Abstract Background Diagnosis of early heart failure with preserved ejection fraction (HFpEF) may be challenging because exertional dyspnea is not specific for heart failure, and biomarkers and indicators of volume overload may be absent at rest. We aimed to characterize the contribution of abnormal left atrial (LA) mechanical properties to exercise intolerance in early HFpEF (normal left ventricular filling pressures at rest but elevated during exercise). Methods Diastolic stress testing (DST) was performed in 104 patients with left ventricular ejection fraction ≥50%, in sinus rhythm, and no more than LV diastolic dysfunction grade I, referred for assessment of exertional dyspnoea. Patients exercised supine cycle ergometry at 60 rpm starting with a 3-min period of low-level 25-W workload followed by 25-W increments in 3-minute stages to maximum tolerated levels. According to DST, 43 patients were diagnosed with HFpEF (average mitral E-to-annular e′ ratio [E/e′] > 14, and peak TR velocity >2.8 m/sec at maximal exertion) and 61 as non-cardiac dyspnea (NCD). During the test, two-dimensional images, mitral E/e′, peak tricuspid regurgitation (TR) velociry, and two-dimensional LA mechanical parameters (longitudinal LA strain [LASR] and strain rate [LASRR] during reservoir phase and LA stiffness assessed as a ratio of mitral E/e′ ratio to LASR) were analysed at baseline, and at peak. Results HFpEF and NCD patients were similar in regard to the LA volume index (34.4 [30.2;39.4] vs. 33.6 [28.4;37.1] ml/m2), and NT-proBNP level (132 [80;238] vs. 129 [80;197] pg/ml). As compared with NCD patients, HFpEF patients displayed reduced LA reservoir function assessed by LASR (22.3 [18.9;25.6] vs. 24.2 [21.2;29.8] % at rest, and 25.3 [21.4;30.2] vs. 29.0 [24.2;33.3] % with exercise) and LASRR (0.78 [0.58;0.96] vs. 0.90 [0.68;1.12] /s at rest, and 1.10 [0.79;1.31] vs. 1.24 [1.03;1.56] s–1 with exercise) with increased LA stiffness (0.57 [0.44;0.70] vs. 0.42 [0.30;0.49] mmHg/% at rest, and 0.61 [0.46;0.74] vs. 0.40 [0.32;0.51] mmHg/% with exercise, all P < 0.05). Additionally, HFpEF patients showed smaller exercise elevation in LASRR (+31 [-5;77] vs. +47 [12;85] % as compared with resting values, P < 0.05). Exercised LA stiffness and reservoir strain correlated with exercise LV filling pressures estimated by mitral E/e′ ratio (r = 0.72 and r =–0.35, P < 0.001). LA stiffness showed a good diagnostic accuracy (area under the curve 0.75), and LA stiffness > 0.46 mmHg/% demonstrated reasonable sensitivity (79%) and specificity (71%) to diagnose HFpEF. Neither LV global longitudinal strain and ejection fraction at rest nor their exercise-induced elevation differed between HFpEF and NCD. Conclusion Impaired LA reservoir function and increased stiffness are associated with exercise intolerance in patients with early HFpEF, while LV systolic function seems preserved in this stage of the disease. LA stiffness provides HFpEF diagnostic potential in ambulatory patients with dyspnea