivo data collected for the study and data from sample analyses will be made available by the corresponding author, Yau-Sheng Tsai (yaustsai@mail.ncku.edu.tw), upon reasonable request. The data analysed from NHIRD was remotely accessed by the authors from the data centre of the Ministry of Health and Welfare in Taiwan. Researchers interested in accessing this dataset could submit a formal application to the Taiwan Ministry of Health and Welfare to request access (No 488, Sect. 6, Zhongxiao E Rd, Nangang District, Taipei 115, Taiwan; website: https://dep.mohw.gov.tw/DOS/cp-2516-59203-113.html). No additional data available.

Declaration of interests

We received support from the Ministry of Science and Technology of Taiwan (110-2314-B-006-023 and 112-2314-B-006-058 to CJL), (110-2320-B-006-017MY3 and 111-2320-B-006-022MY3 to YST). CJL also received research grants (NCKUH-11202005, -11210020) from the National Cheng Kung University Hospital, Tainan, Taiwan. KTH, HSH, ZHL, MHCH, YSC, WHW, and ECCL have no conflicts of interest.