diff --git a/tests/test_optimizers.py b/tests/test_optimizers.py
index decebe76e..0acb9835a 100644
--- a/tests/test_optimizers.py
+++ b/tests/test_optimizers.py
@@ -907,3 +907,57 @@ def energy_converged(current_energy: torch.Tensor, prev_energy: torch.Tensor) ->
             f"Energy for system {step} doesn't match position only optimization: "
             f"system={energy_unit_cell}, individual={individual_energies_fire[step]}"
         )
+
+
+# Test for charge and spin preservation
+# GitHub Issue https://github.com/TorchSim/torch-sim/issues/389
+@pytest.mark.parametrize(
+    ("optimizer_fn", "cell_filter"),
+    [
+        (ts.Optimizer.fire, None),
+        (ts.Optimizer.gradient_descent, None),
+        (ts.Optimizer.fire, ts.CellFilter.unit),
+        (ts.Optimizer.gradient_descent, ts.CellFilter.frechet),
+    ],
+)
+def test_optimizer_preserves_charge_spin(
+    optimizer_fn: ts.Optimizer,
+    cell_filter: ts.CellFilter | None,
+    ar_supercell_sim_state: SimState,
+    lj_model: ModelInterface,
+) -> None:
+    """Test that optimizers preserve charge and spin through initialization and steps."""
+    # Add perturbation to positions for meaningful optimization
+    ar_supercell_sim_state.positions = (
+        ar_supercell_sim_state.positions
+        + torch.randn_like(ar_supercell_sim_state.positions) * 0.1
+    )
+
+    # Set non-zero charge and spin values
+    original_charge = torch.tensor(
+        [5.0], device=ar_supercell_sim_state.device, dtype=ar_supercell_sim_state.dtype
+    )
+    original_spin = torch.tensor(
+        [6.0], device=ar_supercell_sim_state.device, dtype=ar_supercell_sim_state.dtype
+    )
+    ar_supercell_sim_state.charge = original_charge.clone()
+    ar_supercell_sim_state.spin = original_spin.clone()
+
+    init_fn, step_fn = ts.OPTIM_REGISTRY[optimizer_fn]
+    opt_state = init_fn(
+        model=lj_model, state=ar_supercell_sim_state, cell_filter=cell_filter
+    )
+
+    # Verify after initialization
+    assert torch.allclose(opt_state.charge, original_charge)
+    assert torch.allclose(opt_state.spin, original_spin)
+
+    # Run several optimization steps and verify preservation
+    for _ in range(3):
+        if optimizer_fn == ts.Optimizer.fire:
+            opt_state = step_fn(state=opt_state, model=lj_model, dt_max=0.3)
+        else:
+            opt_state = step_fn(state=opt_state, model=lj_model, pos_lr=0.01, cell_lr=0.1)
+
+        assert torch.allclose(opt_state.charge, original_charge)
+        assert torch.allclose(opt_state.spin, original_spin)
diff --git a/torch_sim/optimizers/fire.py b/torch_sim/optimizers/fire.py
index 5583cc0c6..e69a4955f 100644
--- a/torch_sim/optimizers/fire.py
+++ b/torch_sim/optimizers/fire.py
@@ -82,6 +82,8 @@ def fire_init(
         "system_idx": state.system_idx.clone(),
         "_constraints": state.constraints,
         "pbc": state.pbc,
+        "charge": state.charge.clone(),
+        "spin": state.spin.clone(),
         # Optimization state
         "forces": forces,
         "energy": energy,
diff --git a/torch_sim/optimizers/gradient_descent.py b/torch_sim/optimizers/gradient_descent.py
index d6bf52f57..283fe7c80 100644
--- a/torch_sim/optimizers/gradient_descent.py
+++ b/torch_sim/optimizers/gradient_descent.py
@@ -62,6 +62,8 @@ def gradient_descent_init(
         "atomic_numbers": state.atomic_numbers,
         "system_idx": state.system_idx,
         "_constraints": state.constraints,
+        "charge": state.charge,
+        "spin": state.spin,
     }
 
     if cell_filter is not None:  # Create cell optimization state