To solve the given optimization problem, we first need to translate the natural language description into a symbolic representation. This involves defining variables and formulating the objective function and constraints using these variables.

Let's define:
- $x_1$ as the number of bags of type A used,
- $x_2$ as the number of bags of type B used.

The objective is to minimize the total cost, which can be represented by the objective function: $4x_1 + 6x_2$.

The constraints are:
1. The mixture must contain at least 20 units of sardines: $2x_1 + 4x_2 \geq 20$.
2. The mixture must contain at least 25 units of anchovies: $3x_1 + x_2 \geq 25$.
3. Non-negativity constraints: $x_1 \geq 0$, $x_2 \geq 0$, since we cannot use a negative number of bags.

Thus, the symbolic representation of the problem is:
```json
{
    'sym_variables': [('x1', 'number of bags of type A'), ('x2', 'number of bags of type B')],
    'objective_function': '4*x1 + 6*x2',
    'constraints': ['2*x1 + 4*x2 >= 20', '3*x1 + x2 >= 25', 'x1 >= 0', 'x2 >= 0']
}
```

To solve this linear programming problem using Gurobi in Python, we will use the following code:

```python
from gurobipy import *

# Create a new model
model = Model("Aquarium_Feeding")

# Define variables
x1 = model.addVar(vtype=GRB.CONTINUOUS, name="bags_A")
x2 = model.addVar(vtype=GRB.CONTINUOUS, name="bags_B")

# Set the objective function
model.setObjective(4*x1 + 6*x2, GRB.MINIMIZE)

# Add constraints
model.addConstr(2*x1 + 4*x2 >= 20, "sardines_constraint")
model.addConstr(3*x1 + x2 >= 25, "anchovies_constraint")

# Optimize the model
model.optimize()

# Print the results
if model.status == GRB.OPTIMAL:
    print("Optimal solution found:")
    print(f"Number of bags of type A: {x1.x}")
    print(f"Number of bags of type B: {x2.x}")
    print(f"Total cost: {model.objVal}")
else:
    print("No optimal solution found")
```